Original Research
Human Blood Monocyte Subsets
A New Gating Strategy Defined Using Cell Surface Markers Identified
by Mass Cytometry
Graham D. Thomas, Anouk A.J. Hamers, Catherine Nakao, Paola Marcovecchio, Angela M. Taylor,
Chantel McSkimming, Anh Tram Nguyen, Coleen A. McNamara, Catherine C. Hedrick
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Objective—Human monocyte subsets are defined as classical (CD14++CD16−), intermediate (CD14++CD16+), and
nonclassical (CD14+CD16+). Alterations in monocyte subset frequencies are associated with clinical outcomes, including
cardiovascular disease, in which circulating intermediate monocytes independently predict cardiovascular events.
However, delineating mechanisms of monocyte function is hampered by inconsistent results among studies.
Approach and Results—We use cytometry by time-of-flight mass cytometry to profile human monocytes using a panel of 36
cell surface markers. Using the dimensionality reduction approach visual interactive stochastic neighbor embedding, we
define monocytes by incorporating all cell surface markers simultaneously. Using visual interactive stochastic neighbor
embedding, we find that although classical monocytes are defined with high purity using CD14 and CD16, intermediate
and nonclassical monocytes defined using CD14 and CD16 alone are frequently contaminated, with average intermediate
and nonclassical monocyte purity of ≈86.0% and 87.2%, respectively. To improve the monocyte purity, we devised a
new gating scheme that takes advantage of the shared coexpression of cell surface markers on each subset. In addition
to CD14 and CD16, CCR2, CD36, HLA-DR, and CD11c are the most informative markers that discriminate among
the 3 monocyte populations. Using these additional markers as filters, our revised gating scheme increases the purity
of both intermediate and nonclassical monocyte subsets to 98.8% and 99.1%, respectively. We demonstrate the use of
this new gating scheme using conventional flow cytometry of peripheral blood mononuclear cells from subjects with
cardiovascular disease.
Conclusions—Using cytometry by time-of-flight mass cytometry, we have identified a small panel of surface markers that
can significantly improve monocyte subset identification and purity in flow cytometry. Such a revised gating scheme will
be useful for clinical studies of monocyte function in human cardiovascular disease. (Arterioscler Thromb Vasc Biol.
2017;37:00-00. DOI: 10.1161/ATVBAHA.117.309145.)
Key Words: autoimmunity ◼ dengue ◼ flow cytometry ◼ inflammation ◼ monocytes
M
onocytes are a major component of human peripheral
blood, accounting for ≈10% of all circulating leukocytes. Human monocytes are traditionally divided into 3
phenotypically and functionally distinct populations based
on differences in expression of CD14 and CD16 encoding
for the lipopolysaccharide receptor and the low-affinity FC γ
receptor, respectively. Classical (CD14hiCD16neg) monocytes
account for 80% to 90% of human blood monocytes, intermediate (CD14hiCD16hi) monocytes comprise ≈2% to 5%, and
the nonclassical (CD14lowCD16hi) monocytes account for the
remaining 2% to 10%.1,2 Multiple diseases, including bacterial
and viral infections, autoimmunity, and chronic inflammation,
are associated with changes in monocyte subsets. Intermediate
monocytes are more abundant in bacterial sepsis,3 dengue
fever,4 Crohn’s disease,5 cardiovascular disease (CAD),6 and
rheumatoid arthritis,7 whereas the nonclassical monocytes are
more prevalent in periodontitis8 and reduced in stroke.9 There
has been considerable interest surrounding the precise contributions of the various monocyte subsets to human disease.2
Yet, there is still uncertainty concerning the relative contribution of each subset in disease, in part due to conflicting reports
concerning the functions of each subset. Monocytes are typically gated in flow cytometry using either a quadrant-based or
trapezoid-based gating scheme using CD14 and CD16 to distinguish among the 3 monocyte subsets.10,11 Regardless of the
approach, the placement of gates discriminating among the
monocyte subpopulations is an inherently subjective exercise.
Genome-wide analyses and cytometric profiling studies have
identified a repertoire of cell surface markers that are differentially expressed on the 3 monocyte subsets, including CCR2,
Received on: February 2, 2017; final version accepted on: May 1, 2017.
From the Division of Inflammation Biology, La Jolla Institute for Allergy and Immunology, CA (G.D.T., A.A.J.H., C.N., P.M., C.C.H.); and Division of
Cardiology and Robert M. Berne Cardiovascular Center, University of Virginia, Charlottesville (A.M.T., C.M., A.T.N., C.A.M.).
This manuscript was sent to Kathryn Moore, Consulting Editor, for review by expert referees, editorial decision, and final disposition.
The online-only Data Supplement is available with this article at http://atvb.ahajournals.org/lookup/suppl/doi:10.1161/ATVBAHA.117.309145/-/DC1.
Correspondence to Graham D. Thomas, PhD, or Catherine C. Hedrick, PhD, Division of Inflammation Biology, La Jolla Institute for Allergy and
Immunology, 9420 Athena Cir, La Jolla, CA 92037. E-mail [email protected] or [email protected]
© 2017 American Heart Association, Inc.
Arterioscler Thromb Vasc Biol is available at http://atvb.ahajournals.org
1
DOI: 10.1161/ATVBAHA.117.309145
2 Arterioscler Thromb Vasc Biol July 2017
Nonstandard Abbreviations and Acronyms
CAD
CyTOF
DI
viSNE
cardiovascular disease
cytometry by time-of-flight
discrimination index
visual interactive stochastic neighbor embedding
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
CD36, CD64, CD62L, HLA-DR, CX3CR1, SLAN, and
CD11c.2,12,13 One approach to minimize the inconsistencies
that arise from the arbitrary gate placement using CD14 and
CD16 alone is to combine multiple informative cell surface
markers.14,15 However, technical limitations in flow cytometry
have limited the number of cell surface markers that can be
accurately analyzed simultaneously. Furthermore, there is a
lack of analytic approaches that can effectively demonstrate
the improvement provided by a given gating scheme.
The advent of mass cytometry (cytometry by time-of-flight
[CyTOF]) and the development of flow cytometers capable of
detecting in excess of 25 colors have provided the technical
means to overcome these challenges. CyTOF uses heavy metal–
conjugated antibodies in tandem with time-of-flight mass
spectrometry to measure protein abundances at a single-cell
resolution.16 The discrete nature of heavy metal mass to charge
ratios overcomes the limits imposed by fluorescence spectral
overlap in flow cytometry,16,17 which dramatically increases the
number of markers that can be measured simultaneously. This
huge increase in dimensionality has precipitated an analytic
burden rendering the manual, iterative, approach to gating and
cell subset classification ineffective.18 Computational methods
that rely on dimensionality reduction are a powerful solution
to these issues. One such approach, termed visual interactive
stochastic neighbor embedding (viSNE), provides a single-cell
visual representation of cell–cell relationships using a t-distributed stochastic neighbor embedding algorithm.18,19 The resulting plot generated by viSNE positions cells on a 2-dimensional
scatter plot using information borrowed from every marker
used in an experiment.
We reasoned that using CyTOF and viSNE would provide
a powerful approach to phenotype the human monocyte subsets. Here, we use a panel of 36 cell surface markers expressed
on myeloid cells to profile human blood monocytes, confirming previous reported surface marker expression profiles.
Importantly, using viSNE, we find that monocytes gated using
CD14 and CD16 alone lead to frequent misclassification and
cross-contamination between populations. Using a factorial
analysis approach, we identify the cell surface markers that
best discriminate between the monocyte subsets. Furthermore,
we introduce a revised gating scheme for flow cytometry with
considerably lower intersubset contamination to facilitate
high-purity monocyte subset cell preparations for bulk analyses. We also show that flow cytometry experiment analyzed
using multiple redundant markers is an attractive approach to
ensure accurate and reproducible identification of monocyte
subsets in variable human assays.
Materials and Methods
Materials and Methods are available in the online-only Data
Supplement.
Results
Traditional Monocyte Gating Leads to Frequent
Misclassification of Monocyte Subsets
We performed CyTOF mass cytometry on human peripheral
blood in 6 healthy volunteers using a panel of 36 cell surface
markers chosen to interrogate the monocyte system (Table I
in the online-only Data Supplement). We reasoned that the 3
principal monocyte subsets would be more accurately defined
using the full panel of cell surface markers rather than the traditional approach based solely on CD14 and CD16. Monocyte
subsets were defined in the conventional manner by excluding lineage-positive cells (CD66b, CD19, CD3) followed by
selection of monocytes using CD14 and HLA-DR expression. Classical, intermediate, and nonclassical monocytes
constituted 68.3%, 5.7%, and 20.2% of total blood monocytes, respectively (Figure 1A; Figure IA in the online-only
Data Supplement). viSNE was performed on total peripheral
blood mononuclear cells using all cell surface markers. The 3
monocyte subsets were superimposed on the resulting viSNE
map leading to the identification of 3 subpopulations corresponding to each monocyte subset (Figure 1B; Figure IB in
the online-only Data Supplement). This shows that intermediate and nonclassical populations occupied fixed coordinates
on the viSNE map between individuals, whereas the classical
monocytes were more heterogeneously distributed between
individuals, implying heterogeneity within human classical
monocytes as has been observed in mouse.20
To define the monocyte subsets in an unbiased fashion, we
adopted a semisupervised approach. The viSNE results from
6 individuals were combined onto a single plot, and monocyte subset gates were then manually drawn on the viSNE
plot to classify each monocyte population using information
borrowed over all 36 cell surface markers. It was clear that
there was not a distinct separation of the 3 monocyte subsets
(Figure 1C). We quantified the number of each monocyte subset within these newly defined gates and compared them to the
conventional gating strategy (Figure 1D). Using the viSNEbased gating, we observed monocyte frequencies that were
consistent with previous expectations and significantly higher
than we observed with the conventional gating approach, presumably because of the strict lineage gate used in our initial
subset identification. Thus, viSNE-based gating of cell populations is an effective strategy for removing lineage-positive
cells that avoids under-representing cells of interest that may
otherwise be lost. It was apparent from this viSNE-based gating approach that monocytes which were classified as intermediate using the conventional approach were present within
all 3 viSNE-defined gates, and a similar trend was seen for
leakage of classical and nonclassical monocytes into the intermediate monocyte gate (Figure 1C). This led us to speculate
that monocytes that are conventionally classified as 1 subset
may actually be mixed, an assertion that is not surprising, considering the arbitrary nature of gate placement.
To investigate this proposition, monocytes from each subset defined using viSNE were plotted onto a CD14 versus
CD16 biaxial plot (Figure 1E). Classical monocytes predominantly resided within the classical monocyte gate, although a
significant fraction was found to contain low levels of CD14
Thomas et al CyTOF and Monocyte Subsets 3
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Figure 1. Human monocyte gating. A, Traditional gating of classical (CD14++CD16−), intermediate (CD14++, CD16+), and nonclassical
(CD14+CD16+) monocytes. Monocytes are previously gated on Live, Lin (CD3, CD19, CD66b)−, HLA−DR+CD14+ cells. B, Overlay of classical, intermediate, and nonclassical monocytes on a visual interactive stochastic neighbor embedding (viSNE) map of all peripheral
blood mononuclear cell (PBMC) acquired in our experiment. C, viSNE-based gating of monocyte subsets. Classical, intermediate, and
nonclassical monocytes from 6 individuals superimposed on a single biaxial plot showing the red region in B is shown. D, Peripheral
blood monocyte frequencies defined using conventional gating or with gates drawn on the viSNE axis. E, The relative distribution of
viSNE-defined classical, intermediate, and nonclassical monocytes within the conventional monocyte subset gates. F, Quantification
of the distribution of viSNE-defined monocytes in conventional monocyte gates. t-SNE indicates t-distributed stochastic neighbor
embedding.
4 Arterioscler Thromb Vasc Biol July 2017
and was excluded from conventional monocyte subset gating.
Some classical monocytes were also found within the intermediate monocyte gate. viSNE-defined intermediate monocytes
were centered on the conventional intermediate monocyte
gate, although many of these cells were distributed among
all 3 of the conventional gates. Nonclassical monocytes were
principally found within the expected gate; however, these
cells were also found to contaminate the intermediate monocyte population, and some cells were excluded from conventional monocyte subset gating.
These misclassification events equate to contamination
that can bias or lead to the misinterpretation of cell subset function based on bulk measurements, such as RT-PCR,
RNA-Seq, or ELISA. We quantified the contamination that
occurs with conventional gating using monocyte gates defined
by viSNE to determine the true population identity of each
monocyte formerly gated using the conventional approach
(Figure 1F). Classical monocytes were well defined with
>99% of conventionally gated classical monocytes falling
within the viSNE-defined classical monocyte gate. Strikingly,
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Figure 2. Automatic gating of human monocytes. A, visual interactive stochastic neighbor embedding (viSNE) analysis of human monocyte subsets. A total of 1000 of each classical, intermediate, and nonclassical monocytes defined by viSNE were subset and clustered
using all 33 conjugated antibodies. B, Principal component analysis of the 3 monocytes using the 11 most informative cell surface markers defined as the 6 top markers for each pairwise comparison between subsets based on the discrimination index. C, Principal component loading plot for B showing the relative contribution of each surface marker to each principal component. D, Isomap plots showing
expression profiles of the most informative cell surface markers (defined in Table) during the monocyte developmental cascade. t-SNE
indicates t-distributed stochastic neighbor embedding.
Thomas et al CyTOF and Monocyte Subsets 5
intermediate and nonclassical monocytes were only found
to contain 86% (range 78.8%–93.4%) and 87.2% (range
84.2%–94.3%) of the respective subsets as defined by viSNE
(Figure 1F). Intermediate monocytes were principally contaminated with nonclassical monocytes, and to a lesser extent,
also with classical monocytes. We found that conventionally
gated nonclassical monocytes typically contained 5% to 10%
intermediate monocytes according to viSNE. These findings
show that conventional monocyte gating leads to frequent
intersubset contamination. Importantly, these false-positive
events cannot be identified or accounted for using the conventional gating approaches.
Prioritization of Cell Surface Markers That
Discriminate Human Monocyte Subsets
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
One approach to overcome the issue of intersubset contamination is to use additional cell surface markers that, in combination, will increase the specificity of monocyte subset
identification. To identify the marker pairs that would most
optimally work in combination for this purpose, we analyzed
cell surface marker profiles of classical, intermediate, and
nonclassical monocytes. viSNE analysis performed on an
equal number of cells within each monocyte subset showed
a good separation of the 3 populations (Figure 2A). However,
viSNE is a highly nonlinear approach that is extremely powerful for classifying populations, yet with this approach, it is
not possible to directly determine the specific contribution of
individual surface markers to the classification.
To determine these most informative markers, we calculated a discrimination index (DI) for each surface marker
between each pair of monocyte populations (Table I in the
online-only Data Supplement). The DI is calculated as the difference in the mean signal intensity divided by the sum of the
SDs for each marker between a pair of populations and is akin
to the staining index commonly used to determine the brightness of fluorochrome-conjugated antibodies in flow cytometry.21 The 6 markers with the highest DI between each pair
of monocyte subsets were chosen, providing a shortlist of a
total of 11 informative surface markers (Table). To validate
Table. Discrimination Index of Markers
Marker
Class_v_
NonClass
Class_v_Int
NonClass_v_Int
CD16
3.211
2.611
0.300
CD14
2.395
0.350
1.714
CCR2
1.970
0.400
1.553
CD64
1.935
0.469
1.197
CD36
1.415
0.000
1.505
CD62L
1.336
0.926
0.292
CD163
1.280
0.206
1.088
HLA-DR
0.398
1.248
0.914
CD11c
0.600
0.935
0.356
CD40
0.276
0.831
0.555
CD86
0.425
0.603
0.155
the potential of these 11 markers to discriminate between the
monocyte subsets, we performed principal component analysis on viSNE-defined classical, intermediate, and nonclassical monocytes (Figure 2B). In combination, these 11 markers
explained a considerable amount of variability in the data
set, with the primary principal component axis explaining
40.3% of the overall variance and the second principal component accounting for an additional 28.3% of the variance.
Importantly, the 3 monocyte subsets, although overlapping,
did occupy distinct regions of the PCA. Analysis of the principal component loadings was used to identify the markers
responsible for the variability captured in the primary and
secondary principal components. The primary principal component axis was dominated by CD16, CD14, CCR2, CD64,
CD36, and to a lesser extent CD62L and CD86 (Figure 2C),
all of which have high DI values for the comparison between
classical and nonclassical monocytes (Table). The secondary
principal component axis was explained by HLA-DR, CD163,
CD40, CD11c, and CD62L (Figure 2C). Markers dominating
the second principal component axis were found to discriminate between the intermediate and classical populations based
on their DI scores (Table).
Thus, we have found that cell surface markers discriminating among the 3 monocyte populations have a tendency to
separate the intermediate and classical monocytes or the intermediate and nonclassical monocytes, but rarely both (Table).
The main exception to this is HLA-DR which was found to
possess high DI values with respect to both the comparisons
between both classical and intermediate and nonclassical and
intermediate populations. To visualize transitional cell surface
marker expression between the monocyte subsets, we used
isometric mapping to order clusters of monocyte subsets.20,22
Isometric mapping positioned clusters composed of classical,
intermediate, and nonclassical monocytes in the anticipated
order over the primary isometric mapping axis (Figure 2D;
Figure IIA and IIB in the online-only Data Supplement).
We shortlisted the 6 most informative markers (highlighted
in blue, Table) that discriminate the 3 monocyte subsets and
plotted the expression of each against the isometric mapping–ordered clusters (Figure 2D; Figure IIC in the onlineonly Data Supplement). This clearly shows the retainment
of CD14, CCR2, and CD36 on both classical and intermediate monocytes and the loss of these markers on nonclassical
monocytes. Conversely, HLA-DR, CD11c, and CD16 were
expressed more highly on the intermediate and nonclassical
monocytes relative to the classical monocytes, with HLA-DR
being highest on the intermediate subset (Figure 2D).
Rational Improvement of Monocyte Subset Gating
By quantifying the purity of monocyte subsets gated using
the conventional CD14/CD16 method, we have determined
that intermediate and nonclassical monocytes are frequently
misclassified. It has been proposed that leveraging the shared
coexpression of cell surface markers can improve the accuracy
of monocyte subset identification.15 To develop this approach,
we must rationally maximize the accuracy of monocyte sorting while minimizing the requirement for additional surface
markers needed to ensure compatibility with conventional flow
cytometry. Therefore, we incorporated additional markers into
6 Arterioscler Thromb Vasc Biol July 2017
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Figure 3. Revised gating scheme for human monocyte subsets. A, Revised gating scheme for monocyte subsets. Conventionally gated
classical monocytes are subjected to a second gating step, defined as CD11clowHLA-DRhi. Conventionally gated nonclassical monocytes
are subjected to a second gating step, defined as CD36lowCCR2low. Finally, conventionally gated intermediate monocytes are first subject to a gating step to filter contaminating nonclassical monocytes and are thus defined as CD36hiCCR2hi, then to a final gating step to
remove contaminating classical monocytes using a CD11chiHLA-DRlow gate. B, Monocyte subsets gated using either the conventional (A,
left) approach or our revised approach (A, right) overlaid on to the monocyte subset viSNE map. C, Monocyte subset purity assessed by
measuring the frequency of gated cells within viSNE-defined monocyte subset gates in B. D, Recovery of monocyte subsets gated using
the additional gating steps showing only a minor loss of total monocytes with the additional gating. t-SNE indicates t-distributed stochastic neighbor embedding.
Thomas et al CyTOF and Monocyte Subsets 7
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
the conventional gating scheme such that a single extra gating step defines the classical and nonclassical monocytes.
For this step, we use the 2 highest scoring DI markers after
CD14 and CD16. Thus, classical monocytes were defined
as CD14hiCD16lowHLA-DRlowCD11clow, and nonclassical
monocytes were defined as CD14lowCD16hiCD36lowCCR2low
(Figure 3A). Because intermediate monocytes are frequently
contaminated by both classical and nonclassical monocytes,
the inverse of these above gating steps is introduced to filter out contaminants belonging to either population, leading
to a revised definition of CD14hiCD16hiCD36hiCCR2hiHLADRhiCD11chi for intermediate monocytes (Figure 3A). To test
the efficacy of this sorting scheme, we asked what percentage of the monocyte subsets sorted using our new strategy
fall within the viSNE-defined monocyte subset gates. The
mean purity of the classical monocyte subset increased from
99.2% to 99.9%, intermediate monocyte purity increased from
86.2% to 98.7%, and nonclassical monocyte purity increased
from 87.2% to 99.1% (Figure 3B and 3C). We also considered loss of monocytes belonging to each subset as a result
of this revised gating approach. Relative to conventional gating, our revised strategy recovered 75.1% of the classical
monocytes, 67.9% of the intermediate monocytes, and 78.4%
of the nonclassical monocytes (Figure 3D). After accounting
for removal of the misclassified monocytes, this equates to a
loss of 24.1% of the classical, 18.3% of the intermediate, and
8.8% of the classical monocytes, respectively. In summary,
the incorporation of additional gating steps can considerably
improve the classification and purity of monocyte subsets by
excluding false-positive results at the cost of a relatively small
loss of true-positive monocytes.
Revised Gating Approach Improves Subset
Definition in Conventional Flow Cytometry
We tested the application of our new gating method devised
using CyTOF with conventional multicolor flow cytometry.
Human peripheral blood was stained with a panel of 8 cell
surface markers (7 fluorochrome-conjugated antibodies and a
viability dye: CD14, CD16, CD11c, HLA-DR, CD36, CCR2,
and a lineage cocktail composed of CD3, CD19, CD66b
CD56). Consistent with our CyTOF data, classical monocytes
contained the highest expression of CD14, CD36, and CCR2,
intermediate monocytes expressed the highest level HLA-DR
and high levels of CD14, CD16, CD11c and CD36, whereas
nonclassical monocytes highly expressed CD16 and CD11c
with somewhat lower levels of HLA-DR (Figure 4A).
Figure 4. Cell surface marker expression on monocyte subsets. A, Cell surface expression levels of markers in our gating scheme on
classical, intermediate, and nonclassical monocytes gated using the conventional approach showing the expected patterns of expression as determined by cytometry by time-of-flight. B and C, Visual interactive stochastic neighbor embedding (viSNE) analysis of human
monocytes acquired by 7 color flow cytometry, (B) traditional gating, (C) revised gating strategy. t-SNE indicates t-distributed stochastic
neighbor embedding.
8 Arterioscler Thromb Vasc Biol July 2017
To show that this combination of cell surface markers is
sufficient to accurately classify the monocyte subsets, we performed viSNE on total monocytes and overlaid the subsets
gated using the traditional approach. The resulting viSNE
map clearly shows that the 3 monocyte populations occupy
distinct regions of the plot; however, no clear boundary
separated the 3 subsets (Figure 4B). In contrast, monocytes
gated according to our revised strategy showed clearly distinct populations (Figure 4C), demonstrating that our revised
gating strategy leads to a considerable improvement in the
accuracy of monocyte subset classification using conventional flow cytometry.
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Figure 5. Monocyte subsets in human subjects with cardiovascular disease. A, Flow cytometry analysis showing conventional monocyte
subset gating for cardiovascular disease (CAD) high and CAD low individuals. B, Visual interactive stochastic neighbor embedding (viSNE)
maps showing monocyte subsets gated in A and their subsequent definition using viSNE. C, Monocyte subset frequencies defined by
viSNE in CAD hi and CAD low individuals. D, Rainbow plots showing the expected expression profiles of cell surface markers on the
monocyte subsets. E, Revised gating strategy applied to CAD high and CAD low individuals showing improved monocyte subset classification in cardiovascular disease using our strategy. t-SNE indicates t-distributed stochastic neighbor embedding.
Thomas et al CyTOF and Monocyte Subsets 9
Improvement in Monocyte Classification
in Clinical CAD Samples
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Consistently identifying the 3 monocyte subsets in clinical
samples can pose a significant challenge because of the compounding effects of variability that result from genetic diversity, disease severity, sample collection methods, and discrete
sample collection time points. Furthermore, the necessity of
freezing samples for later analysis can impose a considerable and unavoidable loss of cell surface marker integrity. We
reasoned that defining the monocyte subsets using our panel,
which incorporates the most informative redundant markers for each population, would improve subset classification
in the face of this variability. To explore this possibility, we
obtained peripheral blood mononuclear cells from patients
stratified on the basis of CAD severity using the Gensini score
index as CAD low (<10) and CAD high (>20), respectively.
Monocytes defined using the conventional approach
(Figure 5A) were used to locate each subpopulation on the
viSNE map. Nonclassical, intermediate, and classical monocytes occupied distinct regions that were not affected by
disease status (Figure 5B; Figure III in the online-only Data
Supplement). We did however observe variability in CD16
cell surface staining (Figure IV in the online-only Data
Supplement) that would typically require the arbitrary adjustment of gates on an individual basis to ensure that the correct populations were identified. To overcome this inherent
bias, monocyte frequencies were quantified using gates drawn
directly onto the viSNE biaxial plot. Increased frequencies of
both intermediate and nonclassical monocytes were observed
among CAD high individuals (Figure 5C) in line with expectations14; however, given the small sample size in this study,
these differences were not statistically significant. We observed
similar trends in monocyte subset frequencies (Figure VA in
the online-only Data Supplement) and no major differences
in monocyte numbers as a percentage when we gated using
the traditional method (Figure VB in the online-only Data
Supplement). However, the ranges of subset frequencies are
lower using viSNE-based gating, implying a more accurate
classification of monocytes by viSNE. Importantly, visualization of cell surface marker expression on the viSNE axes using
rainbow plots confirmed that the 3 monocyte populations possess the expected cell surface marker profiles (Figure 5D).
Furthermore, our optimized gating approach recovered 3 distinct subpopulations in both low CAD and high CAD individuals (Figure 5E). Thus, we have shown that our monocyte
subset gating scheme facilitates the isolation of highly purified cell preparations during CAD. Furthermore, monocyte
subset identification using viSNE coupled with highly informative cell surface markers is a powerful approach for monocyte identification when dealing with heterogeneous clinical
samples.
Discussion
Monocytes play diverse roles in infection, inflammation,
and wound repair, and changes in the population frequencies of human monocytes are associated with multiple diseases.2 Consequently, there has been considerable interest
in understanding the roles of the various monocyte subsets,
particularly during CAD in which the intermediate monocyte
population is known to expand, as observed in large-scale
epidemiological studies.15,23 It is widely appreciated that care
must be taken to ensure accurate quantification of monocytes
by flow cytometry because of the potential for contamination of CD16-positive monocytes with other CD16-positive
cell types, such as neutrophils and NK cells.23 To avoid these
issues, monocytes are defined using pan monocytic markers
such as CD86 or HLA-DR used in combination with CD14.24
It is similarly appreciated that care is also required to accurately distinguish among the 3 monocyte subsets,10 yet this
problem still poses a considerable challenge that has not been
fully resolved. Indeed, previous studies have reached opposing conclusions concerning the relative relatedness of intermediate monocytes to the classical and nonclassical subsets,
which, in fact, may be easily explained by subtle differences
in gating between studies.12,13
Here, we have presented a novel approach that takes
advantage of viSNE and high-dimensional cellular phenotyping using CyTOF to characterize human classical, intermediate, and nonclassical monocytes in unprecedented detail. Our
results show that defining the monocyte subsets using CD14
and CD16 alone is insufficient and frequently leads to intersubset contamination, regardless of the applied gating strategy. Furthermore, we observed quite considerable variability
between individuals in terms of the degree of contamination.
This cross-contamination of monocyte preparations has the
potential to compromise the interpretation of monocyte subset
function in ex vivo assays for readouts, such as RNA and cytokine expression. Naturally, the consequence of these effects is
a reduction in the consistency of results reported between laboratories and increased variability within experiments, leading
to a loss of statistical power. Thus, defining monocyte subsets
with a single pair of continuously expressed cell surface markers can explain the conflicting reports on subset relationships
and functions.
By measuring 36 myeloid-relevant cell surface markers in parallel, we defined the monocyte subsets using the
dimension reduction approach viSNE. Subsequently, the
cell surface markers that best discriminate between the
viSNE-defined monocyte subsets were identified using our
DI metric. Classical and intermediate monocytes were best
distinguished by CD16, HLA-DR, and CD11c, which were
all more highly expressed on intermediate monocytes relative to classical monocytes. We have also obtained excellent
results using CD62L as a marker that is more abundant on the
classical subset; however, we have found that CD62L is shed
on freezing, and so this marker is not compatible with frozen peripheral blood mononuclear cell preparations (data not
shown). Intermediate monocytes were best distinguished from
nonclassical monocytes using CD14, CCR2, and CD36, all of
which were lost on monocyte maturation into the nonclassical
subset. Using these highly informative cell surface markers,
we have rationally developed a gating strategy that considerably improves the purity of monocyte preparations.
We did not demonstrate significant quantitative differences in monocyte population frequencies when using
our revised gating scheme vis-a-vis the traditional gating
approach in our atherosclerosis cohort because of the small
sample size analyzed. However, our study does indeed
10 Arterioscler Thromb Vasc Biol July 2017
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
demonstrate the use of this new gating scheme in a disease setting in which cell surface marker expression levels
could change drastically. Further, our approach of combining viSNE with our optimized panel of redundant markers
greatly improves the gating and identification of human
monocyte subsets. This provides a true practical advantage
for both the quantification and isolation of both nonclassical
and intermediate monocyte subsets, in which identification
of these 2 subsets is often difficult or simply just not possible
without data of the highest quality—a major hurdle when
dealing with precious clinical samples, often ones that have
been frozen for several years. We have clearly shown that our
new panel of cell surface markers is of considerable benefit
when identifying monocyte subsets of either low-quality or
previously frozen samples.
Our new approach still relies on the sequential gating
of continuously distributed cell surface markers and so, at
least conceptually, does not distinguish between each subset with absolute certainty. However, using monocyte subset
gates defined using viSNE, we empirically demonstrate that
our revised method leads to a considerable improvement in
the purity of monocyte preparations. More generally, our
approach involves cross-validating the accuracy of cell populations defined using hierarchical gating with a small number of markers using viSNE as an unbiased high-dimensional
classifier. Our approach is akin to backgating analysis with
the advantage that cells that are not in the target viSNE gate
are easily classified based on the 2-dimensional viSNE axes.
This allows the rapid phenotyping and quantification of contaminating cell types in gating scheme, enabling the rational
exclusion of such contaminants. Because viSNE and similar
tools are commonly available in flow cytometry analysis packages such as Cytobank, this approach is broadly accessible to
the research community.
There are clear benefits to the use of this gating strategy
for molecular studies of monocyte subset function; however,
the use of this approach in population studies may warrant
consideration on a case-by-case basis. However, we do note
that the dedication of 6 fluorophore channels for monocyte
identification does impose drawbacks, such as increased antibody costs and restrictions on the number of other cell types
that may be assayed concurrently. The choice to commit 6 cell
surface markers for monocyte phenotyping may present a significant cost relative to the use of 2 cell surface markers in the
traditional gating scheme.
In summary, we have shown that viSNE is a powerful tool
for the classification of cell populations in heterogeneous and
difficult to handle samples such as frozen peripheral blood
mononuclear cells. Specifically, using our flow cytometry
staining panel that is designed to incorporate multiple, redundant, cell surface markers, the monocyte subsets can be robustly
identified. Given these benefits, we hope that the adoption of
our revised gating scheme will improve consistency in reports
of monocyte functions between studies enabling a more robust
consensus on monocyte subset functions.
Acknowledgments
We thank Cheryl Kim, director of the La Jolla Institute flow cytometry facility, for assistance with cytometry by time-of-flight.
Sources of Funding
This work was funded by NIH R01 HL118765, R01 CA202987, and
R01 HL112039 (all to C.C. Hedrick); P01 HL55798 (to C.C. Hedrick,
C.A. McNamara, and A.M. Taylor); and by a fellowship from the
American Heart Association 16POST27630002 (to G.D. Thomas).
Disclosures
None.
References
1. Ziegler-Heitbrock L, Ancuta P, Crowe S, et al. Nomenclature of monocytes and dendritic cells in blood. Blood. 2010;116:e74–e80.
2. Wong KL, Yeap WH, Tai JJ, Ong SM, Dang TM, Wong SC. The three
human monocyte subsets: implications for health and disease. Immunol
Res. 2012;53:41–57. doi: 10.1007/s12026-012-8297-3.
3.Poehlmann H, Schefold JC, Zuckermann-Becker H, Volk HD, Meisel
C. Phenotype changes and impaired function of dendritic cell subsets
in patients with sepsis: a prospective observational analysis. Crit Care.
2009;13:R119. doi: 10.1186/cc7969.
4.Azeredo EL, Neves-Souza PC, Alvarenga AR, Reis SR, TorrentesCarvalho A, Zagne SM, Nogueira RM, Oliveira-Pinto LM, Kubelka CF.
Differential regulation of toll-like receptor-2, toll-like receptor-4, CD16
and human leucocyte antigen-DR on peripheral blood monocytes during mild and severe dengue fever. Immunology. 2010;130:202–216. doi:
10.1111/j.1365-2567.2009.03224.x.
5. Grip O, Bredberg A, Lindgren S, Henriksson G. Increased subpopulations
of CD16(+) and CD56(+) blood monocytes in patients with active Crohn’s
disease. Inflamm Bowel Dis. 2007;13:566–572. doi: 10.1002/ibd.20025.
6.Rogacev KS, Cremers B, Zawada AM, et al. CD14++CD16+ monocytes independently predict cardiovascular events: a cohort study of 951
patients referred for elective coronary angiography. J Am Coll Cardiol.
2012;60:1512–1520. doi: 10.1016/j.jacc.2012.07.019.
7. Rossol M, Kraus S, Pierer M, Baerwald C, Wagner U. The CD14(bright)
CD16+ monocyte subset is expanded in rheumatoid arthritis and promotes
expansion of the Th17 cell population. Arthritis Rheum. 2012;64:671–
677. doi: 10.1002/art.33418.
8. Nagasawa T, Kobayashi H, Aramaki M, Kiji M, Oda S, Izumi Y. Expression
of CD14, CD16 and CD45RA on monocytes from periodontitis patients. J
Periodontal Res. 2004;39:72–78.
9. Urra X, Villamor N, Amaro S, Gómez-Choco M, Obach V, Oleaga L,
Planas AM, Chamorro A. Monocyte subtypes predict clinical course and
prognosis in human stroke. J Cereb Blood Flow Metab. 2009;29:994–
1002. doi: 10.1038/jcbfm.2009.25.
10. Ziegler-Heitbrock L, Hofer TP. Toward a refined definition of monocyte
subsets. Front Immunol. 2013;4:23. doi: 10.3389/fimmu.2013.00023.
11.Zawada AM, Fell LH, Untersteller K, Seiler S, Rogacev KS, Fliser D,
Ziegler-Heitbrock L, Heine GH. Comparison of two different strategies
for human monocyte subsets gating within the large-scale prospective
CARE FOR HOMe Study. Cytometry A. 2015;87:750–758. doi: 10.1002/
cyto.a.22703.
12. Cros J, Cagnard N, Woollard K, Patey N, Zhang SY, Senechal B, Puel
A, Biswas SK, Moshous D, Picard C, Jais JP, D’Cruz D, Casanova JL,
Trouillet C, Geissmann F. Human CD14dim monocytes patrol and sense
nucleic acids and viruses via TLR7 and TLR8 receptors. Immunity.
2010;33:375–386. doi: 10.1016/j.immuni.2010.08.012.
13.Wong KL, Tai JJ, Wong WC, Han H, Sem X, Yeap WH, Kourilsky P,
Wong SC. Gene expression profiling reveals the defining features of the
classical, intermediate, and nonclassical human monocyte subsets. Blood.
2011;118:e16–e31. doi: 10.1182/blood-2010-12-326355.
14. Shantsila E, Tapp LD, Wrigley BJ, Pamukcu B, Apostolakis S, MontoroGarcía S, Lip GY. Monocyte subsets in coronary artery disease and their
associations with markers of inflammation and fibrinolysis. Atherosclerosis.
2014;234:4–10. doi: 10.1016/j.atherosclerosis.2014.02.009.
15. Weber C, Shantsila E, Hristov M, Caligiuri G, Guzik T, Heine GH, Hoefer
IE, Monaco C, Peter K, Rainger E, Siegbahn A, Steffens S, Wojta J, Lip
GY. Role and analysis of monocyte subsets in cardiovascular disease.
Joint consensus document of the European Society of Cardiology (ESC)
Working Groups “Atherosclerosis & Vascular Biology” and “Thrombosis”.
Thromb Haemost. 2016;116:626–637. doi: 10.1160/TH16-02-0091.
16. Bendall SC, Nolan GP, Roederer M, Chattopadhyay PK. A deep profiler’s
guide to cytometry. Trends Immunol. 2012;33:323–332. doi: 10.1016/j.
it.2012.02.010.
Thomas et al CyTOF and Monocyte Subsets 11
17. Mair F, Hartmann FJ, Mrdjen D, Tosevski V, Krieg C, Becher B. The end of
gating? An introduction to automated analysis of high dimensional cytometry data. Eur J Immunol. 2016;46:34–43. doi: 10.1002/eji.201545774.
18.Saeys Y, Gassen SV, Lambrecht BN. Computational flow cytometry:
helping to make sense of high-dimensional immunology data. Nat Rev
Immunol. 2016;16:449–462. doi: 10.1038/nri.2016.56.
19. Amir el-AD, Davis KL, Tadmor MD, Simonds EF, Levine JH, Bendall SC,
Shenfeld DK, Krishnaswamy S, Nolan GP, Pe’er D. viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leukemia. Nat Biotechnol. 2013;31:545–552. doi: 10.1038/nbt.2594.
20.Becher B, Schlitzer A, Chen J, Mair F, Sumatoh HR, Teng KW, Low
D, Ruedl C, Riccardi-Castagnoli P, Poidinger M, Greter M, Ginhoux F,
Newell EW. High-dimensional analysis of the murine myeloid cell system.
Nat Immunol. 2014;15:1181–1189. doi: 10.1038/ni.3006.
21. Maecker HT, Frey T, Nomura LE, Trotter J. Selecting fluorochrome conjugates for maximum sensitivity. Cytometry A. 2004;62:169–173. doi:
10.1002/cyto.a.20092.
22. Tenenbaum JB, de Silva V, Langford JC. A global geometric framework
for nonlinear dimensionality reduction. Science. 2000;290:2319–2323.
doi: 10.1126/science.290.5500.2319.
23. Zawada AM, Rogacev KS, Schirmer SH, Sester M, Böhm M, Fliser D,
Heine GH. Monocyte heterogeneity in human cardiovascular disease.
Immunobiology. 2012;217:1273–1284. doi: 10.1016/j.imbio.2012.07.001.
24. Heimbeck I, Hofer TP, Eder C, Wright AK, Frankenberger M, Marei A,
Boghdadi G, Scherberich J, Ziegler-Heitbrock L. Standardized single-platform assay for human monocyte subpopulations: lower CD14+CD16++
monocytes in females. Cytometry A. 2010;77:823–830. doi: 10.1002/
cyto.a.20942.
Highlights
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
• Human nonclassical and intermediate monocyte subsets gated using the conventional CD14 versus CD16 biaxial plot method are commonly
contaminated, leading to inaccurate quantification and potentially confounding molecular analyses.
• Using cytometry by time-of-flight mass cytometry, we devised an optimal cell surface staining panel for monocyte analysis that incorporates
the 4 additional cell surface markers CD36, CCR2, HLA-DR, and CD11c.
• We demonstrate using flow cytometry that this revised cell surface panel leads to considerably lower levels of intersubset contamination.
Downloaded from http://atvb.ahajournals.org/ by guest on July 28, 2017
Human Blood Monocyte Subsets: A New Gating Strategy Defined Using Cell Surface
Markers Identified by Mass Cytometry
Graham D. Thomas, Anouk A.J. Hamers, Catherine Nakao, Paola Marcovecchio, Angela M.
Taylor, Chantel McSkimming, Anh Tram Nguyen, Coleen A. McNamara and Catherine C.
Hedrick
Arterioscler Thromb Vasc Biol. published online June 8, 2017;
Arteriosclerosis, Thrombosis, and Vascular Biology is published by the American Heart Association, 7272
Greenville Avenue, Dallas, TX 75231
Copyright © 2017 American Heart Association, Inc. All rights reserved.
Print ISSN: 1079-5642. Online ISSN: 1524-4636
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://atvb.ahajournals.org/content/early/2017/06/08/ATVBAHA.117.309145
Data Supplement (unedited) at:
http://atvb.ahajournals.org/content/suppl/2017/06/12/ATVBAHA.117.309145.DC1
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Arteriosclerosis, Thrombosis, and Vascular Biology can be obtained via RightsLink, a service of the
Copyright Clearance Center, not the Editorial Office. Once the online version of the published article for
which permission is being requested is located, click Request Permissions in the middle column of the Web
page under Services. Further information about this process is available in the Permissions and Rights
Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Arteriosclerosis, Thrombosis, and Vascular Biology is online
at:
http://atvb.ahajournals.org//subscriptions/
SupplementalFigures
G.D.Thomas,etal.
TableI.AntibodiesandheavymetalconjugatesusedinCyTOF study
a)
b)
FigureI.a)MonocytesubsetfrequenciesineachindividualusedforCyTOF analysis.b)
viSNE overlayshowingthelocationofclassical(green),intermediate(orange)and
nonclassical(blue)monocytessuperimposedonallPBMCs.
cluster Scatter Plot
20
tsne_2
10
0
−10
−20
b)
● ● ●● ●
●
●● ●● ●
●
●
●●
●
● ●●
●
●
●
●
●●
●
●●●●
●
●●
●
●
●
●●
● ● ● ● ● ●●●
●●●
●
●●
● ●● ●
● ●
● ●
●●●
●● ●
●
●● ●
●
●
●●●●●●●
●● ●●
● ●
●
●
●●● ●
●
●
●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●●●●●●● ● ●
●
●
●●● ● ●
●●●●
●●
●● ●
●
● ●●
●
●● ●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●●●●●●●
●●
● ●● ●
●●
●
●●
●
●
●● ● ●
●
●
● ●●
●
●● ●●
●●
●●
●
●
●●
●●
●●
●
●● ●
●
● ●●
●●
●●● ●●●
●●
●
●
● ●●●●●
●●
●
●● ●
●●● ●●
●●
●●
●
●●
●●●● ●●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
● ●
●
●●●
●● ●
●● ●●●●●●●
●
●
●●●
●●
●
● ●●● ● ●
●●
●
● ● ●● ●●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●●
●
●
●●● ●
●● ●● ● ●●●
●
●
● ●●
●
●
●● ● ●●
● ●
●
●●
●●
●●
●●● ●●●
● ●●
●
●●●
●●
●●
●
● ●●
●
●●
●
●● ●
●●
●●●
● ●●
●●
●
●● ● ●●
●
●
●
●● ●
●●
●
●●● ●●● ●●
●
●
●●●
●
●●
●
●
●●●●●●●
● ●
●●
● ●●
●●●
●
●●
●
●●●
●
●
●
●
●●
●
●
● ● ●●
●
●
●●●● ●
●
●
●
● ●●
●● ● ●
● ●● ●●
●
●●
●●
●●●●●
●●
●
●●
●● ●●
●
●●
●● ●
●●● ●
●●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●●
●●● ● ● ●
● ●●
●
●
●●
●
●●
●
● ●
●
●●
● ●●
●●●
●
●●
●
●●
●
●●
●● ●
●
●●
●●●●●● ● ●● ●● ●●
●●
●●●
●●
●
●
●
●●
●
● ●
●
●●●
●
●
●●●●
●●
●●●
●● ●
● ●●
● ●●
●●
●
● ●●
●●
●
●●
●●●
●
●●
●●● ●
●
●●
●●
●
●
●
●●
●●
●●●●●●
●●●
●● ●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ●● ● ●●●
●
●●
●
●●
●●●
● ● ●●●●●
● ●
●
●
●
●
●● ●●
●●
● ●
●
●
● ●
●
●
●● ●
●● ●●
●
●
●
●●● ●
●●
●●
● ● ●●
●
●
●
●●
●
●
●●
● ●●●●
●
● ●●
●●●
●●
●
●
● ●
●
●●
●●● ●
● ●●●
●● ● ●
●●●
●● ● ●
● ● ●● ● ● ●
● ●
●●●
●
●●●●●●
●●●●●
●●
● ●●
●
● ●●
● ●●●
●
● ● ●
●● ●
●
●●
● ●
● ●● ●● ● ●
●
●
●●
●
●
● ●●
●
●●
● ●●
●
●
●●
●
●
●●●●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●● ●
● ●●
●
●●
● ●
●●
●● ●●●● ●
●
● ●●
●
●●
●●● ●●
●●
●●●● ●● ●
●
●
●
●
●
●●
●
●●
●
●
●
● ●●
●
●
●
●
●
●
●●
●
●
● ●●
● ● ● ●
●●●●
●
●
●
● ●
●
●
●
●●●●
●
●
● ●●
●●
●
●
●●
●
●●
●●
●
●● ● ● ●
●●
●●
●
●
●●
●●
●
●
●
● ●●
●
● ●
●
●
●●●●●
●
●●●
●●
●●
●●
●
●●
●
●●●
●●●
●
●
●
●●
●●
●●●
●●
●●● ●
●
●●●
●●
●
●
●●
●●● ●
●
●●
●●
●
●● ●
●●
●
●●●●●
●
●●
●●●●
●●
● ●
●
● ●
● ●
●
● ● ●●
●●
●● ●
●
●●● ●●
●
●● ●
●
●● ●
●● ●
●●●● ●
●
●
●● ● ●
●
●
●● ● ●
●
●●
●●
●
●
● ● ●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ●
●
●● ●
●●
●●● ●
● ●●●
●
●
●
●
●●
●
● ●
●
●
●●●●
●
● ●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
● ●●
●●
●
●
●● ● ●●●●
●● ●
●
●●●●
●●● ●● ● ●
●●●
●●
● ●
●
●
●
●
● ●●
●●
●
●●
●
●
●
●●●● ● ●
●●
●
● ●●
●●
●
●
●
●●
●
● ●●
●
●
●
●
● ●●
●
●
●
●● ●
●
●● ●
●●●
●
●
●
●●●
●
● ●
● ●●
● ●
●
● ● ●●
● ●
●●
●●●
●● ● ●
●
● ●
●●
●
●
●● ●
●●●●●
● ●
●
●●
●
●● ●●●
●
● ● ●●
● ●
●
●
● ●
●
● ● ●
●●
●
●● ●
● ●
●●
●● ●
● ● ●● ●
●
●●
●●●
●● ●
●
●●
●●● ●
● ● ●
● ●● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●
●●● ● ●
●
● ●
●●●
●●
●
●●
●
●● ●●● ●●
●
●●●
●
●
●
● ● ●●
●●
● ●
●
●
●
●
●●●●
●
●
●●
●●
● ●● ●
●●
●
●
●
●
●
●
● ●
●
●
●●
●
●
●
●
●
●
●
●
●●●
● ● ●●● ●●
●● ●●
●●
● ●●
● ●●
● ●●●
●● ●●
●
●
●
●●●
●
●●
●● ●
●●
●●
●●●●
●●
●
●
●
●
●●
●
● ●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●●
●● ●
●
●
●● ●
●
●●
●● ● ●
●●●
● ●
●
●●
●
●
●
●●
●● ●
● ●
●●
●
●●
●
●●
●
●●
●● ●
●●
● ●
●
●
●
●
●
● ●
●●● ●
●
●
● ●●
●●●● ●●
●●
●
● ●●●●●
●●
●
●●
●●
●●
●
●● ●
●●●
●
●
● ●●
●
● ●●●●●● ●
● ●
● ●
●●●
●●●●● ●
●●
●
●●
●●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●●
●● ●
●
●
●
●
●●●●
●●●●
●
●
● ● ●●
●● ●
●
●
●●
●● ● ●
●●
●
●
● ●● ●
●●●
● ●●
●●
●●
●
●● ●●●●
●
●●●● ●
●
●●●●● ●●
● ●
●●● ●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●● ● ●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●●● ●●
●●● ●● ● ●● ●
●● ●
●
●●
● ●●
●
●● ●
●●●
●●●
●
●
●
● ● ●● ●●
●●
●●
●
●●
●●●●
●
● ●
●
●
●
●●
● ●●
●●●●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●●●
●
●
●
●●● ●●●
●●●
● ●
●● ●
●● ●
●●●
●
●● ●●
●
●
● ● ●●●
●
●●
●●
● ●●
● ●
●
●●
●
●
●●
●
●●
●
●
●● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ●●
●
●●
●
●●●
●●●● ●
●
●
●● ●
●●
●●
●● ● ●● ● ●● ●●
●
●●●
●
●
●
● ●
●●
● ●●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
● ●
●
●●
●●●
●●
●● ●●●
●●●
●●
●●
●
●
●
●
●●
●
●● ●
●
●
●●
●
●●
●
●
●
●
●●
●●
●
● ●●●
●●
●●
●
● ●
●
●
●
●●●
●●●
●●●●●
● ●
●● ●●
●
●●●
●
●●
●●
● ●●
●
●
●
●●● ●
● ●●
●●
●● ●
●●●
●
●
●
●
●●
●
●●
● ●●●
●●
●
●●
●
●
●●
●●
●● ●
●
●●●
● ●
●
●
●●
●●
●
● ●●
● ●
●
●
●
●●
●
●
●
●
5
2
9
6
15
4
10 1714
1
12
7
11
3
−30
−20
−10
0
10
20
50
25
8 13
16
cluster Scatter Plot
isomap_2
a)
30
0
−25
−50
●
●
●
●
●●
● ●●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
● ●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●●●
●●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
● ●
●
● ●
●
●●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
● ●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
● ● ●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●●
● ●●
● ●●
●●
●
●●
●●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
● ●●●
●●● ●●
●
●●
●
●
●
●
●
●●●
●
●
●
●●●
●●● ●
●
●●
●
●
●
● ●●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●● ●
●
●
●
●●
●
●
●
●
●
●
● ●●●
●
●
●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●● ●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●● ● ● ●●●●
●●
●●
●
●
●●
●
● ●
●
●●
●
●●
●●
●
●●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● ●●
●
●●
●
●
●●
●●●
●
●
●●● ●● ● ●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
● ●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●●●
●
●
●
●●●● ●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
● ●
●
●● ●●
● ● ● ● ●●
●●
●
●●●
●
●
●
●●
●
●
●●
●
●
●●
●
● ●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●
●
●
●
●●
●●
●●
●●
●●●
●
●●
●
●
●
●
●
●●●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●●
●
●
●●● ●
●
●●
●
●●
●
● ● ●●
●
●
●
●
●
●
●●
●
●●
●
●
●●●
●●
●
●
●●●●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
● ●
●
●
●
●
●
●●
●● ●●●●● ●● ● ●●
●
●
●
●
●●
●●
●●●
●
●
●●
●
●
●
●● ●
●
●
●●●
●
●
●
●
●●
●
●
●●●● ● ● ●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
● ●●
●
●●
●●●●
●
●
●
●●●
●●
●
●
●
●
●
●●●
●
●
●● ●● ●
●●
●●
●●●
●
●●
●
● ●● ●
●
● ●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
● ●
●
●●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
●
●●● ● ● ●
●
●●●
●
●
●●
●
●
●●
●●
●
●●●●●● ● ●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●●
●
●●
●
●
●●
●
●●● ● ● ● ●
●
●●●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●●●●●● ●
● ● ●●● ●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
● ● ●
●
●
●
●
●
●●
●
●
●
● ●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●●● ● ● ● ●●
● ● ●●●●
●●●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●●●● ●●●
●●
●●
●●●● ●● ●
●
●●●
●●
●●●
●●
●
●
●
●
●
●
●●●
●
●●
●
●●●●
●
●
●
●
●●
●
●
●
●
●● ●
●
●
●
●
●
●●●
●
●
●●●●
●
●●●●●
● ●●
●●
●●●
●●
● ● ●●
●
●
● ●
−80
30
−40
0
cluster_ClusterX
c)
2.5
●
●
1
2
●
●
1.5
1.0
Expression
2.5
2.0
1.5
1.0
2.5
2.0
1.5
1.0
4
●
●
5
6
●
●
7
8
●
●
9
10
●
●
11
12
●
●
13
14
●
●
15
●
17
cluster
16
●
●
1
2
●
●
●
●
3
4
5
6
●
●
7
8
●
●
9
10
●
●
11
12
●
●
13
14
160Gd_CD14 (v)
158Gd_HLA−DR (v)
● ●
●
●● ●
●
●
●
● ●●●●
●
●●
●
●
●
● ●●●●
●
●
●
●●●
●
●
● ● ●●
●
●
●●
●
●●●
●●
●
●●
●
●
●●
●
●●
●●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
● ●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●●●
●●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●
●●●
●
●
●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
● ●●●●●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●● ●●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●●●
●●●
●
● ●
● ●
●
●●
● ●●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●●
●●●
●
●
●●
●
● ●
●
●
●●●
●●
●
●
●
●
●●●
●
●
● ● ● ●●
●
●●
●●
●
●●
●
●
●
●
●●
●
●●
●
●●
●
●●
●●●
●●
●
●●
●
●
●● ●
●
●●●●
●●●● ●
●●
● ●●
● ● ●●
●●●
●●
●
●
●
●●
●●
● ●
●
●●
●●
●
●●
●●●●
●
●● ●●●●● ●●
●●
●
●
●●
●
●
● ●
●● ●
●
●●
●●● ●
●●●
●●●
●●●
● ●
●●
●
●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●●●● ●
●
● ●
●
●● ●
●
●
●
●
●
●
●●●●
●
●
● ● ●● ● ●
●●
●
●● ●● ● ●
●
●●
●
●
● ●●
●●
● ●●
●●
●●●
●
●● ●
●
● ● ●
●
●
● ●●●●
● ● ●● ●
●
●
●
●
● ●●
● ●●
●●
●● ●
●
● ●
● ●
●
●
●●
●● ●
●● ●
●●
●
●●●● ●●● ●
●●
●●●●
●●
●
●
●
●
●●●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●●
● ● ● ●
● ●
●
●●
●●●
●●
● ●
●●●● ●●●●
●●●●
●●
●●
● ●
● ● ●
●●
●●●●●● ●
●●
●
●
●●●
●
●
●
●●
●●
●
●●
● ● ● ●●
●●●●
●
●
●●
●
●●
●
●
●●● ●●
●
●
●●
●●
●●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●● ● ●●●
● ●●
●●
●
●
●
●●
●
●
●
●●
●● ●●● ● ● ●
●●●
●
●
●
●
●
●●
●●
●●●
●
●
●●
●●
●● ●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●
● ●●●
●
●
● ●
●
●
●
●
●
●●●●●
●
●
●
●●
●
●●
●
●
●● ●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●● ● ●
●
●
●●●● ●●●● ●●●
●
● ●
●●
●●
●
●●
●●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●●●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●●
●●
●
●
●
●
●●
●
●●●●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●● ●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●●
●●
●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●●
●
●
● ●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●●● ●
●
●
●●
●
●
●
●
●● ●●●●●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●● ●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●●● ●
●
●
●
●●
●
●
●
●
●
●●●●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●● ●●
●
●
●
●
●
●
●
●●●
●●●●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●●●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●●
●●●
●
●
●●●
●
●
●●
●●
●
●
●
●
●● ●
●● ●●
●●
●●
●
●
● ● ●●● ●● ● ●
●●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●● ●
●
●
●
●●●
●●
●
●
●
●●
●●
● ●●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●● ● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●● ●
●
●
●
●
●
●
●●
● ●●
●
●
●●
●
●
●●●
●●
●
●●
●
●●
●
●
●● ●● ●●● ●
●
●●
● ●●● ● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●●●
● ●
●
●
● ●●
● ●
●●
●
●●
●
●
●●●
● ●
●●●
●
●
●●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●● ●●
●●
●●
●
●
●●
●
●●
●
●●●●
●
●
●●
●
●
●
● ●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●●
●
●●
●●
●
●
●
●
●●●
●
●
●●
●●
●●
●
● ●●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●●●●●
●●
●
●●
●
●
●
●
●
●
●●
●
●●●
●
●●
●●●●
●
●
●
● ●●● ● ●● ●
●●●
●●
●●
●
●
●
●
●
●
●
●●●●●
●
●● ●
●
●
●
●●
●
●●●●●
●
● ●
●
●
●
●
●
●●
●
●
●
● ●●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●●
●
●●
●
●●
●
●●
●●
●
●
●
●
● ●
● ●● ● ●
●●
●
●●
●●●
●
●
● ●
●●
●
●
●
●
●●
● ●●
●●
●●
●●
● ●
●●
●● ● ●
●
●●
●
●
●● ●
●●●●
● ● ●●
●
●●●●
●
●●●●●
● ●
●
●
●
●
●● ●●●●● ●
● ●●
●
●●
●● ● ●
●●
●
●
●
●
●
●
●
●● ●
● ●●
9
2.0
3
40
isomap_1
tsne_1
16
15
1
12 5
2
14
6
8
4 10
13
7
3
●
2.5
2.0
1511 7 6
1
1.5
11
1.0
9
5
13
2
16
3
8
10
● ●●
●
●
●
●●
●● ●
● ● ● ● ●●●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●●●●
●
●●
●
●●
●
●
●
●●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●●
● ● ●●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●● ●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●● ● ● ●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●● ● ●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●● ●●
●
●●
●
●
●●
●
●●
●
●●
●●
●
●
●●●
●
●
●●
●
●● ● ● ●● ●
●●
●●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
● ●●●
●
●●
●●
●●
●●●
●
●
●●
●
●
● ●
●●●●
●
●
●
●
●●
●
●
●●
●
●
●
●●●●
●
●
●●
●●●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●●●●
●●
●
●
●●●
●
●
●
●
● ●
●
● ●●
●
●●
●
●
●
●●●
●
●●
●
●●
●
●●
● ●●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●● ● ●●●●● ●
●
●
●●●●
●●
●
●●●
●●
●●
●●
●●●
●●● ●
●●● ●● ●
●
●●
●
●●
●●
●●●●
●
●● ●
●●
●●
●
●●
●
● ●
●●●●●
●●
●●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●●
●
● ●
●●
●
● ●
●
●
●
●●
●●
●
●
● ●●
●
●
●●
●●●
●
●
●● ●
●●
●
●
●●●●●●
●
●
●
●●
●
●●
●
●●
●
● ●●
●●
●
●●
●
●
●
● ●
●
●●
●●
●
●●●
●●
●
●
●●
●●
●
●
●●
●
●
●●
●● ● ●
● ●
●●
●
●●
●
●
●
●●
●
●
●●●
●●● ●● ●
●
●●●●
●
●
● ●●● ●
●●
●●
●●
●●
● ● ●● ●
●
●● ●
●●
●●● ●●
●
●●
●
●
●
● ●● ●
●●
●●
●●
●●
● ●●●
●
●●
●
●
● ●●●●
● ● ●
●
●●
●●●●
●●●●
●
●
●
●●●
●●
●
●●
● ●●
● ●
●
● ●
●
●
●
●
●●
●●
●●
●
●
●
●
● ●
●
●●●
●●●●
●
●
●●●●
●●●
●
●
●
●
●●
●
●
● ●●●
●●●●●●●●
●
●
●
●●
●●●
●
●
●●●
●●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●●
● ●●●● ●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●●●●
●
●●●
●● ●
●
●● ●●●
●●●
●
●
●
●●
●
●●
●●
●
●●● ● ● ●
●
●●
●
●
●
●
●●●●●
●
● ●●
●
●
●●
●● ●
●●
●●
●
●● ●
●●●
●
●
●●
●●
●
●●●
●
●
●●●●
●
●
●
●
●
●●
●
●
●
●
●● ●
●●●
●
●
●
●
●
●
●
●
●●
●●
●●
●●●
●●
●
●
●
● ●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ● ●
●
●
●
●
●
●●
●
●
● ●● ●● ●●
●●●
●
●●●●●
●
●
●
●●
●
●
●
●
●
●
●
●●●●
●
●● ●● ● ●●
●
●
●
●●
●
●
●●
●
●●●
●
●
●● ●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●●
●● ●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●●
●
●●●
●
●
●
●●
●●
●
●
●●
●
●
●●●
●
●●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ● ●● ● ●●●●●●●
●
●
● ●●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
● ●● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●● ● ●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
● ● ● ●● ●●●
●
●
●
●
●
●
●
●●●●
● ●
●
●
●
●
●
●●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●● ●●●
●
●
●
●
●●
●
●
●
●●● ●● ● ●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●●
●
●
●
●● ●
●
●
●
●
●●
●
●
● ●●
● ●●
●
●●
●
●
●
●●
●
●●
● ●●
●●●
●
●
●● ●●
●
●
●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●● ●
●●●●●●● ● ●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●● ●
●● ●●
● ●●
●●
●●
●
●●
●
●
●●
●●●
● ●●
● ●
●
●●●●● ●
●●
●
●
●
● ●●
●
●●
●
●
●
●
●
●● ●
● ●
●
●
●
●
●
●
11
14
7
6
15
16
1
12
4
8
5
2.5
2.0
15 11 16 6
1
7
3
1.5
12 4 14
9
5
13 2
8
10
165Ho_CD16 (v)
●
● ●
●●●
●
●●
●
●
●
●
●●
●
●●●
●●
●●
● ●●
●
●
● ●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● ● ●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●●
●
●●
●
●●● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●●●●●●●
●
●
●
●
●
●●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ● ● ●●●
●
●●
●●
●●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●●
●●
●●●
●
●
●
●● ●
●
●●
●
●
●
●
●
●
●●
●
●●●●
●
● ● ●● ●
●●
●
●
●●●
●
●●
●●●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ●
●
●●●● ● ● ●
●●
●●
● ●●●●●●●
●
●
●
●● ●
●
●●
●●
●●
●
●●
●
●
●●
●
●
● ●
●
●
●
● ●
●●●●
●●●●●
●●
●
●
●
●●●
●●● ●
●●●●●●
●
●
●●●
●●
●
●
●
●●
●
●●
●
●●
●
●
●
●● ●
●●●
●
●●
●● ●●●
●●
●●
●●●●
●● ● ● ●●
●
●
●
●● ●
●● ●
●●
●
●
●
●
●
●●
●●
●
● ●
●●●
●
●
●
●
●
●
●●
●●
●● ●
●
●●●
●●
●●
●
●●●●
●●●
● ●
●
●
●
●
●● ●●
●
●
●
●
●
●
●
●
●
●
●
●● ●●●●
●
●
●●
●
● ●● ●●●
●
●
●
●
●
●
●
●●● ●
●
●
●
●
●●
●●
●
●●
●
●
●
● ●●
●●●
●
●
●●●
●●●
●●●●
● ●●
●
●
●●
●
● ●●
●●
●
●
● ●●
●
●●
●●
●
●
●●
●
●●
●●
●●●
●●
●●
●●●●●
●
●
●
●●●
●●
●●
●●
●
●
●
●
●
●●
●
●
●
●
● ●●
●
●●
●●
●
●
●
●●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●●
●
●●●●●
●
●● ●
●
●● ●
●
● ●●
●●
●
●●
●
●
●●
●●●●
●●●●
●●
●
●
●● ●●●
●
● ● ●●
● ●
●
●
●
●
●
●●●
●
●
●●
●●
●
● ●●
●●●
●
●
●
●
●
●
●●●●●
●
● ●
●●● ●
●
●● ●
●● ●●
●
●
●
●
●
●
●
●
● ●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●● ●
●● ●
●●
●
●
●●●
●
●● ● ●●●
●
●
●●●●●
●
●●
●
●●
●
●●●
●
●
●
● ●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●●
●
●
●
●●●●●●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●●●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●●
●
●
●● ●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●●
●
●
●
●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●● ●● ●
●●
●
●● ●
●
●●
●
●
●
●
●●●●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●●●
● ●●
●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●● ●
●
●
●●●
●
●●
●
●
●
●●
●●
●
●●
● ●●●●
●
●●
●
●
●● ●
●
●●
●●
●
● ●
●
●●●
●
●
●
●
●
●
●●●
●●
●
●
●
●●●
●
●● ●
●●●
●●
●●
●
●
●
● ●
●
● ●
●
● ●●
●
●●
●
●
●
●
●
● ●
●
●
●
●
●●
●●●
●●
● ●● ●● ● ●●
●●●
●
●●
●
●
●
●
●
●●●
●
● ●
●
●
●● ●
●
● ●●●
●●
●
●●
●
●
● ●●●
●● ● ● ● ●
●●
●
● ●
●
●●
●● ●
●●●
●●●
●
●● ●
●
●
●
●●
● ● ●●
●
● ●
●
●
●●
●
● ●● ● ●
●
●●
●
●
●
●● ●
●
●●
●●
● ●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
● ●● ●●●
● ●●
●● ● ● ●
●●●●●●
●● ●●
●
●
●
● ●●
●
●
● ●
●
●
●
●●●●
●
●
● ●●
●●●
●●● ●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
● ●●
●
●
●
● ●● ●
● ●●
● ●
●●
●
●●
●
●●
●
●●
●
●
● ●● ● ●
●
●●●
●●
●
●
●
●●
●
●●
●
●
●
●●●●●● ●
●
●
●●
●
●●
●
●
●
● ● ●●
●
●
● ●
●
●
●●
●●●●
●
●●
●
●
● ●
●
●●
●●
●
●● ●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
● ●
●
●
● ●●●
●● ●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●●●
●●
●● ●
●● ●●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
−80
−40
0
40
1.6
1.2
0.8
● 2
● 3
● 5
● 6
● 7
● 8
● 9
● 12
155Gd_CD36 (v)
2.0
● 1
● 11
●
1.0
2.4
Cluster
● 10
3
10 14
13 9 4
16
8
11
12
5
15
7 6
2
3
1
sample ● data
16
● 4
159Tb_CD11c (v)
10
15
12 4
14
153Eu_CCR2 (v)
13 9 2
●
●
15
11 1613
9
1
3
7
12
6
5
14
10
8
0
40
2
−80
−40
4
● 13
● 14
● 15
● 16
isomap_1
FigureII.a)ClusterX partitioningofthemonocyteviSNE analysisinFigure2a.b)Isometric
mapping(ISOMAP)analysisofClusterX .c)Expressionpatternofmostinformativecell
surfacemarkersplottedagainstISOMAPaxis1showingthepositionsoftheClusterX
clusters.
CADlow
CADhigh
FigureIII.Classical(blue),intermediate(orange)andnonclassical(green)monocyte
subsetsinCADlowandCADhighindividualsdisplayedonaviSNE analysisofall
monocytes.
C
Int.
NC
C
Int.
NC
FigureIV.CD16expressiononClassical(left,C),intermediate(center,Int.)and
nonclassical(right,NC)monocytesubsetsdefinedbyviSNE fromCADsamplesshownon
CD14vsCD16biaxialplot.
Materials and Methods
Healthy human subjects
Heparinized blood from 6 healthy volunteers was obtained after written informed
consent under the guidelines of the Institutional Review Board of the La Jolla Institute
for Allergy and Immunology according to NIH guidelines for the protection of human
research subjects.
Blood preparation
Peripheral blood mononuclear cells (PBMCs) were isolated from 10mL freshly drawn
heparinized blood using Ficoll-PaqueTM PLUS (GE Healthcare Biosciences AB) and
SepMateTM-50 (Stemcell Technologies Inc) according to the manufacturer’s protocol. As
a blocking step, cells were resuspended in 1mL FACS buffer (1mM EDTA, 1% BSA, 1%
Human AB serum, 0.1% NaN3 in 1xPBS) and counted on the Accuri C6 (BD
Biosciences).
Samples from human subjects with cardiovascular disease
PBMCs were obtained from the Coronary Assessment in Virginia cohort (CAVA)
cohort1, frozen in FBS and 10% DMSO, stored in liquid nitrogen and shipped on dry ice
to LJI for analysis. This cohort includes patients between the ages of 30 and 80
undergoing a medically necessary cardiac catheterization. Patients were excluded if
they had any of the following: a current acute coronary syndrome, cancer of any type,
autoimmune disease of any type, anemia, pregnancy, HIV infection,
immunosuppressive therapy, or prior organ transplantation. All protocols and
procedures were approved by the Institutional Review Board at the University of Virginia
(IRB HSR #15328).
Quantitative Coronary Angiography
All CAVA patients underwent standard cardiac catheterization with two orthogonal views
of the right coronary artery and four of the left coronary artery according to accepted
standards. Quantitative coronary angiography (QCA) was performed using automatic
edge detection at an end diastolic frame selected based on demonstration of the most
severe stenosis with minimal foreshortening and branch overlap. Computer software
(Artis Workplace, Siemens Medical Solutions) was used to calculate the minimum
lumen diameter, reference diameter, and percent stenosis. Analysis was performed by
two blinded, independent investigators. The Gensini Severity Scoring system was used
to calculate angiographic scores and estimate CAD burden.
CyTOF staining and data acquisition
First, cells were washed with 10mL PBS, resuspended at 1x10 7 cells/mL in PBS and
stained with 1mL 5mM Cisplatin 195Pt (Fluidigm) for 5 min at room temperature to
exclude dead cells. Thereafter, the cisplatin was quenched with Cell Staining Buffer
(CSB; 2mM EDTA, 0.1% BSA, 0.05% NaN3 in 1xPBS; 5x the volume of the cell
suspension). Cells were then washed twice with 1mL CSB and stained with 100μL of
heavy-metal isotope labeled antibody cocktail (Table S1) on ice. After 30 min, cells were
washed twice with 1mL CSB and fixed overnight at 4 oC with 100μL 2% PFA (Electron
Microscopy Sciences). Following fix, cells were washed twice with CSB and
permeabilized with 100μL of 1xPermeabilization buffer (eBiosciences) for 45 min on ice.
Cells were washed twice with CSB and incubated with 100μL 125nM nucleic acid IrIntercalator (Fluidigm) for 30 min at room temperature. Finally, cells were washed twice
with CSB, counted on Accuri C6 and washed thrice with MilliQ water. Cells were left
pelleted until ready to run.
Immediately before acquisition on CyTOF 2 Helios (DVS Sciences/Fluidigm), cell pellets
were resuspended at 1x106 cells/mL in 0.1x EQTM Four Element Calibration Beads
(Fluidigm) and filtered into FACS tubes with cell strainer caps (Fisher Scientific). Quality
control and tuning were performed on a daily basis before acquisition. After acquisition,
files were automatically transformed into an fcs-format by the Helios Software
(Fluidigm).
Antibody conjugation
For mass cytometry stainings, already conjugated antibodies were purchased from
Fluidigm and purified antibodies were ordered from the companies listed in Table S1.
Antibody labeling was performed with the Maxpar X8 Multi-Metal Labeling Kit (Fluidigm)
according to the manufacturer’s instructions. Conjugations were verified using BD
CompBead Plus (BD Biosciences) and antibodies were titrated before use.
CyTOF data normalization
Data was normalized (including EQ bead-removal) using Matlab-based normalization
software NormalizerR2013a_Win64.
Analysis of CyTOF data
Preliminary analysis of CyTOF data were performed in Cytobank and in most instances
viSNE analysis was performed in Cytobank using the default settings with all samples
and all heavy-metal conjugated cell-surface markers used for input, except for DNA
controls. Biaxial plots for figures 1-4 were made in R using the ‘flowWorkspace’ and
‘ggcyto’ packages. PCA analysis in figure 2 was performed in R using the PCA function
in the ‘factoMineR’ package , while in figure 2 the viSNE and isomap analyses were
performed using functions available in the the ‘cytofkit’ package with the default settings
2. WIthin R FCS data were transformed using the asinh transform method with
parameters a=1, b=1, c=0 prior to plotting and downstream analysis.
Discrimination Index calculation
The discrimination index to determine the ability of cell-surface markers to determine
between populations was based on the staining index proposed in 3. Specifically, for
each marker in each population, the mean signal intensity and the signal standard
deviation were obtained. The DI for a marker between a pair of populations was then
calculated as the absolute difference in the mean signal intensity divided by the sum of
the standard deviations for each population, in this instance a larger number indicates a
greater separation between populations.
Calculation of monocyte purity
To quantify the impurity in the conventional monocyte gating we performed viSNE on
live singlet events from the six healthy individuals using 200,000 events sampled
equally between individuals. viSNE was run in Cytobank with the default settings
(iterations=1000, perplexity=30, theta=0.5) using all 36 cell surface markers. Monocytes
gated using the conventional gating approach were used to locate the position of each
monocyte subset on the viSNE map and non-overlapping gates were then drawn
directly onto the viSNE plot to define the monocyte subsets. Following this we asked
whether cells gated using the conventional monocyte gating scheme were recovered in
the viSNE-defined classical, intermediate or nonclassical gate or elsewhere. Monocyte
purity (either conventionally gated or using our revised approach) was then calculated
as the percentage of traditionally gated monocytes that were recovered in the cognate
viSNE-defined gate.
Flow cytometry
For flow cytometric validation we used the following human antibody-fluorochrome
conjugates CD14 PE (clone 61D3, Tonbo Biosciences, San Diego, CA), CD16 FITC
(clone 3G8, BioLegend, San Diego, CA), CD11c PE-Cy7 (clone 3.9, BioLegend, San
Diego, CA), HLA-DR APC-Cy7 (clone L243, BioLegend, San Diego, CA), CD36 APC
(clone 5-271, BioLegend, San Diego, CA), CCR2 BV421 (clone K036C2, BioLegend,
San Diego, CA), CD3 PerCP-Cy5.5 (clone OKT3, Tonbo Biosciences, San Diego, CA),
CD19 PerCP-Cy5.5 (clone HIB19, Biolegend, San Diego, CA), CD56 PerCP-Cy5.5
(clone 5.1H11, BioLegend, San Diego, CA), CD66b PerCP-Cy5.5 (clone G10F5,
BioLegend, San Diego, CA). 2x106 frozen PBMCs were thawed at room temperature
and immediately upon thawing resuspended in Iftach’s media (RPMI + 10% FCS, 1% Lglut, 1% Pen/Strep, 1% NEAA, 1% Sodium Pyruvate) and incubated for 30 mins at
37oC. Cells were then spun at 350g for 10 minutes and resuspended in 1ml ice cold
hFACS buffer to block (PBS + 2mM EDTA + 5% BSA, + 10% human serum) for 30
minutes. Cells were stained in a final volume of 100 μl hFACS buffer containing each
antibody at a 1:200 dilution for 30 minutes on ice in the dark. Cells were washed twice
in 200 μl hFACS buffer and fixed for 10 minutes in 3.7% CH2O (Sigma) in PBS at room
temperature. Cells were washed twice more in 200 μl PBS and acquired on a BD LSRII
flow cytometer. Compensation was performed using single stain controls with
UltraComp eBeads (eBiosciences, San Diego, CA) using the cytometer software.
viSNE analysis of flow cytometry data
Flow cytometry analysis was performed in Cytobank. For healthy volunteer validation
viSNE was run on gated monocytes (live, singlets Lin-, CD14+HLA-DR+) using CD14,
CD16, CD11c, HLA-DR, CD36 and CCR2 with the default parameters. 100,000 total
monocytes were subsampled, selecting an equal number from each population. For
CAD samples, viSNE was run in the same manner as the healthy blood volunteers,
except that 400,000 total monocytes were used with equal subsetting between samples.
References
1.
Manichaikul A, Rich SS, Perry H, Yeboah J, Law M, Davis M, Parker M, Ragosta M,
Connelly JJ, McNamara CA, Taylor AM. A Functionally Significant Polymorphism in ID3 Is
Associated with Human Coronary Pathology. PLoS One. 2014;9(3):e90222.
2.
Chen H, Lau MC, Wong MT, Newell EW, Poidinger M, Chen J. Cytofkit: A Bioconductor
Package for an Integrated Mass Cytometry Data Analysis Pipeline. PLoS Comput Biol.
2016;12(9):e1005112.
3.
Maecker HT, Frey T, Nomura LE, Trotter J. Selecting fluorochrome conjugates for
maximum sensitivity. Cytometry A. 2004;62(2):169-173.