Embed Size (px)
Citation preview
Supplemental Materials
0
-10
3
- 10
4
- 10
5
-
Supplemental Figure 1 Multi-parameter flow cytometry panel and gating strategy
MDSC panel
Ab Fluorophore cat #
CD11b AF700 CD11b29
HLA-DR APC 559866
CD14 APC H7 560180
CD15 PerCP 555400
CD33 PE 555450
T cell panel
Ab Fluorophore Cat #
CD4 PerPC 347324
CD3 488 555332
CD25 PE 555432
CD8 PE cy7 557746
CD107a APC-H7 561343
Cd127 Af647 558598
A B
Degra
nula
tin
gC
D8 T
cells
T r
egula
tory
cells
MD
SC
s
C
D
E
Live/Dead UV
SS
C-A
HLA-DR
SS
C-A
CD33
CD
11b
CD14
CD
15
Live/Dead UV
SS
C-A
Live/Dead UV
SS
C-A
CD4
CD
8
CD4
CD
8
CD3
CD
8
CD3
CD
8
CD107a
SS
C-A
CD107a
SS
C-A
Live/Dead UV
SS
C-A
CD4
CD
8
CD127
CD
25
300k
-
200k
-
100k
-
0 -
0
-
10
3-
10
4-
10
5-
0
-
10
3-
10
4-
10
5-
0
-
10
3-
10
4-
10
5-
0
-
10
3-
10
4-
10
5-
300k
-
200k
-
100k
-
0 -
105 -
104 -
103 -
0 -
105 -
104 -
103 -
0 -
300k
-
200k
-
100k
-
0 - 0
-10
3
- 10
4
- 10
5
-
0
-10
3
- 10
4
- 10
5
-
0
-10
3
- 10
4
- 10
5
-
105 -
104 -
103 -
0 -
105 -
104 -
103 -
0 -
105 -
104 -
103 -
0 -
105 -
104 -
103 -
0 -
0
-10
3
- 10
4
- 10
5
-
0
-10
3
- 10
4
- 10
5
-
0
-10
3
- 10
4
- 10
5
-
0
-10
3
- 10
4
- 10
5
-
250k
-200k
-150k
-
0 -
100k
-50k -
250k
-200k
-150k
-
0 -
100k
-50k -
300k
-
200k
-
100k
-
0 -
300k
-
200k
-
100k
-
0 -
0
-
10
3-
10
4-
10
5-
0
-
10
3-
10
4-
10
5-
0
-
10
3-
10
4-
10
5-
105 -
104 -
103 -
0 -
105 -
104 -
103 -
0 -
Benign (n= 134)
Non-glial malignancy (n= 34)
Grade I/II (n= 37)
Grade III (n=11)
Grade IV (n=32)
Other (n= 11)
Benign (n= 134) Non-glial malignancy (n= 34) Grade I/II (n= 37)
Grade III (n=11) Grade IV (n=32) Other (n= 11)
Meningioma (n= 55)
Fibroma (n=1)
Pilocytic Astrocytoma (n=7)
Pituitary Adenoma (n=49)
Schwannoma (n=14)
Choroid Plexus Papiloma (n=1)
Craniopharhyngioma (n=2)
Gangliocytoma (n=1)
Hemangioblastoma (n=3)
LG Glioneuronal Tumor (n=1)
Metastasis (n=5)
Adenoid Cystic Carcinoma (n=1)
Breast Cancer Met (n= 5)
Lung Cancer Met (n=9)
Cervical Met (n=1)
Esophageal Met (n=2)
Lymphoma (n=3)
Myeloma (n=1)
Chondrosarcoma (n=1)
Embryol Tumor Grade IV (n=1)
Metastatic Neuroendocrine Carcinoma (n=1)
Plasma cell Neoplasm (n=1)
Recurrent HG Sarcoma (n= 1)
Renal Cell Carcinoma (n=1)
Small B cell Neoplasm (n=1)
Astrocytoma Grade I (n=2)
Astrocytoma Grade II (n=9)
Ependymoma (n=4)
Glioma Grade I (n=14)
Glioma Grade II (n= 1)
Oligoastrocytoma (n=1)
Oligodendroglioma Grade I (n=2)
Oligodendroglioma Grade II (n= 4)
Anaplastic Oligodendroglioma (n= 1)
Anaplastic Astrocytoma (n=8)
Meningioma (n=1)
Ganglioma (n=1)
Glioblastoma (n=32) Abscess (n=1)
Cholesterol Granuloma (n=1)
Cyst (n=1)
Epidermoid Cyst (n=1)
Gliosis (n=1)
Non-neoplastic, Gliosis (n=1)
Gliosis and Myelin loss (n=1)
May represent a Rathke cleft cyst (n=1)
Necrosis (n=1)
Pituitary inflammatory cyst (n=1)
Rathke Cleft Cyst (n=1)
Supplemental Figure 2 Categorizing patient samples for multi-parameter flow cytometry analysis.
CD33+ Iba1+ HLA-DR
MDSCs
CD33+/Iba1+/HLA-DR-/low
22 Matched Primary and Recurrent GBM Samples
Primary Recurrent
HLA-DR
Iba1
CD33
HLA-DR
Iba1
CD33
CD33+, Iba1+, HLA-DR+CD33+, Iba1+, HLA-DR low
CD33+, Iba1+, HLA-DRneg
CD33+, Iba1+, HLA-DR+CD33+, Iba1+, HLA-DR low
CD33+, Iba1+, HLA-DRneg
A
B
Supplemental Figure 3 Representative image of MDSC staining for one patient who had low MDSCs at primary resection and
increased levels upon recurrence.
Patient Characteristics (n=22)
Age at presentation (mean, SD) 60.6 (10.5)
Sex, Male 10 (45%)
Race, Non-Hispanic White 22 (100%)
BMI at presentation (mean, SD) 26.5 (4.5)
Obese at presentation 3 (15%)
Comorbidities
CHF (0%)
Peripheral vascular disease (0%)
Hypertension 5 (23%)
Diabetes mellitus (0%)
Hypothyroidism 1 (5%)
HIV/AIDS (0%)
Lymphoma (0%)
Rheumatoid arthritis, collagen vascular disease 1 (5%)
Alcohol abuse 1 (5 %)
Drug abuse 1 (5%)
Tobacco use (0%)
Ex-smoker 7 (32%)
Current smoker 4 (18%)
Never smoker 11 (50%)
C
Supplemental Figure 4 Immunofluorescence staining of MDSCs and myeloid cells in matched GBM resections
Low CD33 High CD330.0
0.1
0.2
0.3
CD
33
+a
r ea
of
t ot a
lt u
mo
ra
r ea
CD33 Grouping
MDSCs Grouping
Low MDSC High MDSC0.0
0.2
0.4
0.6
0.8
1.0M
DS
C+
ar e
a
of
CD
33
+a
r ea
A
B
A B
C D
Recurrent CD33 Survival after 2nd surgery
0 12 24 36 48 60 720
20
40
60
80
100 Low CD33 (n=11)
HIgh CD33 (n=11)
HR 0.3495% CI 0.12-0.95p=0.033
Survival after 2nd surgery
(Months)
Pe
r ce
nt
su
r vi v
al
Recurrent MDSC in CD33 survival
after 2nd surgery
0 12 24 36 48 60 720
20
40
60
80
100 Low MDSC (n=11)
High MDSC (n=11)
HR 4.9495% CI 1.76-13.8p<0.001
Survival after 2nd surgeryP
erc
en
ts
ur v
i va
l
Recurrent MDSC in CD33 time
between surgeries
0 12 24 360
20
40
60
80
100 Low MDSC (n=11)
High MDSC (n=11)
HR 1.9395% CI 0.80-4.69p=0.14
Time between 1st and 2nd surgery
(Months)
Pe
r ce
nt
su
r vi v
al
Recurrent MDSC in
PFS
0 12 24 360
20
40
60
80
100 low MDSC (n=11)
High MDSC (n=11)
HR 1.7295% CI 0.71-4.15p=0.22
PFS
(months)
Pe
rce
nt
su
r vi v
al
Supplemental Figure 5 Kaplan–Meier survival analysis of significant survival differences indicated in Table 2
Supplemental Figure 6 Longitudinal study of 6 newly diagnosed glioblastoma patients for MDSC and T cell levels.
A % MDSCs
0
5
10
15
20
25
%M
DS
Cs o
f Liv
e C
ells
B 2W 2M 4M 6M
% CD4 of CD3
20
40
60
80
B 2W 2M 4M 6M
100
% C
D4 T
cells
Of C
D3
0.0
% CD8 of CD3 T cells
0.0
0.2
0.4
0.6
0.8
B 2W 2M 4M 6M
% C
D8 T
cells
Of C
D3
B
C
label
signal
_di target Clone # Cat #
209Bi 226CD11b
(Mac-1)ICRF44 3209003B
170Er 590 CD3 UCHT1 3170001B
167Er 204 CD27 L128 3167006B
165Ho 1000 CD61 VI-PL2 3165010B
164Dy 133
CD15
(SSEA-
1)
W6D3 3164001B
163Dy 60CD56
(NCAM)
NCAM16.
23163007B
146Nd 508 CD8a RPA-T8 3146001B
159Tb 500 CD11c Bu15 3159001B
158Gd 142 CD33 WM53 3158001B
169Tm 100CD45R
AHI100 3169008B
89Y 800 CD45 HI30V 3089003B
153Eu 290 CD7 CD7-6B7 3153014B
151Eu 150 CD14 M5E2 3151009B
150Nd 70 CD161 HP-3G10 3159004B
149Sm 333 CD66aCD66a-
B1.13149008B
148Nd 70 CD16 3G8 3148004B
147Sm 300 CD20 2H7 3147001B
145Nd 70 CD4 RPA-T4 3145001B
143Nd 575CD25
(IL-2R)2A3 3149010B
142Nd 232 CD19 HIB19 3142001B
141Pr 216CD196
(CCR6)11A9 3141014A
139La 1000
CD107a
(LAMP1
)
H4A3 3151002B
174Yb 3000 HLA-DR L243 3174001B
155Gd 167CD279
(PD-1)EH12.2H7 3155009B
176Yb 200CD127
(IL-7Ra)A019D5 3176004B
PMN-MDSCs (minimum CD45+, CD66a-, CD3-, CD19-, CD20-,
CD56-, CD14-, CD11b +, CD15+) (better CD45+, CD66a-, CD3-,
CD19-, CD56-, HLADRlow/-, CD11b+, CD33+, CD14-, CD15+)
M-MDSCs (CD45+, CD66a-, CD3-, CD19-, CD20-, CD56-,
HLADRlow/-, CD11b+, CD33+, CD14+, CD15-)
e-MDSCs (CD45+, CD66a-, CD3-, CD19-, CD20-, CD56-, HLADR-
, CD33+)
Classical monocytes (CD45+, CD66a-, CD3-, CD19-, CD20-,
CD56-, CD14high, CD16-, HLA-DR+)
Intermediate monocytes (CD45+, CD66a-, CD3-, CD19-, CD20-,
CD56-, CD14high, CD16+)
Non-classical monocytes (CD45+, CD66a-, CD3-, CD19-, CD20-
, CD56-, CD14low/+, CD16+)
Myeloid dendritic cells (CD45+, CD66a-, CD3-, CD19-, CD20-,
CD56-, CD14-, CD11b+, CD11c+, HLA-DR+)
Monocyte-derived dendritic cells (CD45+, CD66a-, CD3-,
CD19-, CD20-, CD56-, CD14+, CD11b+, CD11c+, HLA-DR+,)
Natural killer cells 1 (CD45+, CD66a-, CD3-, CD19-, CD20-,
CD14-, CD11c-, CD56-, CD16+)
Natural killer cells 2 (CD45+, CD66a-, CD3-, CD19-, CD20-,
CD14-, CD11c-, CD56+, CD16-)
Granulocytes (CD3-, CD20-, CD 14-, CD11c-, CD45-, CD66a+)
Naïve CD8+ T cells (CD45+, CD66a-, CD3+, CD8a+, CD45RA+,
CD27+, CD127+)
Effector T killer cells (CD45+, CD66a-, CD3+, CD8a+,
CD45RA+, CD27-)
Activated T killer cells (CD45+, CD66a-, CD3+, CD8a+, HLA-
DR+)
Cytotoxic T cells (CD45+, CD66a-, CD3+, CD8a+, CD107a+)
Memory T killer cells (CD45+, CD66a-, CD3+, CD8a+, CD45RA-,
CD27+)
DP T cells (CD45+, CD66a-, CD3+, CD8a+, CD4+, CD27+,
CD161+)
Naïve CD4+ T cells (CD45+, CD66a-, CD3+, CD4+, CD45RA+,
CD25-, CD127+, CD27+)
Activated T helper cells (CD45+, CD66a-, CD3+, CD4+, HLA-
DR+)
Effector T helper cell (CD45+, CD66a-, CD3+, CD4+, CD45RA+/-
, CD25+, CD127-, CD27-)
Effector regulatory T helper cells (CD45+, CD66a-, CD3+,
CD4+, CD45RA-, CD25+, CD127-)
Resting regulatory T helper cells (CD45+, CD66a-, CD3+,
CD4+, CD45RA+, CD25+, CD127-)
Memory T helper cells (CD45+, CD66a-, CD3+, CD4+, CD45RA-
, CD25+, CD127+, CD27+)
Th17 cells (CD45+, CD66a-, CD3+, CD4+, CD161+, CD196+)
Naïve B cells (CD45+, CD66a-, CD3-, CD19+, CD20+, HLA-DR+,
CD27-)
Plasma B cells (CD45+, CD66a-, CD3-, CD19+, CD20-, HLA-DR-
, CD27+)
Memory B cells (CD45+, CD66a-, CD3-, CD19+, CD20+, HLA-
DR+, CD27+, CD196+)
Platelets (CD45-, CD61+)
Supplemental Figure 7 CyTOF study using 25 markers to identify immune cell populations that change over time.
A B
Supplemental Figure 8 CyTOF cleaning gates was performed to choose live single cells for analysis.
Defining clusters GBM timepoints
Supplemental Figure 9 tSNE cluster analysis of 30 clusters via histogram for expression of each marker is an unbiased approach to
identify cell populations
tSN
E 1
tSNE 2
Basal 2
Basal 3
Basal 4
Basal 6
Basal 7
Basal 9
T1_2
T1_ 3
T1_ 4
T1_ 6
T1_ 7
T1_ 9 T2_ 9
T2_ 7
T2_ 6
T2_ 4
T2_ 3
T2_ 2
Baseline Timepoint 1 Timepoint 2
Patie
nt
2P
atie
nt
3P
atie
nt
4P
atie
nt
6P
atie
nt
7P
atie
nt
9
tSNE 2
400 20-20-40
0
20
-20
-40
40
tSN
E 1
Cell-type specific cluster analysis
B cells
CD4 T cells
CD45 low
CD8 T cells
Dendritic cells
Double-negative
T cells
DP T cells
G-MDSCs
M-MDSCs
NK cells
Platelets
T/B cell markers
Supplemental Figure 10 tSNE plots identify immune shifts over time in GBM patients
A B
Supplemental Figure 11 Of 6 newly diagnosed GBM patients 3 had a good prognosis and decreasing MDSC while 3 had a poor
prognosis and increasing MDSCs
p2 ccf 3219 survival 1,211 days
0
2
4
6
8
10
12
%M
DS
Cs
p6 ccf 3578 survival 247 days
0
2
4
6
8
10
12
14%
MD
SC
s
p7 ccf 3620 survival- 725 days
0
2
4
6
8
10
12
%M
DS
Cs
p4 ccf 3545 survival- 583 days
0
2
4
6
8
10
12
%M
DS
Cs
p5 ccf 3132 survival- 398 days
0
10
20
30
40
50
%M
DS
Cs
p9 ccf 3765 survival-409 days
0
2
4
6
8
10
12
14
16
18
%M
DS
Cs
Red=Baseline
Green= Time-point 1
Blue= Time-point 2
Decreasing MDSCs p6,7 Decreasing MDSCs p2,6,7
Increasing MDSCs p4,5,9
Patient 4 Patient 2
CD4CD25CD19CD196CD107aCD45CD127HLA-DRCD3CD45RACD27CD61
CD8CD20CD16
CD66aCD161CD14CD7
CD279CD33
CD11cCD56CD15
KEY
Cluster
1
2
3
4
5
6
7
8
9
1
0
tSN
E1
tSNE2
tSN
E1
tSNE2
tSN
E1
tSNE2
Supplemental Figure 12 IDH1 mutant GBM patient has similar MDSCs changes over time compared to WT patients with a good
prognosis, but a distinct immune landscape from those with a poor prognosis
A
B
% L
ive c
ells
Patie
nt
4
IDH
1 W
T
Patie
nt
2
IDH
1 M
uta
nt
A%
Liv
e c
ells
% L
ive c
ells
% L
ive c
ells
0
2
4
6
8
m-DC mo-DC M-MDSC
B
0
2
4
6
8
m-DC mo-DC M-MDSC
0
5
10
NK1 NK2
NK1 NK2
0
5
10
FLT-3L
TimepointB 1 2
0
50
100
150
200
FL
T-3
L(p
g/m
l)
GM-CSF
TimepointB 1 2
0
100
200
300
GM
-CS
F (
pg/m
l)
B
Patient 4 (IDH WT)Patient 2 (IDH Mut)
1 2 B 1 2 B 1 2 B 1 2 B 1 2
B 1 2 B 1 2 B 1 2 B 1 2 B 1 2
Timepoints:
Timepoints:
Supplemental Figure 13 Dendritic cells and antigen-presenting cells are increased in a patient with a good prognosis
LGG vs GBM
LG
G1
LG
G2
LG
G3
LG
G4
LG
G5
LG
G6
LG
G7
LG
G8
LG
G9
LG
G10
LG
G11
GB
M B
asal 2
GB
M B
asal 4
GB
M T
1_4
GB
M T
1_2
GB
M T
2_4
GB
M T
2_4
Lung M
et
Menin
gio
ma
IL-12P70IL-5FractalkineIL-17AENA-78IL-1BIL-7IL-13Flt-3LGM-CSFIL-12P40CTACKBCA-1EGFIFNa2IFNyIL-15Eotaxin-16CKineG-CSFEotaxin-3I-309IL-10IL-3FGF-2IL-2MCP-4MCP-1TPOIL-16IL-28AMCP-3TNFBMIP-1aTGF-aIL-1aIL-33IL-23TSLPIP-10IL-1RAIL-8RANTESIL-18GRO alphaMIP-1dIL-9SCFLIFIL-20IL-21MIP-1BTRAILIL-4IL-6EotaxinSDF-1a+BTNFaMDCMCP-2sCD40LTARCVEGF-APDGF-AAPDGF-BB
-4
-2
0
2
4
0
2
4
6
8
10
Supplemental Figure 14 Cytokine array of glioblastoma and low-grade glioma patients identifies a unique cytokine signature for
glioblastoma patients as they progress through disease
Fo
ld c
hange n
orm
aliz
ed to r
ow
mean
% M
DS
Cs o
f Liv
e C
ells
LG
G1
LG
G2
LG
G3
LG
G4
LG
G5
LG
G6
LG
G7
LG
G8
LG
G9
LG
G10
LG
G11
GB
M B
asal 2
GB
M B
asal 4
GB
M T
1_4
GB
M T
1_2
GB
M T
2_4
GB
M T
2_4
Lung M
et
Menin
gio
ma
Supplemental Figure 15 . tSNE analysis of LGG and GBM samples at baseline allows for a visual identification of immune shifts
A
CGBM Basal 2 GBM Basal 3 GBM Basal 4
GBM Basal 6 GBM Basal 7 GBM Basal 9
LGG 1 LGG 2 LGG 3
B
GBM and LGG MDS analysis
MDS1
-0.3-0.5 0.0 0.3
-0.4
-0.2
0.0
0.2
MD
S2
tSN
E 2
GBM and LGG tSNE analysis
tSNE1
-25 0.0 25
-25
0.0
25
tSN
E 1
tSNE 2
Supplemental Figure 16 CyTOF analysis comparing six GBM patients and three LGG patients identifies significant differences in
immune status. Cluster Defining LGG vs GBM
Pro
port
ion
GBM poor prognosis
Survival <20 months post recurrence
LGG GBM favorable prognosis
Survival > 20 months post recurrence
B cells
CD4 T cells
CD45 low
CD8 T cells
Dendritic cells
DN T cells
M-MDSCs
Monocytes
Neutrophils
NK cells
Platelets
T cell/B cell markers
Supplemental Figure 17 Comparing GBM patients with varying prognoses to LGG patients identifies shifts in immune cell populations.
A
Variable No of patients HR (95% CI) P value
Age# (continuous) 22 1.04 (0.98-1.11) 0.22
MGMT promoter status
Methylated 9 1.00
Unmethylated 13 11.01 (2.18-55.61) 0.004
Sex
Male 10 1.00
Female 12 0.65 (0.21-1.97) 0.45
Steroid use##
Yes 13 1.00
No 9 0.51 (0.18-1.48) 0.22
MDSC
Low 11 1.00
High 11 6.78 (1.89-24.37) 0.003
# Age at time of primary diagnosis## Steroid use is defined as chronic use (>7 days) prior to secondary surgery.
IDH1 status is not included in the analysis, as only one patient in the cohort had a tumor
with a R132 IDH1 mutation.
Cox Regression analysis MDSCs
Supplemental Table 1 Cox regression analysis of MDSCs with Age, MGMt, Sex, and steroid use taken into account identify MDSCs as
the most significant variable in the model.
Cox Regression analysis CD33 Myeloid cells
Variable No of patients HR (95% CI) P value
Age# (continuous) 22 1.07 (1.00-1.13) 0.043
MGMT promoter status
Methylated 9 1.00
Unmethylated 13 10.05 (2.08-48.64) 0.004
Sex
Male 10 1.00
Female 12 0.53 (0.17-1.64) 0.27
Steroid use##
Yes 13 1.00
No 9 0.45 (0.15-1.34) 0.15
CD33
Low 11 1.00
High 11 0.12 (0.028-0.54) 0.006
# Age at time of primary diagnosis## Steroid use is defined as chronic use (>7 days) prior to secondary surgery.
IDH1 status is not included in the analysis, as only one patient in the cohort had a tumor
with a R132 IDH1 mutation.
Supplemental Table 2 Cox regression analysis of CD33+ cells with Age, MGMt, Sex, and steroid use taken into account identify
CD33+ cells as the most significant variable in the model.
32
Supplemental Figure Legends
Supplemental Figure 1. Multi-parameter flow cytometry panel and gating strategy. (A) Multi-
parameter flow cytometry analysis of myeloid cells and MDSCs stained with the antibodies and
fluorophores listed and analyzed on a BD LSRFortessa™ cell analyzer. (B) Each sample was run
through a T cell panel to determine lymphoid cell population quantities using the BD
LSRFortessa™. (C) Schematic depicting the gating strategy for MDSCs (live, HLA-DR-/low,
CD33, IBA1+), where CD15+ cells are considered granulocytic and CD14+ are monocytic MDSCs.
(D) Analysis of the lymphoid panel with T cell markers distinguishes CD8+ T cells as inactivated
(top row, CD107a-) or activated (bottom row, CD107a+). (E) T regulatory cells are distinguished
as CD4+ cells that are CD127- and CD25+.
Supplemental Figure 2. Categorizing patient samples for multi-parameter flow cytometry
analysis. Pie chart from Figure 1 depicts the categories of patient samples: benign, non-glial
malignancy, glioma grade I/II, glioma grade III, glioma grade IV, and other. Below the pie chart,
each category is broken down into the pathological diagnoses with the number of each listed to
the right.
Supplemental Figure 3. Representative image of MDSC staining for one patient who had
low MDSCs at primary resection and increased levels upon recurrence. (A) Schematic
representation of the experimental design outlining patients (n=22) with matched primary and
secondary resections triple-stained for CD33, Iba1, and HLA-DR to identify MDSCs. (B)
Representative images of a patient’s primary (left column) and recurrent (right column) samples
where MDSCs were increased in the recurrent specimen (identified by green masking in left and
right panels). (C) Clinical characteristics of the cohort of the 22 patients with matched primary and
secondary tissue samples.
33
Supplemental Figure 4. Immunofluorescence staining of MDSCs and myeloid cells in
matched GBM resections. (A) MDSC grouping from patients (n=22) was divided by the median
area of MDSCs in the total tumor area, which was used for Kaplan–Meier survival analysis. (B)
As a comparison, CD33+ myeloid cells were analyzed in a similar fashion where high and low
groups were defined by their median levels in the recurrent tissue. Data is represented as mean
and standard deviation.
Supplemental Figure 5. Kaplan–Meier survival analysis of significant survival differences
indicated in Table 2. Analysis of immunofluorescence staining of n=22 matched primary and
secondary resection GBM patients. High and low groups were identified based on median
expression of MDSCs at primary resection and at recurrence (Table 2). Here, we have plotted
the significant values identified for CD33 levels and MDSC levels, and survival was analyzed after
the 2nd surgery (A, B). Additionally, MDSC levels, and the time between surgeries and
progression free survival are shown (C, D).
Supplemental Figure 6. Longitudinal study of 6 newly diagnosed glioblastoma patients for
MDSC and T cell levels. (A) Schematic representation of the study design, where samples were
collected for n=6 patients over disease progression after being diagnosed with glioblastoma
(B,C,D) Over disease progression, patients were analyzed at baseline (B), 2 weeks post-op (2W),
2 months post-baseline (2M), and then every following 2 months. One out of six patients were
IDH1 Mutant (red). Multi-parameter flow cytometry for CD4 T cells (CD3+, CD4+, CD8-), CD8+ T
cells (CD3+, CD4-, CD8+), CD8+ T cells (CD3+, CD4-, CD8+). One-way ANOVA, used to compare
34
means across timepoints, yielded no significant differences (p<0.05). Data is represented as
mean and standard deviation
Supplemental Figure 7. CyTOF study using 25 markers to identify immune cell populations
that change over time. (A) A panel of 25 immune markers with heavy metal tags was used to
label cell populations. (B) The immune cell populations that are capable of being identified using
the 25 markers are listed with their marker selection.
Supplemental Figure 8. CyTOF cleaning gates were performed to choose live single cells
for analysis. DNA Ir191/193 was used to mark nucleated cells, which were then further gated for
intact cells by length parameters. Next, cells were gated for live/dead ln115, where negative cells
are live. Cells in this final gate were then subdivided to new FCS files for analysis.
Supplemental Figure 9. tSNE cluster analysis of 30 clusters via histogram for expression
of each marker is an unbiased approach to identify cell populations. In the longitudinal study
of newly diagnosed patients (n=6), 6 patients were analyzed by CyTOF at three timepoints each.
tSNE cluster analysis was used to define unique populations of cells. In this analysis, the 6
samples are combined for cluster analysis of each marker, and histograms show the expression
level of each marker in the cluster, which were then manually determined to be the immune cell
types labeled on the Y axis.
Supplemental Figure 10. tSNE plots identify immune shifts over time in GBM patients. (A)
t-distributed stochastic neighbor embedding (tSNE) plot identifies 30 unique populations that are
color coded among the patient samples (n=6) across all timepoints, representing a total of 18
samples. (B) Individual tSNE plots of each sample demonstrate the quantity of each cell
population by density of color-coded clusters over time.
35
Supplemental Figure 11. Of 6 newly diagnosed GBM patients, 3 had a good prognosis and
decreasing MDSC, while 3 had a poor prognosis and increasing MDSCs. Percent MDSCs
was assessed by flow cytometry in 6 GBM patients over time, where 3 patients had overall
decreasing MDSCs and a good prognosis (top row). The other 3 patients (bottom row) had overall
increasing MDSCs over time and a poor prognosis (bottom row).
Supplemental Figure 12. An IDH1-mutant GBM patient has similar MDSC changes over time
compared to WT patients with a good prognosis but a distinct immune landscape from
those with a poor prognosis. (A) tSNE analysis of the MDSC population from patients with
decreasing MDSCs (p2, 6, 7) and increasing MDSCs (p4, 5, 9) was utilized to determine whether
inclusion or exclusion of patient 2 (IDH1 mutant) altered the MDSC expression profile over time.
In this analysis, removing patient 2 from the analysis did not significantly alter the expression
profile of MDSCs (A), and the MDSC profile is still distinctly different between the two groups of
patients with differing prognoses. FlowSOM analysis of Patients 2 (IDH1 mutant, good prognosis,
decreasing MDSCs) and 4 (IDH1 WT, poor prognosis, increasing MDSCs) creates an unbiased
clustering of 10 groups, with each node of the clusters identifying the size of the cell population
and pie charts showing their expression of CyTOF markers, demonstrating distinct differences in
immune landscape at baseline (B).
Supplemental Figure 13. Dendritic cells and antigen-presenting cells are increased in a
patient with a good prognosis. (A) Manual gating of DC populations, M-MDSCs, and NK cells
from Patients 2 and 4 at baseline (B), timepoint 1 (1), and timepoint 2 (2), where B and 1 are at
the same point in time post-diagnosis and 2 is the final time point collected. (B) Multi-parameter
flow cytometry-based cytokine array where the serum levels (in pg/ml) of 65 cytokines were
36
examined. FLT-3L and GM-CSF were increased in Patient 2 over time. Data is represented as
mean and standard deviation
Supplemental figure 14. Cytokine array of GBM and LGG patients identifies a unique
cytokine signature for glioblastoma patients as they progress through disease. Heat map
representation of the pg/ml of each cytokine on the Y axis, where each row is internally normalized
to the highest value. Dendrogram clustering groups the cytokines based on hierarchical
clustering. MDSCs of each patient are plotted below their corresponding cytokine levels in the bar
chart.
Supplemental Figure 15. tSNE analysis of LGG and GBM samples at baseline allows for a
visual identification of immune shifts. (A) Multi-dimensional scaffold analysis of low-grade
glioma (LGG) and glioblastoma (GBM) patient samples from CyTOF analysis were compared to
determine whether large differences existed between samples based on their grouping within the
plot, where similar samples cluster together. (B) Combining CyTOF analysis of six GBM and three
LGG patient samples, tSNE analysis was performed to identify unique cell clusters. (C) tSNE
analysis of each patient individually is separated to identify immune shifts between GBM and
LGG.
Supplemental Figure 16. CyTOF analysis comparing six GBM patients and three LGG
patients identifies significant differences in immune status. (A) Unbiased tSNE clustering
analysis identifies 12 immune cell populations in n=6 GBM patients at baseline and n=3 LGG
patients at baseline. (B) Quantification of the immune cell populations comparing baseline GBM
samples to baseline low-grade glioma (LGG) samples identifies significant differences between
cohorts for DCs, NK cells, and mixed T-B marker cells. Statistics were determined by comparing
baseline to each timepoint using linear models of the data with t-test comparisons and Benjamini-
37
Hochberg adjust to control for multiple comparisons *p<0.05, ** p<0.001, ***p<0.0001. Graphs
represent data sets as median with 1st and 3rd quartiles.
Supplemental Figure 17. Comparing GBM patients with varying prognoses to LGG patients
identifies shifts in immune cell populations. (A) Baseline samples from n=6 patients in the
longitudinal CyTOF study were stratified into three with a good prognosis (p2, 6, 7) and three with
a poor prognosis (p4, 5, 9) and compared to n=3 LGG patients. These patients were then
compared to three low-grade glioma (LGG) patients at baseline using a pie chart with their
immune cell populations as defined by tSNE clustering and histogram defining the characteristics.
Supplemental Table 1. Cox regression analysis of MDSCs with age, MGMT status, sex, and
steroid use taken into account identifies MDSCs as the most significant variable in the
model. As a confirmation of the survival differences seen by log-rank tests in Figure 1E, F and
Table 2, a Cox regression model was used to test the effect of possible confounding clinical
variables. MDSCs remained the most significant predictor of survival (p=0.003); n=22.
Supplemental Table 2. Cox regression analysis of CD33 with age, MGMT status, sex, and
steroid use taken into account identifies CD33+ cells as the most significant variable in the
model. As a confirmation of the survival differences seen by log-rank tests in Figure 1E, F and
Table 2, a Cox regression model was used to test the effect of possible confounding clinical
variables. CD33+ myeloid cells also remained the most significant predictor of survival (p=0.006);
n=22.
