ArticleThe local microenvironment drives activation of
neutrophils in human brain tumorsGraphical abstractHighlightsd Neutrophils infiltrate human IDH WT gliomas and brain
metastasis in substantial numbers
d Pro-inflammatory pathways are enriched in brain tumor-
associated neutrophils (TANs)
d Brain TANs have a prolonged survival and
immunosuppressive and pro-angiogenic phenotype
d TNF-a and Ceruloplasmin secreted by tumor-associated
myeloid cells induce TAN phenotypesMaas et al., 2023, Cell 186, 4546–4566
October 12, 2023 ª 2023 The Authors. Published by Elsevier Inc.
https://doi.org/10.1016/j.cell.2023.08.043Authors
Roeltje R. Maas, Klara Soukup,
Nadine Fournier, ..., Roy T. Daniel,
Monika E. Hegi, Johanna A. Joyce
Correspondence
johanna.joyce@unil.ch
In brief
Comprehensive analyses of brain tumor-
associated neutrophils in humans and
mice reveal a multifaceted pro-tumoral
phenotype, which is induced by soluble
factors secreted by tumor-associated
myeloid cells.ll
OPEN ACCESS
llArticle
The local microenvironment drives
activation of neutrophils in human brain tumors
Roeltje R. Maas,1,2,3,4,5,6 Klara Soukup,1,2,3 Nadine Fournier,3,7,8,14 Matteo Massara,1,2,3,4,14 Sabine Galland,1,2,3,4,9
Mara Kornete,1,2,3 Vladimir Wischnewski,1,2,3,4 Joao Lourenco,3,7 Davide Croci,1,2,3 A´ngel F. A´lvarez-Prado,1,2,3,4
Damien N. Marie,1,2,3 Johanna Lilja,1,2,3 Rachel Marcone,3,7 Gabriel F. Calvo,10 Rui Santalla Mendez,1,2,3,4
Pauline Aubel,1,2,3,4 Leire Bejarano,1,2,3,4 Pratyaksha Wirapati,3,8 Iva´n Ballesteros,11 Andre´s Hidalgo,11,12
Andreas F. Hottinger,1,2,3,9 Jean-Philippe Brouland,13 Roy T. Daniel,4,6 Monika E. Hegi,4,5,6 and Johanna A. Joyce1,2,3,4,15,*
1Department of Oncology, University of Lausanne, Lausanne 1011, Switzerland
2Ludwig Institute for Cancer Research, University of Lausanne, Lausanne 1011, Switzerland
3Agora Cancer Research Centre Lausanne, Lausanne 1011, Switzerland
4L. Lundin and Family Brain Tumor Research Center, Departments of Oncology and Clinical Neurosciences, Centre Hospitalier Universitaire
Vaudois, Lausanne 1011, Switzerland
5Neuroscience Research Center, Centre Hospitalier Universitaire Vaudois, Lausanne 1011, Switzerland
6Department of Neurosurgery, Centre Hospitalier Universitaire Vaudois, Lausanne 1011, Switzerland
7Translational Data Science Group, Swiss Institute of Bioinformatics, Lausanne 1011, Switzerland
8Bioinformatics Core Facility, Swiss Institute of Bioinformatics, Lausanne 1011, Switzerland
9Department of Oncology, Centre Hospitalier Universitaire Vaudois, Lausanne 1011, Switzerland
10Department of Mathematics & MOLAB-Mathematical Oncology Laboratory, University of Castilla-La Mancha, Ciudad Real 13071, Spain
11Program of Cardiovascular Regeneration, Centro Nacional de Investigaciones Cardiovasculares Carlos III, Madrid 28029, Spain
12Vascular Biology and Therapeutics Program and Department of Immunobiology, Yale University School of Medicine, New Haven, CT
06519, USA
13Department of Pathology, Centre Hospitalier Universitaire Vaudois, University of Lausanne, Lausanne 1011, Switzerland
14These authors contributed equally
15Lead contact
*Correspondence: johanna.joyce@unil.ch
https://doi.org/10.1016/j.cell.2023.08.043SUMMARYNeutrophils are abundant immune cells in the circulation and frequently infiltrate tumors in substantial
numbers. However, their precise functions in different cancer types remain incompletely understood,
including in the brain microenvironment. We therefore investigated neutrophils in tumor tissue of glioma
and brain metastasis patients, with matched peripheral blood, and herein describe the first in-depth analysis
of neutrophil phenotypes and functions in these tissues. Orthogonal profiling strategies in humans and mice
revealed that brain tumor-associated neutrophils (TANs) differ significantly fromblood neutrophils and have a
prolonged lifespan and immune-suppressive and pro-angiogenic capacity. TANs exhibit a distinct inflamma-
tory signature, driven by a combination of soluble inflammatory mediators including tumor necrosis factor
alpha (TNF-ɑ) and Ceruloplasmin, which is more pronounced in TANs from brain metastasis versus glioma.
Myeloid cells, including tumor-associatedmacrophages, emerge at the core of this network of pro-inflamma-
tory mediators, supporting the concept of a critical myeloid niche regulating overall immune suppression in
human brain tumors.INTRODUCTION
Central nervous system tumors comprise primary and metasta-
tic malignancies and often confer a poor prognosis. Among
primary brain tumors, diffuse gliomas represent the most
aggressive types and are classified based on the occurrence
of isocitrate dehydrogenase 1 and 2 (IDH) mutations.1 IDH
mutant (mut) tumors are generally low-grade gliomas, whereas
IDHwild-type (WT) tumors represent high-grade aggressive glio-
blastomas (GBMs). Brain metastases (BrMs) are more frequent4546 Cell 186, 4546–4566, October 12, 2023 ª 2023 The Authors. Pu
This is an open access article under the CC BY license (http://creativeand originate predominantly from primary lung, breast, mela-
noma, and kidney tumors.2 Prognosis following the current
standard of care therapy remains poor,3,4 and ongoing trials to
identify new treatment strategies increasingly focus on immuno-
and targeted therapies, which have shown some efficacy in a
subset of BrMs but only very limited effects in gliomas.5–7 The
suboptimal efficacy of these treatments is likely driven in part
by immune-suppressive mechanisms operating in the brain.8
The brain tumor microenvironment (TME) is a critical regulator
of cancer progression and metastasis,9 with unique cell typesblished by Elsevier Inc.
commons.org/licenses/by/4.0/).
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4547
Article
ll
OPEN ACCESS Article(e.g., microglia [MG], astrocytes, and neurons), a specialized im-
mune composition,10,11 and physiological regulation by the
blood-brain barrier (BBB).12 The overall tumor immune land-
scape of human gliomas and BrMs has been recently re-
ported,10,11 with a particular focus on analyzing MG and mono-
cyte-derived macrophages (MDMs). Importantly, a substantial
proportion of tumor-associated neutrophils (TANs) among
brain-infiltrating immune cells was evident, specifically in human
IDH WT gliomas and BrMs.10
However, to date, only a handful of studies have explored
the roles of TANs in brain tumors, and mostly reported pro-tu-
moral roles in mousemodels of gliomas13–16 and BrM.17,18 Given
certain differences between mouse and human neutrophil
biology,19 a comprehensive and multifaceted investigation of
TANs in human brain tumors is of critical importance and
currently lacking. Moreover, considering the diversity of neutro-
phil functions in distinct organ environments, in both health20
and disease,21 an emphasis must be placed on understanding
the contribution of TANs to each specific tumor type and stage.22
Here, we present the first in-depth analysis of neutrophils in
patients with diverse brainmalignancies.We identified common-
alities and differences between neutrophils in blood vs. tumor
tissue and determined how primary vs. metastatic brain TMEs
co-opt these cells. We were thereby able to address the
following questions: (1) what determines neutrophil abundance
in human brain tumors, (2) where do TANs localize within the
brain TME and which cell types do they interact with, and (3)
how do specific tissue vs. tumor types dictate neutrophil pheno-
types and functionalities.
RESULTS
Neutrophils are more abundant in human BrMs than
primary gliomas and their phenotype is altered within
the brain TME
In this study, we investigated whether neutrophils are altered by
distinct brain TMEs in tumors of intracranial vs. extracranial
origin. We analyzed freshly resected human glioma and BrMs,
matched blood samples, and non-tumor brain tissue (epilepsy
lobectomies) using a range of complementary approaches
(flow cytometry [FCM], immunofluorescence [IF] staining forFigure 1. Neutrophils are more abundant in the TME of human BrMs th
(A) Experimental design and methodology for analysis of human brain tumor tiss
Abbreviations: BrM, brain metastasis; TANs, tumor-associated neutrophils; PBNs
RNA-seq, RNA sequencing; FACS, fluorescence-activated cell sorting; MEC, mic
species; HD, healthy donor.
(B) Neutrophil proportion among CD45+ immune cells in non-tumor (n = 11) and tum
41, nOther-BrM = 24) using FCM. Wilcoxon rank-sum test.
(C) Neutrophil-to-lymphocyte ratio measured in whole blood from HD (n = 12)
nLung-BrM = 25, nOther-BrM = 17). Wilcoxon rank-sum test.
(D) Density of neutrophils (per mm2) in the non-perivascular niche (non-PVN) and P
using IF as shown in (E), (F), and Figure S1C. Wilcoxon signed-rank test.
(E and F) (E) Representative IF images of neutrophils (CD45+, CD15+) in relation to
Scale bars: 100 mm, highlighted boxed areas are shown at higher magnification
(G) Fold-changemedian fluorescence intensity (MFI) inmatched PBNs and TANs f
HD PBNs (n > 14). Wilcoxon signed-rank test, comparing matched patient PBNs a
values: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
See also Figure S1 and Table S1A for clinical details.
4548 Cell 186, 4546–4566, October 12, 2023spatial TME analyses, RNA sequencing [RNA-seq], microenvi-
ronmental cultures [MECs], protein arrays, and various ex vivo
functional assays) (Figure 1A). This allowed us to comprehen-
sively interrogate the phenotype, transcriptome, and function-
ality of brain TANs and peripheral blood neutrophils (PBNs) for
the first time.
Analysis of 192 human samples by FCM revealed a significant
increase in relative TAN abundance in IDHWT gliomas and BrMs
compared with IDH mut gliomas and non-tumor tissue, with
the highest TAN proportions in BrMs (Figures 1B and S1A;
Table S1A; methodology for detailed sample processing and
gating strategy of different immune cell populations is described
in Maas et al.23). Similarly, the blood neutrophil-to-lymphocyte
ratio (NLR) was significantly higher in both IDH WT glioma and
BrM patients vs. healthy donors (HDs; Figures 1C and S1B;
Table S1A) and also elevated compared with IDHmut glioma pa-
tients. High NLR is associated with reduced patient survival in
various cancer types including brain tumors.24–27
Considering the increased relative abundance of PBNs in
IDHWT glioma and BrM patients, we investigated whether neu-
trophils infiltrate the tumor parenchyma or primarily remain
within blood vessels in the brain TME. By employing IF staining
of whole tissue sections, we analyzed neutrophil (CD15+
CD45+) localization relative to the vasculature (CD31+) and
demonstrated that neutrophils indeed penetrate the brain (tu-
mor) tissue (Figures 1D–1F). Quantifying neutrophil proximity to
blood vessels revealed that they localize at higher densities
within the perivascular niche (PVN = within 20 mm distance of
vessels) in brain tumor tissue (Figures 1D and S1C). Notably,
BrMs exhibited the highest spatial TAN density (Figure 1D), con-
firming their overall elevated abundance as determined by FCM
(Figure 1B).
Given the differential abundance of TANs between BrMs and
gliomas, we next asked whether neutrophils were phenotypically
altered (1) compared with HDs, (2) in different tissues (blood vs.
tumors), and (3) in distinct brain tumor types (gliomas vs. BrMs).
We first analyzed canonical neutrophil activation markers and
functional molecules, previously reported to be differentially ex-
pressed by TANs in other tumors.17,18,26,28–32 Using FCM, we
found that brain TANs—independent of tumor type—exhibited
a pronounced activation profile characterized by increasedan primary gliomas and are phenotypically altered
ue and matched blood.
, peripheral blood neutrophils; IF, immunofluorescence; FCM, flow cytometry;
roenvironmental culture; TME, tumor microenvironment; ROS, reactive oxygen
or tissue (nIDHmut glioma = 31, nIDHWT glioma = 70, nBreast-BrM = 15, nLung-BrM =
vs. brain tumor bearing patients (nIDH mut = 20, nIDH WT = 43, nBreast-BrM = 8,
VN in non-tumor (n = 5) and tumor tissue (nIDH mut = 12, nIDH WT = 15, nBrM = 27)
the vasculature (CD31+) in non-tumor, IDHmut andWT glioma, and a lung-BrM.
in (F). Scale bars: 20 mm.
rom tumor-bearing patients (nIDHmut > 10, nIDHWT = 19, nBrM > 31) normalized to
nd TANs only. Data in (B)–(D) and (G) are represented as mean ± SD and p.adj
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4549
Article
ll
OPEN ACCESS Articleexpression of CD11B, CD15, CD66B, and S100A9, and
decreased expression of CD62L, in parallel with marked alter-
ations in levels of the chemokine receptors CXCR1, CXCR2,
and CXCR4, when compared with matched PBNs (Figures 1G
and S1D). S100A9 was predominantly expressed by TANs (Fig-
ure S1E; IF analysis), and functionally, we observed lower
intracellular levels of matrix metalloproteinase 9 (MMP9) and
arginase 1 (Arg1) (Figures 1G and S1D), potentially indicating
their secretion into the TME, as reported in other contexts.33,34
TANs also expressed higher levels of programmed death-ligand
1 (PD-L1) (Figure 1G), which not only inhibits T cell18 but can also
delay neutrophil apoptosis.33,35 While these specific markers
exhibited some inter-individual variation, there were no marked
differences between TANs in BrMs vs. glioma, nor in PBNs
from patients vs. HDs (Figures 1G and S1D).
Brain neutrophils are enriched in inflammatory gene
signatures, which become more pronounced in BrMs
but not in gliomas
Considering these phenotypic differences, we next took an un-
supervised approach to analyze the transcriptome of matched
PBNs and TANs from brain tumor patients, PBNs from HD,
and neutrophils from non-tumor brain tissue, each isolated by
fluroescence-activated cell sorting (FACS) and then subjected
to RNA-seq (Table S1B). Unsupervised clustering and prin-
cipal-component analysis (PCA) revealed that TANs have a
distinct transcriptional profile from PBNs (Figures 2A and S2A),
while patient PBNs closely resemble HD PBNs. This broad sep-
aration was not influenced by treatment history, tumor grade, or
mutational status (Figure 2A). Notably, in BrM patients only, spe-
cific alterations were observed in PBNs vs. HD PBNs (Fig-
ure S2B; Tables S2A and S2B), with a small subset of these
changes also evident in the comparison of BrM TANs vs. HD
PBNs (Figure S2C; Tables S2A and S2C). Over-representationFigure 2. Neutrophil transcription is altered within the brain tissue env
(A) Unsupervised heatmap depicting the top 250 most variable neutrophil genes
(B) Euler diagram showing intersect of upregulated differentially expressed gene
(cutoff: p.adj < 0.05 for TANs and p < 0.005 for non-tumor brain neutrophils; log2
(C) Volcano plot showing DEGs in non-tumor brain neutrophils vs. HD PBNs. Hig
upregulated genes.
(D) Over-representation analysis (ORA) of pathways from Hallmark, KEGG, and
neutrophils, glioma, and BrM TANs vs. HD PBNs (intersect shown in B). Pro-infla
(E) Euler diagram depicting intersect of upregulated DEGs in lung-, breast-BrM
p < 0.005; LFC > 1).
(F) PC plot of neutrophil transcriptional profiles in non-(tumor) brain tissue, calcu
(G) Gene set enrichment analysis (GSEA) using Hallmark and Reactome databas
tumor brain neutrophils. (cutoff: p.adj < 0.05; NES > 2 in at least one of the tumo
(H) Dot plot depicting LFC of individual DEGs between BrM/glioma TANs vs. mat
BrM vs. glioma TANs (cutoff: p.adj < 0.05; LFC > 1 or <
1; dark blue, signif. in glio
PBNs only; green, signif. in BrM TANs vs. glioma TANs and BrM PBNs; light g
BrM TANs).
(I) ORA of top 20 most significant pathways from Hallmark, KEGG, and Reactom
(cutoff: p.adj < 0.05; NES > 0). Pro-inflammatory pathways in green.
(J) Heatmap depicting non-tumor neutrophils and brain-TANs aligned to top 10
Samples were hierarchically clustered and genes of interest are shown on right.
(K) Euler diagram depicting intersect between the top 100 DEGs upregulated in lu
primary lung cancer TANs vs. PBNs.
(L) ORA of significant pathways from Hallmark, KEGG, and Reactome databases
(K) (cutoff: p.adj < 0.05; NES > 0).
See also Figure S2 and Tables S1B, for clinical details, S2, and S3.
4550 Cell 186, 4546–4566, October 12, 2023analysis (ORA) revealed three significant pathways, all broadly
related to inflammatory signaling, as shared between BrM
PBNs and TANs vs. HD PBNs (Figure S2D; Table S2D).
While low in abundance, neutrophils in non-tumor brain tissue
showed a substantially different transcriptional profile compared
with HD PBNs, clustering with brain TANs (Figures 2A and S2A),
with most differentially expressed genes (DEGs) overlapping
with TANs (Figures 2B and S2E; Tables S2C, S2E, and S2F).
Several genes elevated in non-tumor brain neutrophils vs. HD
PBNs included those associated with altered metabolism
(OLR1/LOX1, ALOX15B, CH25H, AREG, PTN, and APOC2)
and multiple neurotrophic factors (NR4A1, NR4A2, NR4A3, RA-
B3IL1, SYNDIG1, TSPAN7, and RASGEF1C) (Figures 2C and
S2F; Tables S2F and S2G). This signature thus likely reflects
the adaptation of neutrophils to the molecular composition and
nutrient availability in the brain. These findings, along with previ-
ous studies investigating neutrophil phenotypes in different tis-
sues under homeostasis in mouse models,20 underscore the
contribution of distinct organ environments as a key determinant
shaping the neutrophil transcriptome.
To assess which transcriptional alterations were shared be-
tween non-tumor brain neutrophils and brain TANs vs. HD
PBNs (Figure 2B), ORA was performed, predominantly identi-
fying inflammatory pathways (Figure 2D; Table S2H). When
investigating the distinguishing features between brain TANs
vs. non-tumor brain neutrophils, we found significantly more
DEGs in breast- and lung-BrM TANs than in IDH mut and IDH
WT glioma TANs, indicating that glioma TANs undergo few tu-
mor-specific alterations (Figure 2E; Tables S3A–S3D). There
was no overlap between BrM and glioma TANs regardless of
mutational status (Figure 2E), which was corroborated by PCA
(Figure 2F). While breast- and lung-BrM TANs were further en-
riched in inflammatory and immune-activation signaling path-
ways (Figure 2G; Tables S3E and S3F), IDH mut, and especiallyironment
, genes of interest indicated on right.
s (DEGs) in BrM/glioma TANs, and non-tumor brain neutrophils vs. HD PBNs
fold-change (LFC) > 1).
hlighted are neurotrophic and metabolic genes within the top 20 significantly
Reactome databases on upregulated DEGs shared between non-tumor brain
mmatory pathways highlighted in red.
, IDH WT, and IDH mut glioma TANs vs. non-tumor brain neutrophils (cutoff:
lated based on top 1% most highly variable genes.
es on DEGs in lung-, breast-BrM, IDH mut, and IDHWT glioma TANs vs. non-
r types). Highlighted are ECM and pro-inflammatory pathways.
ched PBNs (cutoff: p.adj < 0.05). Highlighted are significant DEGs of interest in
ma TANs vs. BrM TANs and glioma PBNs; light blue, signif. in glioma TANs vs.
reen, signif. in BrM TANs vs. PBNs only, salmon pink, shared in glioma and
e databases on DEGs upregulated specifically in BrM-TANs depicted in (H)
most variable genes per ‘‘N’’ neutrophil cluster as defined by Zilionis et al.36
ng- and breast-BrM TANs vs. matched PBNs and the Zilionis et al. dataset of
on the 27 DEGs shared between breast- and lung-BrM TANs only, as shown in
ll
OPEN ACCESSArticleIDH WT TANs mostly exhibited increased extracellular matrix-
related signatures (Figure 2G; Tables S3G and S3H).
We then extracted a core DEG signature distinguishing BrM
from glioma TANs by intersecting DEGs identified from the com-
parison of BrM vs. glioma TANs (disease-specific alterations),
withDEGs in TANsvs.matchedPBNs (tissue-specific alterations).
While only 30 genes were higher in glioma TANs, 247 genes were
BrMTAN specific (Figures 2H, S2G, and S2H; Tables S3I and S3J)
and again substantially enriched in pro-inflammatory signaling
pathways, particularly tumor necrosis factor alpha (TNF-a)
signaling (Figure 2I; Table S3K). While there is substantial overlap
in DEGs between lung- and breast-BrM TANs (Figure 2E), their
divergence is driven by the enrichment of cell-cycle-associated
pathways in breast-BrM TANs and inflammatory signaling path-
ways in lung-BrM TANs (Figure S2I; Table S3L).
We next compared our data with publicly available single-cell
RNA-seq (scRNA-seq) datasets of human TANs in primary lung
cancer36,37 and melanoma-BrM.38 We found that gene expres-
sion signatures of pro-inflammatory and tumor-specific neutro-
phil clusters N4-5, as identified by Zilionis and colleagues in
lung cancer,36 were particularly enriched in lung-BrM TANs (Fig-
ure 2J; Table S3M), suggesting similarity between the microen-
vironments of primary lung cancer and lung-BrM. Similar obser-
vations were made by analyzing an independent lung-TAN
scRNA-seq dataset37 (Figure S2J; Table S3M). Comparison
with themelanoma-BrM scRNA-seq dataset showed that glioma
TANs are enriched in genes associated with interferon-respon-
sive clusters (Figure S2K), whereas BrM TANs correspond to
more pro-inflammatory interleukin (IL)-8- and calprotectin-high
clusters (Figure S2K; Table S3M). This integration underscores
the transcriptional heterogeneity of TANs not only between gli-
omas and BrMs with different grade and mutational background
(Figures 2J, S2J, and S2K) but also within these tumor types,
suggesting the presence of different subsets in each tissue.
To identify a unique BrM TAN signature, we compared the top
100 DEGs of breast- and lung-BrM TANs vs.matched PBNswith
those from primary lung cancer TANs36 (Figure 2K; Table S3N).
Indeed, lung cancer TANs shared more similarities with lung-
BrM vs. breast-BrM TANs (Figure 2K). ORA of 27 DEGs shared
only between breast- and lung-BrM TANs (Figure 2K), revealed
TNF-a signaling and broad pro-inflammatory signaling as signif-
icantly induced specifically in BrM TANs (Figure 2L; Table S3O).
Moreover, TANs in BrMs with TP53 or KRAS mutations
were transcriptionally distinct from TP53/KRAS-WT BrMs
and showed increased inflammatory signaling (Figure S2L;
Tables S3P and S3Q). This suggests that distinct mutations in
the cancer cells can transcriptionally imprint on TANs, either
directly or through induced alterations in the TME.39
Together, these data show that in addition to the brain tissue
environment, the local TME imprints a further layer of alterations
onto TANs, which is more pronounced in BrMs comparedwith gli-
omas,andpredominantlycharacterizedby inflammatorysignaling.
ROS release is elevated in circulating neutrophils from
BrM-bearing patients and is abrogated within the
brain TME
Considering the observed tissue and TME-specific alterations in
brain TANs from patients, we further investigated their functionalalterations using breast-BrM and GBM mouse models. These
fully immunocompetent models closely reflect the increased
TAN abundance observed in human tumors (Figure 3A), although
the overall proportion is lower in murine brain tumors. Notably,
the phenotype of TANs in BrMs and gliomas is similarly altered
for those markers conserved between species (Figure 3B). We
next examined the capacity of isolated murine PBNs and TANs
to produce reactive oxygen species (ROS)—a hallmark property
of activated neutrophils associated with cell killing,40–42 muta-
genesis,21 and T cell suppression.43,44 Interestingly, we found
that BrM and GBM TANs produced less ROS compared with
matched PBNs (Figure 3C).
Given certain differences in the murine and human neutrophil
ROS pathway,19 we sought to validate these findings in patient
samples. In functional assays, we found that PBNs from BrM pa-
tients released more ROS compared with glioma and HD PBNs
(Figures 3D, 3E, and S3A). Transcriptional analysis confirmed
this result, as BrM PBNs (but not glioma PBNs) showed positive
enrichment of ROS-associated pathways compared with HDs
(Figure S3B). Notably, TANs isolated from glioma and BrM sam-
ples produced substantially less ROS vs. matched PBNs (Fig-
ure 3F), similar to our findings in mouse models. This effect
was not influenced by prior treatment (Figure S3C). Consistently,
at the transcriptional level, brain TANs show pronounced dysre-
gulation of genes related to ROS production and oxidative ho-
meostasis (Figure 3G). Expression of NCF1, NCF2, and NCF4
(NADPH oxidase subunits) was decreased in TANs vs. PBNs,
while constituents of potent antioxidant response mechanisms
were induced (e.g., glutathione pathway members GCLC,
GCLM, GPX3, GPX8; peroxiredoxins PRDX1, PRDX2, PRDX4;
and superoxide dismutase subunits SOD1, SOD2, and SOD3)
(Figure 3G). By contrast, non-tumor brain neutrophils expressed
high levels of both pro- and anti-ROS-associated genes (Fig-
ure 3G), suggesting a modest ROS inhibitory potential of the
healthy brain environment.
To examine the effect of the brain TME on neutrophil ROS
levels ex vivo, we separated TANs (CD66B+) from the rest of
the TME (i.e., CD66B
 cells) and found that ROS levels increased
immediately (Figure 3H), indicating a partially reversible, tran-
sient component to the ROS impairment. Consequently, we
found that TAN ROS release was significantly increased in
response to phorbol 12-myristate 13-acetate (PMA), a potent
driver of neutrophil ROS production, and this activation over-
came the CD66B
 TME-driven ROS inhibition (Figure 3H), albeit
not to the same extent as for PBNs. To model PBN entry into the
brain TME using ex vivo cultures, we co-incubated PBNs or
TANs with the matched CD66B
 TME, which confirmed an im-
mediate and cell-ratio-dependent inhibition of ROS release
(Figures 3H and S3D). In sum, these data show that the brain
TME can counteract neutrophil ROS release. Given that ROS
act as potent cytotoxic agents,40–42 ROS neutralization may
represent a fast-acting defense mechanism of the brain TME.
Spatial TME analyses reveal the association of TANs
with PD-1+ CD8+ T cells and endothelial cells
We next assessed which cells of the brain TME can be
modulated by TANs based on their spatial proximity. First, we
addressed the potential effect of TANs on T cell function.Cell 186, 4546–4566, October 12, 2023 4551
AD
H
E F G
B C
(a.
u.)
IDH
 W
T
Figure 3. Murine brain TANs are phenotypically and functionally altered, as are human neutrophil ROS levels
(A) Proportion of Ly6G+ neutrophils of CD45+ immune cells in Ntv-a;iLy6GtdTomato glioblastoma (GBM)(n = 27) and iLy6G
tdTomato BrM(n = 40) mousemodels and non-
tumor bearing mouse brain (nnon-GBM = 27, nnon-BrM = 39). Wilcoxon rank-sum test.
(B) MFI of indicated markers in Ntv-a;iLy6GtdTomato GBM-bearing(n = 27) and iLy6G
tdTomato BrM-bearing(n = 40) mice. Wilcoxon signed-rank test.
(C) Detected ROS in arbitrary units (a.u.) in isolated PBNs and TANs in Ntv-a;iLy6GtdTomato GBM-bearing(n = 7) and iLy6G
tdTomato BrM-bearing(n = 7) mice.Wilcoxon
signed-rank test.
(D) Representative curves of ROS measured over time in PBNs from HD and brain tumor-bearing patients.
(E) Maximum ROS detected in PBNs from HD(n = 18), and tumor-bearing patients (nIDH mut = 12, nIDH WT = 20, nBrM = 51). Wilcoxon rank-sum test.
(F) ROS detected in matched PBNs and TANs (nIDH mut = 6, nIDH WT = 9, nBrM = 9). Wilcoxon signed-rank test.
(G) Heatmap depicting ROS-related significant DEGs in BrM and glioma TANs vs. matched PBNs.
(H) ROS detected in cocultures of matched PBNs and TANs with CD66B
 TME, activation by phorbol 12-myristate 13-acetate (PMA) (nIDH mut = 6, nIDH WT = 8,
nBrM = 9). Wilcoxon signed-rank test. Data in (A)–(F) and (H) are represented as mean ± SD. p.adj values in (A), (B), (E), and (H): *p < 0.05, **p < 0.01, ***p < 0.001,
****p < 0.0001. p values in (C) and (F): *p < 0.05, ***p < 0.001.
See also Figure S3.
ll
OPEN ACCESS ArticleNotably, in early-stage lung cancer, TANs show antigen-pre-
senting capacity.45 However, only few brain TANs express anti-
gen presentation molecules (e.g., HLA-ABC, HLA-DR, CD80,
and CD86) (Figure 4A), similar to matched PBNs (Figure S4A),4552 Cell 186, 4546–4566, October 12, 2023and significantly lower compared with MG and MDMs (Fig-
ure 4A). We also measured CD3+ T cell cytokine production in
coculture with PBNs or TANs and found no change in secretion
of cytotoxic mediators (Figures 4B and S4B), suggesting that
AC
F
I J
G H
D E
B
IDH WT
IDH WT
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4553
Article
ll
OPEN ACCESS Articlebrain TANs do not alter CD3+ T cell cytokine secretion, at least in
the ex vivo setting. However, given that TANs show increased
PD-L1 expression (Figure 1G), we also investigated the spatial
relationship of TANs and T cells in situ using sequential IF. Inter-
rogating IDHWT gliomas and BrMs, we found that CD4+ T helper
cells, Tregs, and CD8+ T cells are in closer proximity to TANs
compared with the average distance of all TME cells to this pop-
ulation (Figure 4C, indicated by the red line, p < 0.0001; Fig-
ure S4C). Since cell proximity is only suggestive of potential in-
teractions, we also assessed the expression of programmed
cell death protein 1 (PD-1) on CD8+ T cells and PD-L1 on
TANs. This revealed that PD-L1+ TANs (52% of TANs, based
on IF) are located closest to PD-1+ CD8+ T cells (Figures 4D,
S4D, and S4E), and the most abundant PD-L1+ cell type within
the PD-1+ CD8+ T cell niche is indeed TANs (Figure 4E). This sug-
gests that TANsmay have an immunosuppressive function in the
brain TME.
Considering the contributions of neutrophils to angiogenesis
regulation in tumors,46 as well as healthy47 and inflammatory
brain tissue,48,49 and that TANs reside proximal to the vascula-
ture (Figure 1D), we investigated the relationship between
TANs and vessels. We first characterized vessels stratified by
the presence of TANs in the PVN, based on the following mea-
surements: (1) form factor, indicating their degree of circularity
(Figure S4F), (2) size, and (3) proliferation. As described in other
contexts, larger and more deformed vessels (with a form fac-
tor < 1) can be less functional and leakier.50,51 Especially in the
brain, where the BBB tightly regulates molecular and cellular
transport under homeostatic conditions, the development of
leaky vessels can substantially impact the influx of cells.52
Notably, we found that vessel form factor decreases in IDH WT
gliomas and BrMs, especially in those vessels containing TANs
within their PVN (Figures 4F and S4G). Moreover, TANs, irre-
spective of brain tumor type, localized preferentially within the
PVN of larger vessels (Figure S4H), and in closer proximity to
proliferating endothelial cells (Figure 4G).
Additionally, we found significant enrichment of the Hallmark
angiogenesis pathway in both glioma and BrM TANs vs.
matched PBNs (Figures 4H and 4I). Expression of various pro-
angiogenic genes (e.g., VEGFA, THBD, and ICAM1) was higher
in TANs compared with other TME populations (Figure S4I).Figure 4. Brain TANs colocalize with PD-1+ CD8+ T cells and the vascu
(A) Proportion of cells expressing antigen presentation molecules in matched MD
(B) Cytokine concentration (pg/mL) in coculture supernatant of peripheral CD3+
Wilcoxon signed-rank test, p.adj values as shown.
(C) Distance of individual T cells (CD4+ T cells(n = 53,692), CD8
+ T cells(n = 54,495), Treg
indicates average distance of all cells(n = 2,571,834) to the nearest TAN. Wilcoxon r
(D) Representative lung-BrM image showing IF cell type quantification of TANs (
(E) Mean proportion of cell types identified in 20 mm radius around PD-1+ CD8+ T
(F) Mean form factor of individual vessels in presence (Pos) or absence (Neg) of ne
nIDH mut = 12, nIDH WT = 15, nBrM = 27). Wilcoxon signed-rank test.
(G) Distance (mm) of Ki67
(n = 191,077) and Ki67
+
(n = 14,718) endothelial cells to t
****p < 0.0001.
(H) GSEA of Hallmark angiogenesis pathway in glioma and BrM TANs combined
(I) Heatmap of individual genes from leading-edge Hallmark angiogenesis pathw
(J) Correlation between mean vessel form factor vs. normalized gene counts o
nIDH mut = 6, nIDH WT = 7, nBrM = 23) using the Pearson method. Data in (A)–(C), (F),
****p < 0.0001.
See also Figure S4.
4554 Cell 186, 4546–4566, October 12, 2023Several pro-angiogenic factors were also upregulated in non-tu-
mor brain neutrophils, albeit not to the same extent as in IDHWT
and BrM TANs (Figure 4I). This corresponds with the enrichment
of angiogenesis pathways in non-tumor brain neutrophils vs. HD
PBNs (Figure S2F). At the protein level, we found that the angio-
genesis-associated factor S100A953 was enriched in TANs vs.
PBNs (Figures 1G and S1E). MMP9, an important VEGF-associ-
ated pro-angiogenic metalloprotease,54 was detected at lower
protein levels (Figure 1G), potentially indicating its extracellular
release. Notably, MMP9 mRNA levels were highly elevated in
TANs compared with other TME populations, designating
TANs as a major MMP9 source in brain tumors (Figure S4I).
Moreover, we revealed a significant correlation between vessel
deformity (lower form factor) and the expression levels of several
pro-angiogenic genes in TANs (Figure 4J), MG, and MDMs (Fig-
ure S4J). Together, these findings support important roles for the
myeloid compartment in promoting aberrant angiogenesis in gli-
omas and BrMs, and potentially via distinct pathways in each
specific myeloid cell population.
The brain TMEhas both transient and permanent effects
on neutrophils and extends their lifespan
Given the pronounced phenotypic and transcriptional alterations
we observed in brain TANs compared with PBNs, and their po-
tential immunosuppressive and pro-angiogenic functions, we
considered it critical to assess their longevity and transit time
in the brain TME. While neutrophils are generally short-lived
cells, a recent study revealed that murine neutrophils in certain
organs have a substantially longer half-life/transit time than in
blood.20 Indeed, we found that isolated murine breast-BrM
TANs show significantly prolonged survival compared with
matched PBNs in culture (Figure 5A). Considering the limitations
of assessing neutrophil lifespan ex vivo, we next utilized a neutro-
phil fate-mapping mouse line (iLy6GtdTomato)20 to investigate
decay in vivo; the first time this has been applied to preclinical
cancer models. In both BrM and GBM, we indeed observed a
trend toward longer mean lifetimes of brain TANs compared
with healthy brain neutrophils (Figures 5B and 5C). Notably,
among BrM and GBM TANs, a subset of neutrophils showed a
further prolonged maximum lifespan, indicating that a larger
fraction of aged neutrophils may persist within the brain tumorlature
Ms, MG, and TANs (nIDH WT = 4, nBrM = 2). Wilcoxon signed-rank test.
T cells (pCD3) with PBNs/TANs for 96 h (nIDH mut = 2, nIDH WT = 8, nBrM = 6).
(n = 26,001)) to the nearest TAN in BrM(n = 20) and IDHWTglioma(n = 7). Dashed line
ank-sum test.
PD-L1
/PD-L1+) and CD8+ T cells (PD-1
/PD-1+). Scale bars: 50 mm.
cells(n = 20,697) in BrM(n = 20) and IDH WT glioma(n = 7).
utrophils in their PVN, based on IF as shown in Figures 1E and 1F (nNon-tumor = 5,
he nearest TAN (nIDH WT = 7, nBrM = 20). Wilcoxon rank-sum test, p value:
, compared with matched PBNs.
ay stratified by brain tissue type and PBN/TAN.
f angiogenesis-associated genes in brain (tumor) neutrophils (nnon-tumor = 3,
and (G) are represented as mean ± SD, p.adj value in (A), (C), and (F): *p < 0.05,
AD
G
H
I J K
E F
B C
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4555
Article
ll
OPEN ACCESS Articleniche (Figure S5A). Importantly, TAN lifespan was similarly pro-
longed in primary mammary fat pad TANs (Figure S5B), and
the presence of a growing tumor did not alter neutrophil lifespan
in any of the peripheral organs assessed, including blood
(Figure S5C).
Next, we analyzed the longevity of human TANs, by isolating
and culturing these cells in tumor MECs ex vivo over time. We
observed that up to 80% of TANs survive for 24 h, and up to
20% remain alive even after 48 h, which was similar across all
brain tumor types and independent of treatment history
(Figures 5D, S5D, and S5E). As expected, considerably fewer
viable PBNs remain at 24 h and almost all PBNs die within 48 h
(Figures 5D and S5D). Notably, PBNs cultured within the
CD66B
 TME exhibited significantly increased survival, reaching
a similar level to TANs (Figure 5D), indicating that the TME indeed
prolongs neutrophil longevity. Critically, separating TANs from
the TME milieu abrogated this longevity effect (Figure S5F),
underscoring the importance of TME-supplied factors.
Neutrophils are continuously released from the bone marrow
in vast numbers55–57; thus, programmed cell death is critical
for maintaining homeostasis. We therefore assessed PBN life cy-
cle stage at 24 h, via annexin V/DAPI staining. This showed that
PBNs cultured within TME MECs remain alive/pre-apoptotic for
longer compared with PBNs cultured alone (Figure 5E). Corrob-
orating these findings, a pronounced number of pro-apoptotic
genes were downregulated in TANs vs. PBNs in both BrMs
and gliomas, and conversely, several anti-apoptotic genes
were induced especially in BrM-TANs (Figure 5F). Non-tumor
brain neutrophils showed an intermediate phenotype between
PBNs and TANs, suggesting that the brain environment itself
already influences apoptosis pathways in neutrophils (Figure 5F).
Having discovered that the brain-TAN phenotype, ROS pro-
duction, and lifespan are altered in both patients and mouse
models, we next queried how the brain TME induces these
changes. Culturing patient PBNs for 24 h within the CD66B

TME broadly recapitulated the TANphenotype (Figure 5G).While
some alterations were induced in 24-h-old PBNs independent of
the presence of the TME (e.g., CD62L, CXCR4, CXCR2, and
S100A9), others were modulated only upon incubation withinFigure 5. The brain TME modulates both neutrophil survival and pheno
(A) Proportion of live neutrophils after 24 h based on annexin V viability staining
(B) Experimental design showing iLy6GtdTomato mice and proportion of tdTomato
brains over 8 days. A minimum of 3 mice from 6 independent experiments are sh
(C) Proportion of tdTomato+ neutrophils normalized to maximum in Ntv-a;iLy6Gtd
independent experiments are shown per time point, mean lifetime in days is sho
(D) Proportion of live TANs in tumors, matched PBNs and PBNs in coculture with
Wilcoxon signed-rank test.
(E) Proportion of PBNs at different viability stages, alone and in cocultures with m
(F) Heatmap of DEGs associated with apoptosis in PBNs vs. neutrophils from br
(G) MFI of phenotypic markers in PBNs over time (0 and 24 h) and in the presence
(H) Uniform manifold approximation and projection of 10-marker FCM panel (CD1
on matched TANs and PBNs in coculture with CD66B
 TME over 48 h. Trajecto
trajectory.
(I) Fold-change in 24 h survival of HD PBNs (n = 25) after culture inMEC-CM (nIDHm
and HD ID as random slope effect: ANOVA p value = 4.12 3 10
8.
(J) Fold-change MFI in HD PBN (n = 15) cultured in MEC-CM (nIDH mut = 4, nIDH W
(K) ROSmeasured in HD PBNs (n = 7) cultured in BrMMEC-CM (n = 13) and treate
(E), (G), and (I)–(K) are represented as mean ± SD. p.adj values in (D), (G), and (I)
See also Figure S5 and Table S1C for clinical information.
4556 Cell 186, 4546–4566, October 12, 2023the TME cultures (e.g., CD66B and PD-L1) (Figure 5G). TANs
maintain certain phenotypic alterations even when isolated
from the TME (Figure S5G). The time-sensitive transition of
PBNs to TANs was further confirmed through an unbiased
approach. Trajectory analysis on single cells analyzed for 10
TAN phenotype markers (by FCM) demonstrated that PBNs
cultured for 24 h in CD66B
 TME converge with TANs (Figure 5H,
black arrows), while PBNs cultured alone follow a separate tra-
jectory (red arrows).
Notably, both increased longevity and TAN phenotypes were
also induced in HD PBNs when cultured in the presence of condi-
tioned media (CM) from 24 h MECs isolated from different brain
tumors (Figures 5I and 5J; Table S1C for clinical information asso-
ciated with these MEC). In addition, BrM-MEC-CM suppressed
PBN-released ROS to a similar extent as diphenyleneiodonium
(DPI), a ROS inhibitor (Figure 5K). This effect was reversed by
the addition of PMA (Figure 5K). Similar findings regarding sup-
pression of ROS (Figure S5H) and prolonged neutrophil survival
(Figure S5I) were obtained for murine PBNs when cultured in
MEC-CM derived from breast-BrM or GBM mouse models.
Together this implicates soluble factors as the main mediators
of these processes, which are conserved between human and
mouse. It is furthermore indicative of combined mechanisms of
transitory and permanent alterations, which appear partially linked
to the persistence of neutrophils in the brain TME.
MEC-derived secreted factorsmodulate TANphenotype
and lifespan
We next asked which soluble factors are the major drivers of
these alterations and began by assessing tumor cells. We
derived cell lines from the patient MECs, which were confirmed
by whole exome sequencing analysis to represent tumor cells
(Figures S6A and S6B). These lines thus enabled an analysis of
the effect of soluble factors released exclusively by tumor cells
on neutrophils. Surprisingly, when culturing HD PBNs in CM
from tumor cell lines (termed TCM), we did not detect any life-
span extension, in contrast to the matched MEC-CM (Figure 6A;
Table S1C for clinical information associated with MEC-CM and
TCM). We therefore assessed whether CM derived from varioustype
in BrM-bearing(n = 10) mice. Wilcoxon signed-rank test, p value: **p < 0.01.
+ neutrophils normalized to maximum in non-tumor and BrM-bearing mouse
own per time point, mean lifetime in days is shown on right.
Tomato non-tumor and GBM-bearing mouse brains. Minimum of 3 mice from 3
wn on top.
CD66B
 TME population at 24 and 48 h (nIDH mut = 5, nIDH WT = 5, nBrM = 4).
atched MEC (nIDH mut = 2, nIDH WT = 2, nBrM = 5).
ain (tumor) tissue.
of CD66B
 TME (nIDH mut = 4, nIDH WT = 5, nBrM = 5). Wilcoxon signed-rank test.
1B, CD15, CD16, CD45, CD54, CD62L, CD66B, CXCR1, CXCR2, and CXCR4)
ries shown based on Slingshot: red arrow; PBN-trajectory, black arrow; TAN-
ut = 4, nIDHWT = 4, nBrM = 15). Mixed effects model withMEC type as fixed effect
T = 4, nBrM = 14) vs. control medium. Mixed effect model as in (I).
d with diphenyleneiodonium (DPI) or PMA.Wilcoxon rank-sum test. Data in (A)–
–(K): *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
AD
F
H I
G
E
B C
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4557
Article
ll
OPEN ACCESS Articlenormal cell types including human brain microvascular endothe-
lial cells (HBMECs), human umbilical vein endothelial cells
(HUVECs), astrocytes (HAs), and in vitro-derived unpolarized
HD MDMs (collectively referred to as non-TCM), confer the ca-
pacity to prolong neutrophil survival. Interestingly, only
HBMEC-CM induced prolonged neutrophil lifespan to some
extent, indicating that this is in part a brain endothelial-specific
effect (Figure S6C). Furthermore, neither TCM nor any of the
non-TCM recapitulated the complete phenotypic transition of
neutrophils induced by the complete MEC-CM (Figures 6B and
S6D). However, it is important to note that these tumor-naive
cell lines and in vitro-derived MDMs can be transcriptionally
distinct from their brain-tumor-associated counterparts, as pre-
viously shown for MDMs.10 Taken together, these results indi-
cate that factors produced by brain microenvironment-educated
populations, rather than by tumor cells (alone), drive the
observed neutrophil alterations.
To identify such factors, we performed a 1,000-protein array
analysis on a panel of 24 h MEC-CM (nIDH mut = 2, nIDH WT = 3,
nBrM = 9), matched TCM (nIDH mut = 1, nIDH WT = 3, nBrM = 2),
and non-TCM (HBMEC, HA, and MDM) compared with media
alone (control). Unsupervised clustering analysis showed that
MEC-CM overall clustered separately from TCM, non-TCM,
and control media, except for one outlier (Figure S6E). Most
BrM-MEC clustered closely together, with some distinction
from glioma-MEC. Subsetting for upregulated protein expres-
sion vs. control, and significant differential expression between
CM types, revealed 57 MEC-specific proteins (Figure 6C). Of
these, 51 were enriched in MEC-CM compared with both TCM
and non-TCM (Figure S6F; Tables S4A–S4C), thereby represent-
ing potential drivers of the TAN alterations.
Applying interactionnetworkanalysis to theseMEC-specificup-
regulated proteins revealed several clusters, centered on a highly
interconnectedpro-inflammatorycytokinegroup (Figure6D). TNF-
a, a potent recruiter and activator of neutrophils,58,59 is positioned
at the core of this cluster (Figures 6D and 6E). We also detected
Ceruloplasmin (CP), Glyoxalase 1 (GLO-1), and IL-9 at high levels
inMEC-CM (Figures 6D and 6E), which are important regulators of
reducedROSproduction60–62andprolongedneutrophil lifespan.63
Additionally,we foundhigh levelsofmajor neutrophil chemoattrac-
tants IL-664and IL-865 inMEC-CMspecifically (Figures6Dand6E).
Notably, we showed that when HD PBNs were exposed to
TNF-a, this resulted in a more TAN-like phenotype compared
with their untreated counterparts, including CD11B/CD66B up-Figure 6. Neutrophil alterations are induced by soluble factors in the b
(A) Fold-change in HD PBN survival (n = 12) after culture in MEC- or matched tu
medium pooled from 6 independent experiments. Mixed-effect model with CM ty
(B) Fold-change MFI of indicated markers in HD PBNs. Mixed effect model as in
(C) Euler plot depicting overlap in differentially detected proteins (cutoff: p < 0.05
(D) Protein interaction network of the 51 differentially expressed proteins in MEC
protein pathways shown in Figure 7C.
(E) Detection of the proteins indicated (a.u.) normalized to control media. Wilcox
(F) Fold-change MFI of indicated markers measured in HD PBNs (n = 8) treated
(G) Fold-change MFI of the indicated proteins in HD PBNs (n = 6) cultured in ME
(H) ROS detected (MFI) in HD PBNs (n = 10) treated with MEC-CM specific med
(I) Proportion of PBNs (n = 10) at different viability stages after treatment with ME
Data in (A), (B), and (E)–(I) are represented as mean ± SD. p.adj values in (A), (B),
See also Figure S6 and Tables S1C, for clinical information and S4.
4558 Cell 186, 4546–4566, October 12, 2023regulation and CD62L shedding, but no alterations in CXCR2/
CXCR4 levels (Figure 6F). Moreover, TNF-a signaling was en-
riched in the brain neutrophil-specific signature (Figure 2D), in
BrM and glioma TANs vs. matched PBNs (Figure S6G), and
even more pronounced in BrM-TANs vs. gliomas (Figure 2I).
Consequently, TNF-a inhibition in MEC-CM partially reversed
the TAN-like phenotype by suppressing CD11B/CD66B expres-
sion, whereas other inhibitors of pro-inflammatory signaling did
not convey such an effect, or even exacerbated it (Figure 6G).
CP was the only other soluble factor identified that was similarly
able to induce a partial TAN-like phenotype, with increased
CD11B/CD66B (Figures 6F and S6H). Combined treatment
with TNF-a and CP further increased the TAN-like phenotype,
with significant reduction in CXCR2 expression (Figure 6F).
Regarding the functional alterations observed in TANs, we found
decreased ROS levels in HD PBNs following CP exposure (Fig-
ure 6H). Similarly, the proportion of live PBNs increased when
incubated with CP (Figure 6I), mimicking the observations
made in TANs vs. PBNs in situ. Other MEC-CM specific soluble
factors (e.g., GLO-1, IL-6, IL-9, and TNF-a) did not confer such
functional alterations (Figures 6H and 6I).
In sum, key aspects of the phenotypic and functional alter-
ations in the PBN-to-TAN transition were induced by TNF-a
and CP treatment, alone or in combination (Figures 6F, 6H,
and 6I), indicating that multiple factors are required to drive the
full TAN phenotype.
Neutrophils interact predominantly with the myeloid
compartment
Lastly, we investigated which cell populations drive these
alterations in TAN phenotype and function. First, we character-
ized the cellular composition of ex vivo MEC. At 24 h, CD45

cells, and all major immune populations were still present
(Figures 7A and 7B), with a gradual reduction in absolute cell
counts that was more pronounced for CD45
 cells, MG and
MDMs vs. CD4+ and CD8+ T cells (Figure S7A)—but overall indi-
cating that all populations can potentially contribute to the ef-
fects observed in situ and upon exposure of PBNs to MEC-CM
ex vivo.
Second, we analyzed our brain TME RNA-seq data to identify
which cell types produce these MEC-CM-specific factors (Fig-
ure 6D). This revealed the myeloid compartment, including
MDMs, MG, and TANs themselves, as the highest expressers
of most MEC-specific pro-inflammatory cytokines (Figure 7C).rain TME
mor cell line CM (TCM, nIDH mut = 1, nIDH WT = 3, nmelanoma-BrM = 2) vs. control
pe as fixed effect and HD ID as random slope effect: ANOVA p value = 0.0011.
(A).
) between MEC-CM, TCM, and non-TCM.
-CM vs. TCM and non-TCM. Color codes based on the unbiased clustering of
on rank-sum test.
with TNF-a and/or CP. Wilcoxon signed-rank test.
C-CM (n = 3) and the indicated inhibitors. Wilcoxon signed-rank test.
iators. Wilcoxon signed-rank test, p value: *p < 0.05.
C-CM specific soluble factors.
and (E)–(G): *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
(legend on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566, October 12, 2023 4559
Article
ll
OPEN ACCESS ArticleThree myeloid-specific clusters emerged: (1) pro-inflammatory
factors produced mainly by TAMs (e.g., TNF, IL6, and IL10), (2)
factors associated with recruitment and myeloid activation pro-
duced predominantly by neutrophils (e.g.,CXCR2, IL9, IL11, and
CP), and (3) activating/recruitment factors produced by the
entire myeloid compartment (e.g., MMP10, MMP12, CXCL8,
S100A8, S100A9, and S100A12) (Figures 6D and 7C). The
TAM-specific inflammatory cluster was of interest as it contains
many factors also found at the core of the secreted protein
network (Figures 6D and 7C), suggesting a key role in shaping
the MEC-CM. By contrast, and consistent with our TCM exper-
iments, the CD45
 population does not appear to be a major
source of these cytokines specifically; however, it is a producer
of extracellular-matrix-associated factors as well as CP in BrMs
specifically (Figure 7C). Together, this led us to consider that the
myeloid compartment is a predominant source of soluble factors
driving brain-TAN alterations.
To explore the interactions within the myeloid compartment,
we assessed whether TANs have a close spatial relationship
with TAMs. Both TANs (Figures 1D, 1F, and S1C) and TAMs10
are predominantly localized in the PVN; however, their spatial
orientation in relation to each other has not been studied to
date across different brain tumors. In a cohort of IDHWT glioma
and BrM tissue samples, we quantified the cell types present
within a radius of 20 mm around TANs, using sequential IF. We
observed that TANs predominantly interact with other TANs
both inside and outside the PVN (Figures 7D and 7E). Interac-
tions with tumor cells were mostly observed in the non-PVN
area. By contrast, MG and MDMs represent the most abundant
non-TAN immune cells proximal to TANs, particularly in the BrM
PVN, where they account for 20% of the cellular neighbor-
hoods (Figure 7E). Querying our RNA-seq data for neutrophil
recruitment factors (Figure 7F), we found that TANs and TAMs
show the highest expression of these chemokines (CXCL8,
CXCL5, CXCL1, CSF3, CXCL2, and CCL20), with a progressive
increase in more malignant tumor types (IDH WT gliomas and
BrMs). Together, this revealed a highly interactive network within
the myeloid cell niche of brain tumors, which is especially pro-
nounced in the PVN. This myeloid network is not only crucial
for the PBN-to-TAN phenotypic and functional transition but
also for the recruitment of neutrophils to the brain TME itself.
DISCUSSION
In this study, we have analyzed the phenotypes and functions of
neutrophils across diverse human brain tumors, comprising
>190 clinical samples. By interrogating neutrophils in the periph-Figure 7. Neutrophils interact with the myeloid compartment
(A) Proportion cells of live cells in ex vivo MEC (nIDH mut = 5, nIDH WT = 7, nBrM = 1
(B) Mean proportions of immune cell populations as percent of CD45+ cells from
(C) Heatmap of gene expression levels in major TME populations for MEC-CM spe
CD8+ T cells. Clustering by overarching protein pathways based on Ward’s meth
(D) Representative IF image and cell type quantification by QuPath in lung-BrM s
P2RY12+,CD49D
), MDMs (CD45+,CD68/P2RY12+,CD49D+) and ‘‘Other’’ (all rem
(E) Mean proportion of cell types identified in 20 mm radius around TANs(n = 62,70
(F) Normalized log2-transformed expression of neutrophil-recruiting cytokines in
See also Figure S7.
4560 Cell 186, 4546–4566, October 12, 2023ery (PBNs) and the tumor (TANs), we found that brain TANs are
highly abundant in IDHWTgliomas and BrMs and have amarked
inflammatory phenotype compared with PBNs. This phenotype
shares similarities with TANs in primary lung cancer66 and in-
flammatory disease.67 Importantly, independent of tumor type,
and even in the absence of a tumor, neutrophils that infiltrate
into the brain adapt transcriptionally to this unique organ micro-
environment. The type of brain tumor confers additional, more
pronounced transcriptional alterations, with BrM TANs showing
the most substantial differences, including enrichment in
multiple pro-inflammatory signaling networks. While BrM TANs
share some transcriptional similarities to TANs in primary extra-
cranial tumors, additional pro-inflammatory pathways are
induced in BrM TANs, indicating an important contribution of
TAN imprinting by both the tumor type and the local brain
TME. Although neutrophil tissue specificity had previously
been demonstrated for other tissue types (mostly in murine
models), we demonstrate for the first time that this is also the
case in human brain tissue. This discovery that the disease-spe-
cific TME imprints a unique layer of alterations onto TANs, in
addition to brain-tissue-driven changes, shows interesting paral-
lels to analyses of other brain TME populations including MG,
MDMs10,11,68 and T cells.69,70 Our findings thereby underscore
the necessity to not only investigate neutrophils within the spe-
cific organ of interest but also by additional stratification into
distinct tumor subtypes.
We have addressed how the brain TME drives the observed
TAN alterations, phenotypes, and functions, through a series of
ex vivo assays and spatial analyses of patient samples, in parallel
with interrogation of neutrophil behavior in mouse models. One
notably pronounced effect of the brain TME on neutrophils, both
in direct contact and via soluble factors, is the increased lifespan
of TANs compared with matched PBNs. Prolonged TAN survival
has similarly been observed in the context of several extracranial
cancers.71,72 We found that secreted factors in the brain TME can
recapitulate the TAN phenotype and prolonged lifespan, when
supplied exogenously to PBNs, and block ROS production. In
extracranial tumors, both pro- and anti-tumoral effects of ROS
have been reported.40,41,43,73 In the intracranial setting, increased
ROS production by tumor cells led to enhanced cytotoxicity in a
glioma model,14 Similarly, administration of engineered chimeric
antigen receptor-neutrophils, with increased ROS production, re-
sulted in glioma cell killing.74 Taken together with our findings
herein, this supports the triggering of a tumor-protective response
by the brain TME via suppression of these damaging molecules.
Spatial analyses of the brain TME revealed perivascular and
myeloid cell niches, in which TANs interact with these specific3) at 0, 24 h (mean and SD).
cultures shown in (A).
cific upregulated proteins depicted in Figure 6D. T cells include both CD4+ and
od.
howing tumor cells (pan-cadherin+), TANs (CD45+,CD15+), MG (CD45+,CD68/
aining cells). Scale bars: 50 mm.
9) in BrM(n = 20) and IDH WT glioma(n = 7).
different brain TME populations.
ll
OPEN ACCESSArticlecell types. The myeloid niche drives neutrophil recruitment into
the brain via IL-8 and granulocyte-colony stimulating factor
(G-CSF,encoded by CXCL8 and CSF3, respectively). A previous
study in a BrM mouse model similarly identified tumor cell-
derived G-CSF as a recruiter of immunosuppressive PD-L1+
neutrophils.18 The contribution of myeloid cells was, however,
not investigated in this earlier study. Importantly, even though
the expression of recruitment factors is effectively higher in
myeloid cells, their overall abundance is lower by comparison
to tumor cells. Thus, glioma and BrM tumor cells may addition-
ally contribute to TAN recruitment. We also found angiogenesis
pathways enriched in TANs and closely associated with
deformed tumor vessels, which likely contribute to the contin-
uous efflux of cells from the peripheral circulation. Indeed, spatial
analyses combined with interrogation of secreted cytokines indi-
cated critical immune-suppressive mechanisms of TANs at play
within the brain TME. Of note, as peripheral T cells have not been
primed to recognize tumor cells, and brain tumor-infiltrating lym-
phocytes have an exhausted phenotype70 elucidating the immu-
nosuppressive contribution of human brain TANs in ex vivo as-
says is challenging. Nonetheless, in several extracranial
cancers, TANs have been demonstrated to similarly confer a
direct inhibitory effect on T cells through their PD-L1 expres-
sion.71,75 We show herein that immunosuppressive capacity ap-
pears to be a critical property of TANs in both gliomas and BrMs.
Investigation of the transition of PBNs to brain-TANs revealed
an intricate network of predominantly pro-inflammatory media-
tors, including TNF-a. Given that the brain TME is rich in secreted
TNF-a, andwith TANs (particularly in BrM) showing a clear eleva-
tion of TNF-a signaling pathways, our data indicate that this
cytokine is a critical mediator of the complex multifactorial
neutrophil alterations herein. In the context of inflammation and
cancer, TNF-a has been associated with induction of either an
anti-, 31 or pro-tumoral76 inflammatory TAN phenotype. Thus,
its role in instructing TANs is highly context dependent; however,
this function appears to be shared between gliomas and BrMs.
Moreover, the TAN alterations we identified can in part be attrib-
utable to CP, as well as TNF-a. These two soluble factors were
specifically found in MEC-CM, and RNA-seq analysis identified
myeloid cells as the major source of these and other pro-inflam-
matory mediators. Integrating their capacity to recruit neutro-
phils and induce the PBN-to-TAN conversion, positions myeloid
cells at the center of the TME networks critical for brain cancer
development.
In summary, this study reports the first multifaceted interroga-
tion of TANs across a large series of diverse human brain tumors.
This revealed prolonged survival and extensive pro-inflamma-
tory alterations of neutrophils within the brain tumormilieu, which
is associatedwith T cell-suppressive and pro-angiogenic pheno-
types. Moreover, we also present the first in vivo analyses of
neutrophil dynamics in cancer, taking advantage of neutrophil-
reportermice20 andmultiple immune-competent cancermodels.
Together, our findings highlight the potential benefit of targeting
solublemediators produced within themyeloid niche that dictate
the pro-inflammatory TAN phenotypes in brain cancers. Future
studies, including in preclinical models, will be of considerable
interest to investigate the efficacy of such therapeutic
approaches.Limitations of the study
In this study, we investigated TANs in both human and murine
brain tumors. As analyzing neutrophils in vivowithin patient brain
tumors is technically not possible, we developed strategies
enabling us to mimic the brain TME ex vivo. Nevertheless, there
are some technical limitations as there is no immune cell recruit-
ment in these MEC cultures and cell death gradually increases
over time. Moreover, we are aware that our isolation methods,
while carefully optimized and controlled, could activate TANs.
Thus, we assessed the phenotype, and functional alteration of
TANs both in isolated neutrophils and whole-tumor MEC cul-
tures. No evident differences were observed; however, we
cannot exclude that there are alterations to other functions
and/or phenotypes that were not investigated herein.
Using ex vivo MEC-CM experiments, we established that sol-
uble factors produced by the complete brain TME induce PBN-
to-TAN alterations. A protein array (1,000 proteins) helped
narrow down the responsible factors to 51 MEC-CM specific
mediators, from which 5 cytokines (TNF-a, CP, GLO-1, IL-6,
and IL-9) were studied in depth. None of these, either alone or
in combination, could induce the complete TAN phenotype, indi-
cating that additional factors are likely required. A larger unbi-
ased screen would thus be needed to assess the potential
contribution of other MEC-CM-specific molecules, including
lipids and metabolites. In addition, it is crucial to validate the in-
hibition of these TME-secreted factors in vivo, given the limita-
tions of ex vivo assays stated above.
Finally, the human and murine brain tumors used in this study
represented late-stage tumors. Thus, it is important to note that
neutrophil function and phenotype may differ at earlier stages of
brain tumor formation. In vivomodels will be required to address
the complex neutrophil-tumor crosstalk over the temporal
course of brain cancer development and progression.STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
d KEY RESOURCES TABLE
d RESOURCE AVAILABILITYB Lead contact
B Materials availability
B Data and code availability
d EXPERIMENTAL MODEL AND STUDY PARTICIPANT
DETAILS
B Human subjects
B Primary cell cultures
d METHOD DETAILS
B Human brain (tumor) enzymatic digestion
B Flow cytometry and FACS
B Immunofluorescence staining of tissue sections
B Image analysis and cell type identification
B Mouse tumor models
B Mouse tissue analysis
B Mathematical modeling of mean neutrophil lifetime
B Murine PBN and TAN isolation
B Human PBN and TAN isolationCell 186, 4546–4566, October 12, 2023 4561
ll
OPEN ACCESS ArticleB Reactive oxygen species (ROS) detection
B T cell cytokine array
B Annexin V apoptosis analysis
B UMAP and trajectory analysis
B Microenvironmental culture (MEC)-conditioned media
(CM) collection and analysis
B Conditioned media (CM) generation from cell lines and
primary cells
B CM education experiments
B Whole exome sequencing (WES)
B Protein array
B Recombinant protein treatment
B Software and visualization
d QUANTIFICATION AND STATISTICAL ANALYSIS
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.cell.
2023.08.043.
ACKNOWLEDGMENTS
We thank the Departments of Neurosurgery, Neuro-oncology, and Pathology at
CHUV Lausanne; Dr. Florian Klemm;members of Prof. Hegi’s lab; and Prof. Ron
Stoop for generous collaboration in obtaining and processing patient samples.
We thank Dr. Benoit Duc for helping collect HD blood and Dr. Spencer Watson
and Paola Guerrero for insightful advice on QuPath image quantification. We
gratefully acknowledge all current and former members of Prof. Joyce’s lab,
and members of the laboratories of Profs. Tacchini-Cottier, Meylan, De Palma,
Petrova and Pittet, for insightful discussions and advice. We thank V. Noguet
Brechbu¨hl (Mouse Pathology Facility) for expert preparation of tissue sections;
all members of the Epalinges and Agora animal facilities; P. Baumgartner and
A.-C. Thierry for advice and technical support on multiplex cytokine arrays;
and all members of the UNIL Flow Cytometry Facility for expert guidance in
designing panels and performing FAC-sorting. We used Biorender to prepare
the graphical abstract and several figure infographics. This research was sup-
portedby theBreastCancerResearchFoundation,CarigestFoundation, Fonda-
tion ISREC, Ludwig Institute for Cancer Research, Swiss Bridge Award, Swiss
National Science Foundation Advanced Grant TMAG-3_209224, and University
of Lausanne (to J.A.J.). K.S. was funded by an Erwin-Schro¨dinger Fellowship
from the Austrian Science Fund (FWF, J4343-B28); M.M. received funding
from AIRC and European Union’s Horizon 2020 research and innovation
programme under the Marie Sk1odowska Curie (grant agreement 800924);
S.G. was supported by Fondation Leenaards and Fondation ISREC;
A.F.A.-P. was supported by an EMBO Long-Term Postdoctoral Fellowship
(EMBO ALTF 654-2019) and a Marie Sk1odowska Curie Actions Individual
Fellowship (MSCA-IF 890933); G.F.C. is supported by MCIN/AEI/10.13039/501
100011033 (grant PID2019-110895RB-I00) and by Junta de Comunidades de
Castilla-LaMancha (SBPLY/19/180501/000211);R.S.M. is supportedby theEu-
ropean Union’s Horizon 2020 research and innovation program under the Sk1o-
dowskaCurie grant agreement (grant agreement 955951, Evomet ITN); L.B.was
supported by the European Molecular Biology Organisation (ALTF 1193-2018)
and the Human Frontier Science Program (LT000038/2019); A.H. is funded by
FET-OPEN (no. 861878) from theEuropeanCommission. TheCNIC is supported
by the MCIN and the Pro CNIC Foundation and is a Severo Ochoa Center of
Excellence (CEX2020-001041-S). Our particular gratitude is extended to all pa-
tients and healthy volunteers who generously donated tissue for analyses, and
without whom this study would not have been possible.
AUTHOR CONTRIBUTIONS
Conceptualization, R.R.M., K.S., and J.A.J.; methodology, R.R.M., K.S., M.M.,
S.G., M.K., and G.F.C.; formal analysis, R.R.M., N.F., J. Lourenco., A.F.A.-P.,
R.M., P.W., and G.F.C.; investigation, R.R.M., K.S., M.M., S.G., M.K., V.W.,4562 Cell 186, 4546–4566, October 12, 2023D.C., A.F.A.-P., D.N.M., J. Lilja, R.S.M., P.A., and L.B.; resources, A.H., I.B.,
A.F.H., J.-P.B., R.T.D., and M.E.H.; writing – original draft, R.R.M., K.S.,
S.G., and J.A.J.; writing – review & editing, all authors reviewed, edited, or
commented on the manuscript; visualization, R.R.M., N.F., S.G., K.S., and
G.F.C.; supervision, J.A.J.; funding acquisition, J.A.J., R.R.M., and K.S.
DECLARATION OF INTERESTS
D.C. has received consulting fees from Seed Biosciences S.A.; P.W. has pro-
vided consulting for Almax, Bayer, Sanofi, and Genentech. A.H. is a paid
consultor for Flagship Pioneering, Inc. for matters unrelated to this study;
M.E.H. has an advisory role at TME Pharma; J.A.J. received an honorarium
for speaking at a Bristol Meyers Squibb research symposium and previously
served on the scientific advisory board of Pionyr Immunotherapeutics.
Received: October 26, 2022
Revised: July 11, 2023
Accepted: August 31, 2023
Published: September 27, 2023
REFERENCES
1. Louis, D.N., Perry, A., Wesseling, P., Brat, D.J., Cree, I.A., Figarella-
Branger, D., Hawkins, C., Ng, H.K., Pfister, S.M., Reifenberger, G.,
et al. (2021). The 2021 WHO classification of tumors of the central ner-
vous system: a summary. Neuro. Oncol 23, 1231–1251. https://doi.org/
10.1093/neuonc/noab106.
2. Achrol, A.S., Rennert, R.C., Anders, C., Soffietti, R., Ahluwalia, M.S.,
Nayak, L., Peters, S., Arvold, N.D., Harsh, G.R., Steeg, P.S., et al.
(2019). Brain metastases. Nat. Rev. Dis. Primers 5, 5. https://doi.org/
10.1038/s41572-018-0055-y.
3. Sperduto, P.W., Mesko, S., Li, J., Cagney, D., Aizer, A., Lin, N.U., Nesbit,
E., Kruser, T.J., Chan, J., Braunstein, S., et al. (2020). Survival in patients
with brain metastases: summary report on the updated diagnosis-spe-
cific graded prognostic assessment and definition of the eligibility quo-
tient. J. Clin. Oncol. 38, 3773–3784. https://doi.org/10.1200/JCO.
20.01255.
4. Mohile, N.A., Messersmith, H., Gatson, N.T., Hottinger, A.F., Lassman,
A., Morton, J., Ney, D., Nghiemphu, P.L., Olar, A., Olson, J., et al.
(2022). Therapy for diffuse astrocytic and oligodendroglial tumors in
adults: ASCO-SNO guideline. J. Clin. Oncol. 40, 403–426. https://doi.
org/10.1200/JCO.21.02036.
5. Cloughesy, T.F., Mochizuki, A.Y., Orpilla, J.R., Hugo, W., Lee, A.H., Da-
vidson, T.B., Wang, A.C., Ellingson, B.M., Rytlewski, J.A., Sanders, C.M.,
et al. (2019). Neoadjuvant anti-PD-1 immunotherapy promotes a survival
benefit with intratumoral and systemic immune responses in recurrent
glioblastoma. Nat. Med. 25, 477–486. https://doi.org/10.1038/s41591-
018-0337-7.
6. Goldberg, S.B., Schalper, K.A., Gettinger, S.N., Mahajan, A., Herbst,
R.S., Chiang, A.C., Lilenbaum, R., Wilson, F.H., Omay, S.B., Yu, J.B.,
et al. (2020). Pembrolizumab for management of patients with NSCLC
and brain metastases: long-term results and biomarker analysis from a
non-randomised, open-label, phase 2 trial. Lancet Oncol. 21, 655–663.
https://doi.org/10.1016/S1470-2045(20)30111-X.
7. Desbaillets, N., and Hottinger, A.F. (2021). Immunotherapy in glioblas-
toma: a clinical perspective. Cancers (Basel) 13. https://doi.org/10.
3390/cancers13153721.
8. Sampson, J.H., Gunn, M.D., Fecci, P.E., and Ashley, D.M. (2020). Brain
immunology and immunotherapy in brain tumours. Nat. Rev. Cancer
20, 12–25. https://doi.org/10.1038/s41568-019-0224-7.
9. Quail, D.F., and Joyce, J.A. (2017). Themicroenvironmental landscape of
brain tumors. Cancer Cell 31, 326–341. https://doi.org/10.1016/j.ccell.
2017.02.009.
10. Klemm, F., Maas, R.R., Bowman, R.L., Kornete, M., Soukup, K., Nassiri,
S., Brouland, J.P., Iacobuzio-Donahue, C.A., Brennan, C., Tabar, V., et al.
ll
OPEN ACCESSArticle(2020). Interrogation of the microenvironmental landscape in brain tu-
mors reveals disease-specific alterations of immune cells. Cell 181,
1643–1660.e17. https://doi.org/10.1016/j.cell.2020.05.007.
11. Friebel, E., Kapolou, K., Unger, S., Nu´n˜ez, N.G., Utz, S., Rushing, E.J.,
Regli, L., Weller, M., Greter, M., Tugues, S., et al. (2020). Single-cell map-
ping of human brain cancer reveals tumor-specific instruction of tissue-
invading leukocytes. Cell 181, 1626–1642.e20. https://doi.org/10.1016/
j.cell.2020.04.055.
12. Andersen, B.M., Faust Akl, C., Wheeler, M.A., Chiocca, E.A., Reardon,
D.A., and Quintana, F.J. (2021). Glial and myeloid heterogeneity in the
brain tumour microenvironment. Nat. Rev. Cancer 21, 786–802. https://
doi.org/10.1038/s41568-021-00397-3.
13. Magod, P., Mastandrea, I., Rousso-Noori, L., Agemy, L., Shapira, G.,
Shomron, N., and Friedmann-Morvinski, D. (2021). Exploring the longitu-
dinal glioma microenvironment landscape uncovers reprogrammed pro-
tumorigenic neutrophils in the bone marrow. Cell Rep. 36, 109480.
https://doi.org/10.1016/j.celrep.2021.109480.
14. Yee, P.P., Wei, Y., Kim, S.Y., Lu, T., Chih, S.Y., Lawson, C., Tang, M., Liu,
Z., Anderson, B., Thamburaj, K., et al. (2020). Neutrophil-induced ferrop-
tosis promotes tumor necrosis in glioblastoma progression. Nat. Com-
mun. 11, 5424. https://doi.org/10.1038/s41467-020-19193-y.
15. Liang, J., Piao, Y., Holmes, L., Fuller, G.N., Henry, V., Tiao, N., and de
Groot, J.F. (2014). Neutrophils promote the malignant glioma phenotype
through S100A4. Clin. Cancer Res. 20, 187–198. https://doi.org/10.1158/
1078-0432.CCR-13-1279.
16. Jiguet-Jiglaire, C., Boissonneau, S., Denicolai, E., Hein, V., Lasseur, R.,
Garcia, J., Romain, S., Appay, R., Graillon, T., Mason, W., et al. (2022).
Plasmatic MMP9 released from tumor-infiltrating neutrophils is predic-
tive for bevacizumab efficacy in glioblastoma patients: an AVAglio ancil-
lary study. Acta Neuropathol. Commun. 10, 1. https://doi.org/10.1186/
s40478-021-01305-4.
17. Liu, Y., Kosaka, A., Ikeura, M., Kohanbash, G., Fellows-Mayle, W.,
Snyder, L.A., and Okada, H. (2013). Premetastatic soil and prevention
of breast cancer brain metastasis. Neuro. Oncol 15, 891–903. https://
doi.org/10.1093/neuonc/not031.
18. Zhang, L., Yao, J., Wei, Y., Zhou, Z., Li, P., Qu, J., Badu-Nkansah, A.,
Yuan, X., Huang, Y.W., Fukumura, K., et al. (2020). Blocking immunosup-
pressive neutrophils deters pY696-EZH2-driven brain metastases. Sci.
Transl. Med. 12. https://doi.org/10.1126/scitranslmed.aaz5387.
19. Eruslanov, E.B., Singhal, S., and Albelda, S.M. (2017). Mouse versus hu-
man neutrophils in cancer: a major knowledge gap. Trends Cancer 3,
149–160. https://doi.org/10.1016/j.trecan.2016.12.006.
20. Ballesteros, I., Rubio-Ponce, A., Genua, M., Lusito, E., Kwok, I., Ferna´n-
dez-Calvo, G., Khoyratty, T.E., van Grinsven, E., Gonza´lez-Herna´ndez,
S., Nicola´s-A´vila, J.A´., et al. (2020). Co-option of neutrophil fates by tis-
sue environments. Cell 183, 1282–1297.e18. https://doi.org/10.1016/j.
cell.2020.10.003.
21. Hedrick, C.C., and Malanchi, I. (2022). Neutrophils in cancer: heteroge-
neous and multifaceted. Nat. Rev. Immunol. 22, 173–187. https://doi.
org/10.1038/s41577-021-00571-6.
22. Quail, D.F., Amulic, B., Aziz, M., Barnes, B.J., Eruslanov, E., Fridlender,
Z.G., Goodridge, H.S., Granot, Z., Hidalgo, A., Huttenlocher, A., et al.
(2022). Neutrophil phenotypes and functions in cancer: a consensus
statement. J. Exp. Med. 219. https://doi.org/10.1084/jem.20220011.
23. Maas, R.R., Soukup, K., Klemm, F., Kornete, M., Bowman, R.L., Bedel,
R., Marie, D.N., A´lvarez-Prado, A´.F., Labes, D., Wilson, A., et al. (2021).
An integrated pipeline for comprehensive analysis of immune cells in hu-
man brain tumor clinical samples. Nat. Protoc. 16, 4692–4721. https://
doi.org/10.1038/s41596-021-00594-2.
24. Cupp, M.A., Cariolou, M., Tzoulaki, I., Aune, D., Evangelou, E., and Ber-
langa-Taylor, A.J. (2020). Neutrophil to lymphocyte ratio and cancer
prognosis: an umbrella review of systematic reviews and meta-analyses
of observational studies. BMC Med. 18, 360. https://doi.org/10.1186/
s12916-020-01817-1.25. Starzer, A.M., Steindl, A., Mair, M.J., Deischinger, C., Simonovska, A.,
Widhalm, G., Gatterbauer, B., Dieckmann, K., Heller, G., Preusser, M.,
et al. (2021). Systemic inflammation scores correlate with survival prog-
nosis in patients with newly diagnosed brain metastases. Br. J. Cancer
124, 1294–1300. https://doi.org/10.1038/s41416-020-01254-0.
26. Shaul, M.E., and Fridlender, Z.G. (2019). Tumour-associated neutrophils
in patients with cancer. Nat. Rev. Clin. Oncol. 16, 601–620. https://doi.
org/10.1038/s41571-019-0222-4.
27. Zhang, L., Hu, Y., Chen,W., Tian, Y., Xie, Y., and Chen, J. (2020). Pre-ste-
reotactic radiosurgery neutrophil-to-lymphocyte ratio is a predictor of the
prognosis for brain metastases. J. Neurooncol. 147, 691–700. https://
doi.org/10.1007/s11060-020-03477-w.
28. Coffelt, S.B., Kersten, K., Doornebal, C.W., Weiden, J., Vrijland, K., Hau,
C.S., Verstegen, N.J.M., Ciampricotti, M., Hawinkels, L.J.A.C., Jonkers,
J., et al. (2015). IL-17-producing gd T cells and neutrophils conspire to
promote breast cancer metastasis. Nature 522, 345–348. https://doi.
org/10.1038/nature14282.
29. Wellenstein, M.D., Coffelt, S.B., Duits, D.E.M., van Miltenburg, M.H.,
Slagter, M., de Rink, I., Henneman, L., Kas, S.M., Prekovic, S., Hau,
C.S., et al. (2019). Loss of p53 triggers WNT-dependent systemic inflam-
mation to drive breast cancer metastasis. Nature 572, 538–542. https://
doi.org/10.1038/s41586-019-1450-6.
30. Peng, Z., Liu, C., Victor, A.R., Cao, D.Y., Veiras, L.C., Bernstein, E.A.,
Khan, Z., Giani, J.F., Cui, X., Bernstein, K.E., et al. (2021). Tumors exploit
CXCR4(hi)CD62L(lo) aged neutrophils to facilitatemetastatic spread. On-
coimmunology 10, 1870811. https://doi.org/10.1080/2162402X.2020.
1870811.
31. Finisguerra, V., Di Conza, G., Di Matteo, M., Serneels, J., Costa, S.,
Thompson, A.A., Wauters, E., Walmsley, S., Prenen, H., Granot, Z.,
et al. (2015). MET is required for the recruitment of anti-tumoural neutro-
phils. Nature 522, 349–353. https://doi.org/10.1038/nature14407.
32. Glodde, N., Bald, T., van den Boorn-Konijnenberg, D., Nakamura, K.,
O’Donnell, J.S., Szczepanski, S., Brandes, M., Eickhoff, S., Das, I., Shrid-
har, N., et al. (2017). Reactive neutrophil responses dependent on the re-
ceptor tyrosine kinase c-MET limit cancer immunotherapy. Immunity 47,
789–802.e9. https://doi.org/10.1016/j.immuni.2017.09.012.
33. Bodac, A., and Meylan, E. (2021). Neutrophil metabolism in the cancer
context. Semin. Immunol. 57, 101583. https://doi.org/10.1016/j.smim.
2021.101583.
34. Garcı´a-Navas, R., Gajate, C., and Mollinedo, F. (2021). Neutrophils drive
endoplasmic reticulum stress-mediated apoptosis in cancer cells
through arginase-1 release. Sci. Rep. 11, 12574. https://doi.org/10.
1038/s41598-021-91947-0.
35. Wang, J.F., Wang, Y.P., Xie, J., Zhao, Z.Z., Gupta, S., Guo, Y., Jia, S.H.,
Parodo, J., Marshall, J.C., and Deng, X.M. (2021). Upregulated PD-L1
delays human neutrophil apoptosis and promotes lung injury in an exper-
imental mouse model of sepsis. Blood 138, 806–810. https://doi.org/10.
1182/blood.2020009417.
36. Zilionis, R., Engblom, C., Pfirschke, C., Savova, V., Zemmour, D., Saat-
cioglu, H.D., Krishnan, I., Maroni, G., Meyerovitz, C.V., Kerwin, C.M.,
et al. (2019). Single-cell transcriptomics of human and mouse lung can-
cers reveals conserved myeloid populations across individuals and spe-
cies. Immunity 50, 1317–1334.e10. https://doi.org/10.1016/j.immuni.
2019.03.009.
37. Salcher, S., Sturm, G., Horvath, L., Untergasser, G., Kuempers, C., Fota-
kis, G., Panizzolo, E., Martowicz, A., Trebo, M., Pall, G., et al. (2022).
High-resolution single-cell atlas reveals diversity and plasticity of tis-
sue-resident neutrophils in non-small cell lung cancer. Cancer Cell 40,
1503–1520.e8. https://doi.org/10.1016/j.ccell.2022.10.008.
38. Alvarez-Breckenridge, C., Markson, S.C., Stocking, J.H., Nayyar, N.,
Lastrapes, M., Strickland, M.R., Kim, A.E., de Sauvage, M., Dahal, A.,
Larson, J.M., et al. (2022). Microenvironmental landscape of human mel-
anoma brain metastases in response to immune checkpoint inhibition.Cell 186, 4546–4566, October 12, 2023 4563
ll
OPEN ACCESS ArticleCancer Immunol. Res. 10, 996–1012. https://doi.org/10.1158/2326-
6066.CIR-21-0870.
39. A´lvarez-Prado, A´.F., Maas, R.R., Soukup, K., Klemm, F., Kornete, M.,
Krebs, F.S., Zoete, V., Berezowska, S., Brouland, J.P., Hottinger, A.F.,
et al. (2023). Immunogenomic analysis of human brain metastases re-
veals diverse immune landscapes across genetically distinct tumors.
Cell Rep. Med. 4, 100900. https://doi.org/10.1016/j.xcrm.2022.100900.
40. Gershkovitz, M., Caspi, Y., Fainsod-Levi, T., Katz, B., Michaeli, J., Kha-
waled, S., Lev, S., Polyansky, L., Shaul, M.E., Sionov, R.V., et al.
(2018). TRPM2 mediates neutrophil killing of disseminated tumor cells.
Cancer Res. 78, 2680–2690. https://doi.org/10.1158/0008-5472.CAN-
17-3614.
41. Granot, Z., Henke, E., Comen, E.A., King, T.A., Norton, L., and Benezra,
R. (2011). Tumor entrained neutrophils inhibit seeding in the premeta-
static lung. Cancer Cell 20, 300–314. https://doi.org/10.1016/j.ccr.
2011.08.012.
42. Takeshima, T., Pop, L.M., Laine, A., Iyengar, P., Vitetta, E.S., and
Hannan, R. (2016). Key role for neutrophils in radiation-induced antitumor
immune responses: potentiation with G-CSF. Proc. Natl. Acad. Sci. USA
113, 11300–11305. https://doi.org/10.1073/pnas.1613187113.
43. Rice, C.M., Davies, L.C., Subleski, J.J., Maio, N., Gonzalez-Cotto, M.,
Andrews, C., Patel, N.L., Palmieri, E.M., Weiss, J.M., Lee, J.M., et al.
(2018). Tumour-elicited neutrophils engage mitochondrial metabolism
to circumvent nutrient limitations and maintain immune suppression.
Nat. Commun. 9, 5099. https://doi.org/10.1038/s41467-018-07505-2.
44. Schmielau, J., and Finn, O.J. (2001). Activated granulocytes and granu-
locyte-derived hydrogen peroxide are the underlying mechanism of sup-
pression of T-cell function in advanced cancer patients. Cancer Res. 61,
4756–4760.
45. Singhal, S., Bhojnagarwala, P.S., O’Brien, S., Moon, E.K., Garfall, A.L.,
Rao, A.S., Quatromoni, J.G., Stephen, T.L., Litzky, L., Deshpande, C.,
et al. (2016). Origin and role of a subset of tumor-associated neutrophils
with antigen-presenting cell features in early-stage human lung cancer.
Cancer Cell 30, 120–135. https://doi.org/10.1016/j.ccell.2016.06.001.
46. De Palma, M., Biziato, D., and Petrova, T.V. (2017). Microenvironmental
regulation of tumour angiogenesis. Nat. Rev. Cancer 17, 457–474.
https://doi.org/10.1038/nrc.2017.51.
47. Hao, Q., Chen, Y., Zhu, Y., Fan, Y., Palmer, D., Su, H., Young, W.L., and
Yang, G.Y. (2007). Neutrophil depletion decreases VEGF-induced focal
angiogenesis in the mature mouse brain. J. Cereb. Blood Flow Metab.
27, 1853–1860. https://doi.org/10.1038/sj.jcbfm.9600485.
48. Ajikumar, A., Long, M.B., Heath, P.R., Wharton, S.B., Ince, P.G., Ridger,
V.C., and Simpson, J.E. (2019). Neutrophil-derived microvesicle induced
dysfunction of brain microvascular endothelial cells in vitro. Int. J. Mol.
Sci. 20. https://doi.org/10.3390/ijms20205227.
49. Qiu, Y.M., Zhang, C.L., Chen, A.Q., Wang, H.L., Zhou, Y.F., Li, Y.N., and
Hu, B. (2021). Immune cells in the BBB disruption after acute ischemic
stroke: targets for immune therapy? Front. Immunol. 12, 678744.
https://doi.org/10.3389/fimmu.2021.678744.
50. Winkler, F., Kozin, S.V., Tong, R.T., Chae, S.S., Booth, M.F., Garkavtsev,
I., Xu, L., Hicklin, D.J., Fukumura, D., di Tomaso, E., et al. (2004). Kinetics
of vascular normalization by VEGFR2 blockade governs brain tumor
response to radiation: role of oxygenation, angiopoietin-1, and matrix
metalloproteinases. Cancer Cell 6, 553–563. https://doi.org/10.1016/j.
ccr.2004.10.011.
51. Izumi, Y., Xu, L., di Tomaso, E., Fukumura, D., and Jain, R.K. (2002).
Tumour biology: herceptin acts as an anti-angiogenic cocktail. Nature
416, 279–280. https://doi.org/10.1038/416279b.
52. Arvanitis, C.D., Ferraro, G.B., and Jain, R.K. (2020). The blood-brain bar-
rier and blood-tumour barrier in brain tumours andmetastases. Nat. Rev.
Cancer 20, 26–41. https://doi.org/10.1038/s41568-019-0205-x.4564 Cell 186, 4546–4566, October 12, 202353. Wang, S., Song, R., Wang, Z., Jing, Z., Wang, S., and Ma, J. (2018).
S100A8/A9 in inflammation. Front. Immunol. 9, 1298. https://doi.org/
10.3389/fimmu.2018.01298.
54. Quintero-Fabia´n, S., Arreola, R., Becerril-Villanueva, E., Torres-Romero,
J.C., Arana-Arga´ez, V., Lara-Riegos, J., Ramı´rez-Camacho, M.A., and
Alvarez-Sa´nchez, M.E. (2019). Role of matrix metalloproteinases in
angiogenesis and cancer. Front. Oncol. 9, 1370. https://doi.org/10.
3389/fonc.2019.01370.
55. Summers, C., Rankin, S.M., Condliffe, A.M., Singh, N., Peters, A.M., and
Chilvers, E.R. (2010). Neutrophil kinetics in health and disease. Trends
Immunol. 31, 318–324. https://doi.org/10.1016/j.it.2010.05.006.
56. Adrover, J.M., Nicola´s-A´vila, J.A., and Hidalgo, A. (2016). Aging: a tem-
poral dimension for neutrophils. Trends Immunol. 37, 334–345. https://
doi.org/10.1016/j.it.2016.03.005.
57. Brostjan, C., and Oehler, R. (2020). The role of neutrophil death in chronic
inflammation and cancer. Cell Death Discov. 6, 26. https://doi.org/10.
1038/s41420-020-0255-6.
58. Comen, E., Wojnarowicz, P., Seshan, V.E., Shah, R., Coker, C., Norton,
L., and Benezra, R. (2016). TNF is a key cytokine mediating neutrophil
cytotoxic activity in breast cancer patients. npj Breast Cancer 2,
16009. https://doi.org/10.1038/npjbcancer.2016.9.
59. Vieira, S.M., Lemos, H.P., Grespan, R., Napimoga, M.H., Dal-Secco, D.,
Freitas, A., Cunha, T.M., Verri, W.A., Jr., Souza-Junior, D.A., Jamur,
M.C., et al. (2009). A crucial role for TNF-alpha in mediating neutrophil
influx induced by endogenously generated or exogenous chemokines,
KC/CXCL1 and LIX/CXCL5. Br. J. Pharmacol. 158, 779–789. https://
doi.org/10.1111/j.1476-5381.2009.00367.x.
60. Chapman, A.L., Mocatta, T.J., Shiva, S., Seidel, A., Chen, B., Khalilova, I.,
Paumann-Page, M.E., Jameson, G.N., Winterbourn, C.C., and Kettle,
A.J. (2013). Ceruloplasmin is an endogenous inhibitor of myeloperoxi-
dase. J. Biol. Chem. 288, 6465–6477. https://doi.org/10.1074/jbc.
M112.418970.
61. Golenkina, E.A., Viryasova, G.M., Galkina, S.I., Gaponova, T.V., Sud’ina,
G.F., and Sokolov, A.V. (2018). Fine regulation of neutrophil oxidative sta-
tus and apoptosis by ceruloplasmin and its derivatives. Cells 7. https://
doi.org/10.3390/cells7010008.
62. Xue, M., Rabbani, N., Momiji, H., Imbasi, P., Anwar, M.M., Kitteringham,
N., Park, B.K., Souma, T., Moriguchi, T., Yamamoto, M., et al. (2012).
Transcriptional control of glyoxalase 1 by Nrf2 provides a stress-respon-
sive defence against dicarbonyl glycation. Biochem. J. 443, 213–222.
https://doi.org/10.1042/BJ20111648.
63. Chowdhury, K., Kumar, U., Das, S., Chaudhuri, J., Kumar, P., Kanjilal, M.,
Ghosh, P., Sircar, G., Basyal, R.K., Kanga, U., et al. (2018). Synovial IL-9
facilitates neutrophil survival, function and differentiation of Th17 cells in
rheumatoid arthritis. Arthritis Res. Ther. 20, 18. https://doi.org/10.1186/
s13075-017-1505-8.
64. Xiao, Y., Cong, M., Li, J., He, D., Wu, Q., Tian, P., Wang, Y., Yang, S.,
Liang, C., Liang, Y., et al. (2021). Cathepsin C promotes breast cancer
lungmetastasis bymodulating neutrophil infiltration and neutrophil extra-
cellular trap formation. Cancer Cell 39, 423–437.e7. https://doi.org/10.
1016/j.ccell.2020.12.012.
65. Zha, C., Meng, X., Li, L., Mi, S., Qian, D., Li, Z., Wu, P., Hu, S., Zhao, S.,
Cai, J., et al. (2020). Neutrophil extracellular traps mediate the crosstalk
between glioma progression and the tumor microenvironment via the
HMGB1/RAGE/IL-8 axis. Cancer Biol. Med. 17, 154–168. https://doi.
org/10.20892/j.issn.2095-3941.2019.0353.
66. Eruslanov, E.B., Bhojnagarwala, P.S., Quatromoni, J.G., Stephen, T.L.,
Ranganathan, A., Deshpande, C., Akimova, T., Vachani, A., Litzky, L.,
Hancock, W.W., et al. (2014). Tumor-associated neutrophils stimulate
T cell responses in early-stage human lung cancer. J. Clin. Invest. 124,
5466–5480. https://doi.org/10.1172/JCI77053.
67. Fortunati, E., Kazemier, K.M., Grutters, J.C., Koenderman, L., and Van
den Bosch v, J. (2009). Human neutrophils switch to an activated pheno-
type after homing to the lung irrespective of inflammatory disease. Clin.
ll
OPEN ACCESSArticleExp. Immunol. 155, 559–566. https://doi.org/10.1111/j.1365-2249.2008.
03791.x.
68. Gonzalez, H., Mei, W., Robles, I., Hagerling, C., Allen, B.M., Hauge
Okholm, T.L., Nanjaraj, A., Verbeek, T., Kalavacherla, S., van Gogh, M.,
et al. (2022). Cellular architecture of human brain metastases. Cell 185,
729–745.e20. https://doi.org/10.1016/j.cell.2021.12.043.
69. Sudmeier, L.J., Hoang, K.B., Nduom, E.K., Wieland, A., Neill, S.G.,
Schniederjan, M.J., Ramalingam, S.S., Olson, J.J., Ahmed, R., and Hud-
son, W.H. (2022). Distinct phenotypic states and spatial distribution of
CD8(+) T cell clonotypes in human brain metastases. Cell. Rep. Med.
3, 100620. https://doi.org/10.1016/j.xcrm.2022.100620.
70. Wischnewski, V., Maas, R.R., Aruffo, P.G., Soukup, K., Galletti, G., Kor-
nete, M., Galland, S., Fournier, N., Lilja, J., Wirapati, P., et al. (2023).
Phenotypic diversity of T cells in human primary and metastatic brain tu-
mors revealed by multiomic interrogation. Nat. Cancer 4, 908–924.
https://doi.org/10.1038/s43018-023-00566-3.
71. Wang, T.T., Zhao, Y.L., Peng, L.S., Chen, N., Chen, W., Lv, Y.P., Mao,
F.Y., Zhang, J.Y., Cheng, P., Teng, Y.S., et al. (2017). Tumour-activated
neutrophils in gastric cancer foster immune suppression and disease
progression through GM-CSF-PD-L1 pathway. Gut 66, 1900–1911.
https://doi.org/10.1136/gutjnl-2016-313075.
72. Wu, L., Saxena, S., Goel, P., Prajapati, D.R., Wang, C., and Singh, R.K.
(2020). Breast cancer cell-neutrophil interactions enhance neutrophil sur-
vival and pro-tumorigenic activities. Cancers (Basel) 12. https://doi.org/
10.3390/cancers12102884.
73. Mensurado, S., Rei, M., Lanc¸a, T., Ioannou, M., Gonc¸alves-Sousa, N.,
Kubo, H., Malissen, M., Papayannopoulos, V., Serre, K., and Silva-San-
tos, B. (2018). Tumor-associated neutrophils suppress pro-tumoral IL-
17+ gd T cells through induction of oxidative stress. PLoS Biol. 16,
e2004990. https://doi.org/10.1371/journal.pbio.2004990.
74. Chang, Y., Cai, X., Syahirah, R., Yao, Y., Xu, Y., Jin, G., Bhute, V.J., Tor-
regrosa-Allen, S., Elzey, B.D., Won, Y.Y., et al. (2023). CAR-neutrophil
mediated delivery of tumor-microenvironment responsive nanodrugs
for glioblastoma chemo-immunotherapy. Nat. Commun. 14, 2266.
https://doi.org/10.1038/s41467-023-37872-4.
75. Xue, R., Zhang, Q., Cao, Q., Kong, R., Xiang, X., Liu, H., Feng, M., Wang,
F., Cheng, J., Li, Z., et al. (2022). Liver tumour immune microenvironment
subtypes and neutrophil heterogeneity. Nature 612, 141–147. https://doi.
org/10.1038/s41586-022-05400-x.
76. Shan, Z.G., Chen, J., Liu, J.S., Zhang, J.Y., Wang, T.T., Teng, Y.S., Mao,
F.Y., Cheng, P., Zou, Q.M., Zhou, W.Y., et al. (2021). Activated neutro-
phils polarize protumorigenic interleukin-17A-producing T helper sub-
sets through TNF-alpha-B7-H2-dependent pathway in human gastric
cancer. Clin. Transl. Med. 11, e484. https://doi.org/10.1002/ctm2.484.
77. Szklarczyk, D., Morris, J.H., Cook, H., Kuhn, M., Wyder, S., Simonovic,
M., Santos, A., Doncheva, N.T., Roth, A., Bork, P., et al. (2017). The
STRING database in 2017: quality-controlled protein-protein association
networks, made broadly accessible. Nucleic Acids Res. 45, D362–D368.
https://doi.org/10.1093/nar/gkw937.
78. Croci, D., Santalla Me´ndez, R., Temme, S., Soukup, K., Fournier, N., Zo-
mer, A., Colotti, R., Wischnewski, V., Flo¨gel, U., van Heeswijk, R.B., et al.
(2022). Multispectral fluorine-19 MRI enables longitudinal and noninva-
sive monitoring of tumor-associated macrophages. Sci. Transl. Med.
14, eabo2952. https://doi.org/10.1126/scitranslmed.abo2952.
79. Holland, E.C., Hively, W.P., DePinho, R.A., and Varmus, H.E. (1998). A
constitutively active epidermal growth factor receptor cooperates with
disruption of G1 cell-cycle arrest pathways to induce glioma-like lesions
in mice. Genes Dev. 12, 3675–3685. https://doi.org/10.1101/gad.12.
23.3675.
80. Bankhead, P., Loughrey, M.B., Ferna´ndez, J.A., Dombrowski, Y., McArt,
D.G., Dunne, P.D., McQuaid, S., Gray, R.T., Murray, L.J., Coleman, H.G.,
et al. (2017). QuPath: open source software for digital pathology image
analysis. Sci. Rep. 7, 16878. https://doi.org/10.1038/s41598-017-
17204-5.81. Schmidt, U., Weigert, M., Broaddus, C., andMyers, G. (2018). Cell detec-
tion with star-convex polygons. In Medical Image Computing and Com-
puter Assisted Intervention – MICCAI 2018 : 21st International Confer-
ence (Springer International Publishing), pp. 265–273.
82. Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S.,
Batut, P., Chaisson, M., and Gingeras, T.R. (2013). STAR: ultrafast uni-
versal RNA-seq aligner. Bioinformatics 29, 15–21. https://doi.org/10.
1093/bioinformatics/bts635.
83. Li, B., and Dewey, C.N. (2011). RSEM: accurate transcript quantification
from RNA-seq data with or without a reference genome. BMC Bioinfor-
matics 12, 323. https://doi.org/10.1186/1471-2105-12-323.
84. Baddeley, A., Rubak, E., and Turner, R. (2015). Spatial point patterns:
methodology and applications with (R (CRC press).
85. Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2010). edgeR: a Bio-
conductor package for differential expression analysis of digital gene
expression data. Bioinformatics 26, 139–140. https://doi.org/10.1093/
bioinformatics/btp616.
86. Ritchie, M.E., Phipson, B., Wu, D., Hu, Y.F., Law, C.W., Shi, W., and
Smyth, G.K. (2015). limma powers differential expression analyses for
RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47.
https://doi.org/10.1093/nar/gkv007.
87. Leek, J.T., and Storey, J.D. (2007). Capturing heterogeneity in gene
expression studies by surrogate variable analysis. PLoS Genet. 3,
1724–1735. https://doi.org/10.1371/journal.pgen.0030161.
88. Wu, T.Z., Hu, E.Q., Xu, S.B., Chen, M.J., Guo, P.F., Dai, Z.H., Feng, T.Z.,
Zhou, L., Tang, W.L., Zhan, L., et al. (2021). clusterProfiler 4.0: a universal
enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141.
https://doi.org/10.1016/j.xinn.2021.100141.
89. Blighe, K. (2021). scDataviz: single cell dataviz and downstream analy-
sesR package version 1.2.0 (Bioconductor. https://doi.org/10.18129/
B9.bioc.scDataviz.
90. Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis
(Springer-Verlag).
91. Trapnell, C., Cacchiarelli, D., Grimsby, J., Pokharel, P., Li, S., Morse, M.,
Lennon, N.J., Livak, K.J., Mikkelsen, T.S., and Rinn, J.L. (2014). The dy-
namics and regulators of cell fate decisions are revealed by pseudotem-
poral ordering of single cells. Nat. Biotechnol. 32, 381–386. https://doi.
org/10.1038/nbt.2859.
92. Street, K., Risso, D., Fletcher, R.B., Das, D., Ngai, J., Yosef, N., Purdom,
E., and Dudoit, S. (2018). Slingshot: cell lineage and pseudotime infer-
ence for single-cell transcriptomics. BMC Genomics 19, 477. https://
doi.org/10.1186/s12864-018-4772-0.
93. Martin, M. (2011). Cutadapt removes adapter sequences from high-
throughput sequencing reads. EMBnet.journal 17, 10–12. https://doi.
org/10.14806/ej.17.1.200.
94. Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with
Burrows-Wheeler transform. Bioinformatics 25, 1754–1760. https://doi.
org/10.1093/bioinformatics/btp324.
95. Danecek, P., Bonfield, J.K., Liddle, J., Marshall, J., Ohan, V., Pollard,
M.O., Whitwham, A., Keane, T., McCarthy, S.A., Davies, R.M., et al.
(2021). Twelve years of SAMtools and BCFtools. GigaScience 10.
https://doi.org/10.1093/gigascience/giab008.
96. McLaren, W., Gil, L., Hunt, S.E., Riat, H.S., Ritchie, G.R., Thormann, A.,
Flicek, P., and Cunningham, F. (2016). The Ensembl variant effect predic-
tor. Genome Biol. 17, 122. https://doi.org/10.1186/s13059-016-0974-4.
97. Gu, Z.G., Eils, R., and Schlesner, M. (2016). Complex heatmaps reveal
patterns and correlations inmultidimensional genomic data. Bioinformat-
ics 32, 2847–2849. https://doi.org/10.1093/bioinformatics/btw313.
98. Csardi, G., and Nepusz, T. (2006). The igraph software package for com-
plex network research. InterJournal Complex Systems 1695, 1–9.
99. Bates, D., Ma¨chler, M., Bolker, B.M., and Walker, S.C. (2015). Fitting
linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48.
https://doi.org/10.18637/jss.v067.i01.Cell 186, 4546–4566, October 12, 2023 4565
ll
OPEN ACCESS Article100. Kuznetsova, A., Brockhoff, P.B., and Christensen, R.H.B. (2017). lmerT-
est package: tests in linear mixed effects models. J. Stat. Softw. 82,
1–26. https://doi.org/10.18637/jss.v082.i13.
101. Ozawa, T., Riester, M., Cheng, Y.K., Huse, J.T., Squatrito, M., Helmy, K.,
Charles, N., Michor, F., and Holland, E.C. (2014). Most human non-GCIMP
glioblastoma subtypes evolve froma common proneural-like precursor gli-
oma. Cancer Cell 26, 288–300. https://doi.org/10.1016/j.ccr.2014.06.005.
102. Becher, O.J., Hambardzumyan, D., Fomchenko, E.I., Momota, H., Main-
waring, L., Bleau, A.M., Katz, A.M., Edgar, M., Kenney, A.M., Cordon-
Cardo, C., et al. (2008). Gli activity correlates with tumor grade in
platelet-derived growth factor-induced gliomas. Cancer Res. 68, 2241–
2249. https://doi.org/10.1158/0008-5472.CAN-07-6350.
103. Gut, G., Herrmann, M.D., and Pelkmans, L. (2018). Multiplexed protein
maps link subcellular organization to cellular states. Science 361.
https://doi.org/10.1126/science.aar7042.4566 Cell 186, 4546–4566, October 12, 2023104. Sternberg, S.R. (1983). Biomedical image processing. Computer
16, 22–34.
105. Baddeley, A., and Turner, R. (2005). spatstat: an R package for analyzing
spatial point patterns. J. Stat. Softw. 12, 1–42. https://doi.org/10.18637/
jss.v012.i06.
106. Pyonteck, S.M., Akkari, L., Schuhmacher, A.J., Bowman, R.L., Sevenich,
L., Quail, D.F., Olson, O.C., Quick, M.L., Huse, J.T., Teijeiro, V., et al.
(2013). CSF-1R inhibition alters macrophage polarization and blocks gli-
oma progression. Nat. Med. 19, 1264–1272. https://doi.org/10.1038/
nm.3337.
107. Hambardzumyan, D., Amankulor, N.M., Helmy, K.Y., Becher, O.J.,
and Holland, E.C. (2009). Modeling adult gliomas using RCAS/t-
va technology. Transl. Oncol. 2, 89–95. https://doi.org/10.1593/
tlo.09100.
ll
OPEN ACCESSArticleSTAR+METHODSKEY RESOURCES TABLEREAGENT or RESOURCE SOURCE IDENTIFIER
Antibodies
FCM: Annexin V PE Thermo Fisher Scientific Cat# 88-8102-72, RRID:AB_2575183
FCM: Arginase 1 PE-Cyanine7, rat
monoclonal anti-human (clone
A1exF5), dilution 1:80
Thermo Fisher Scientific Cat# 25-3697-80, RRID:AB_2734840
FCM: CD11b APC, rat monoclonal
anti-human/mouse (clone M1/70),
dilution 1:300
Thermo Fisher Scientific Cat# 17-0112-82, RRID:AB_469343
FCM: CD11B BUV661, rat monoclonal
anti-human/mouse (clone M1/70),
dilution 1:1280
BD Bioscience Cat# 612977, RRID:AB_2870249
FCM: CD11B BUV737, rat monoclonal
anti-human/mouse (clone M1/70),
dilution 1:1280
BD Bioscience Cat# 612800, RRID:AB_2870127
FCM: CD11B BV421, rat monoclonal
anti-human/mouse (clone M1/70),
dilution 1:1280
BioLegend Cat# 101251, RRID:AB_2562904
FCM: CD11C BV605, mouse monoclonal
anti-human (clone 3.9), dilution 1:320
BioLegend Cat # 301636, RRID:AB_2563796
FCM: CD127 BV510, mouse monoclonal
anti-human (clone A019D5), dilution 1:160
BioLegend Cat # 351332, RRID:AB_2562304
FCM: CD14 AF488, mouse monoclonal
anti-human (clone HCD14), dilution 1:640
BioLegend Cat # 325610, RRID:AB_830683
FCM: CD15 BV510, mouse monoclonal
anti-human (clone W6D3), dilution 1:640
BioLegend Cat # 323028, RRID:AB_2563400
FCM: CD16 BUV737 mouse monoclonal
anti-human (clone 3G8), dilution 1:640
BD Bioscience Cat # 612786, RRID:AB_2833077
FCM: CD16 BV421 mouse monoclonal
anti-human (clone 3G8), dilution 1:1280
BioLegend Cat # 302037, RRID:AB_10898112
FCM: CD16 BV510 mouse monoclonal
anti-human (clone 3G8), dilution 1:1280
BioLegend Cat # 302048, RRID:AB_2562085
FCM: CD16 BV785 mouse monoclonal
anti-human (clone 3G8), dilution 1:640
BioLegend Cat# 360734, RRID:AB_2800994
FCM: CD19 BUV563, mouse monoclonal
anti-human (clone SJ25C1), dilution 1:320
BD Bioscience Cat # 612916,
RRID:AB_2870201
FCM: CD25 PE, mouse monoclonal anti-human
(clone BC96), dilution 1:80
BioLegend Cat # 302606, RRID:AB_314276
FCM: CD274 (PD-L1) BV711 mouse monoclonal
anti-human (clone 29E.2A3), dilution 1:60
BioLegend Cat# 329722, RRID:AB_2565764
FCM: CD3 PerCP/Cy5.5 mouse monoclonal
anti-human (clone HIT3a), dilution 1:80
BioLegend Cat# 300328, RRID:AB_1575008
FCM: CD4 BV650 mouse monoclonal
anti-human (clone OKT4), dilution 1:200
BioLegend Cat# 317436, RRID:AB_2563050
FCM: CD45 AF700 mouse monoclonal
anti-human (clone HI30), dilution 1:640
BioLegend Cat# 304023, RRID:AB_493760
FCM: CD45 AF700 rat monoclonal
anti-mouse (clone 30-F11), dilution 1:500
BioLegend Cat# 103128, RRID:AB_493715
FCM: CD45 BV605 mouse monoclonal
anti-human (clone HI30), dilution 1:640
BioLegend Cat# 304042, RRID:AB_2562106
(Continued on next page)
Cell 186, 4546–4566.e1–e15, October 12, 2023 e1
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
FCM: CD49D APC mouse monoclonal
anti-human (clone 9F10), dilution 1:320
BioLegend Cat# 304308, RRID:AB_2130041
FCM: CD56 PE/Dazzle 594 mouse monoclonal
anti-human (clone HDC56), dilution 1:640
BioLegend Cat# 318348, RRID:AB_2563564
FCM: CD62L BV421 rat monoclonal anti-mouse
(clone MEL-14), dilution1:400
BioLegend Cat# 104435, RRID:AB_10900082
FCM: CD62L PerCP/Cyanine5.5 mouse monoclonal
anti-human (clone DREG-56), dilution 1:500
BioLegend Cat# 304824, RRID:AB_2239105
FCM: CD66B AF700 mouse monoclonal
anti-human (clone G10F5), dilution 1:100
BioLegend Cat# 305114, RRID:AB_2566038
FCM: CD66B PE mouse monoclonal
anti-human (clone 4/60c), dilution 1:200
BioLegend Cat# 392904, RRID:AB_2750202
FCM: CD66B PE/Cyanine7 mouse
monoclonal anti-human (clone G10F5),
dilution 1:640
BioLegend Cat# 305116, RRID:AB_2566605
FCM: CD8a PE/Cyanine7 mouse
monoclonal anti-human (clone HIT8a),
dilution 1:320
BioLegend Cat# 300914, RRID:AB_314118
FCM: CD80 PE/Cyanine7 mouse monoclonal
anti-human (clone L307.4), dilution 1:160
BD Bioscience Cat# 561135, RRID:AB_10561688
FCM: CD86 Pacific Blue mouse monoclonal
anti-human (clone IT2.2), dilution 1:160
BioLegend Cat# 305418, RRID:AB_493663
FCM: cMET AF647 mouse monoclonal
anti-human (clone 3D6), dilution 1:40
BD Bioscience Cat# 566014, RRID:AB_2739459
FCM: CD181 (CXCR1) BV711 mouse
monoclonal anti-human (clone 5A12),
dilution 1:80
BD Bioscience Cat# 743423, RRID:AB_2741496
FCM: CD182 (CXCR2) AF488 mouse
monoclonal anti-human (clone 5E8/CXCR2),
dilution 1:80
BioLegend Cat# 320712, RRID:AB_492938
FCM: CXCR4 PE mouse monoclonal
anti-human (clone 12G5), dilution 1:25
R&D Systems Cat# FAB170P, RRID:AB_357076
FCM: CD184 (CXCR4) PerCP-eFluor710
rat monoclonal anti-human/mouse (clone
2B11), dilution 1:60
Thermo Fisher Scientific Cat# 46-9991-82, RRID:AB_10670489
FCM: HLA-ABC APC mouse monoclonal
anti-human (clone W6/32), dilution 1:400
BioLegend Cat# 311409, RRID:AB_314878
FCM: HLA-DR BV711 mouse monoclonal
anti-human (clone L243), dilution 1:320
BioLegend Cat# 307644, RRID:AB_2562913
FCM: Ly6C BV711 rat monoclonal anti-
mouse (clone HK1.4), dilution 1:800
BioLegend Cat# 128037, RRID:AB_2562630
FCM: Ly6G BV605 rat monoclonal anti-
mouse (clone 1A8), dilution 1:160
BioLegend Cat# 127639, RRID:AB_2565880
FCM: Ly6G FITC rat monoclonal anti-
mouse (clone 1A8), dilution 1:300
BioLegend Cat# 127605, RRID:AB_1236488
FCM: MMP9 AF488 mouse monoclonal
anti-human (clone 56129), dilution 1:80
R&D Systems Cat# IC9111F, RRID:AB_2144867
FCM/IF: MRP-14 (S100A9) PE mouse
monoclonal anti-human (clone MRP1H9),
dilution 1:640 (FCM), 1:100 (IF)
BioLegend Cat# 350706, RRID:AB_2564008
IF: anti-chicken IgG (H+L) AF647, dilution 1:500 Jackson ImmunoResearch Cat# 703-605-155, RRID:AB_2340379
IF: anti-goat DyLight 755, dilution 1:500 Thermo Fisher Scientific Cat# SA5-10091, RRID:AB_2556671
IF: anti-goat IgG (H+L) AF488, dilution 1:500 Thermo Fisher Scientific Cat# A32814, RRID:AB_2762838
(Continued on next page)
ll
OPEN ACCESS
e2 Cell 186, 4546–4566.e1–e15, October 12, 2023
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
IF: anti-mouse IgG (H+L) AF555, dilution 1:500 Thermo Fisher Scientific Cat# A-32773, RRID:AB_2762848
IF: anti-mouse IgG (H+L) AF647, dilution 1:500 Thermo Fisher Scientific Cat# A-31571, RRID:AB_162542
IF: anti-rabbit IgG (H+L) AF555, dilution 1:500 Thermo Fisher Scientific Cat# A-32794, RRID:AB_2762834
IF: anti-rabbit IgG (H+L) AF755, dilution 1:500 Thermo Fisher Scientific Cat# SA5-10043, RRID:AB_2556623
IF: anti-rat IgG (H+L) AF647, dilution 1:500 Abcam Cat# ab150155, RRID:AB_2813835
IF: anti-sheep IgG (H+L) AF488, dilution 1:500 Thermo Fisher Scientific Cat# A11015, RRID:AB_2534082
IF: CD15 mouse monoclonal anti-human
(clone MMA), dilution 1:100
Abcam Cat# ab17080, RRID:AB_443635
IF: CD3 mouse monoclonal anti-human
(clone UCHT1), dilution 1:50
BioLegend Cat# 300401, RRID:AB_314055
IF: CD3 FITC mouse monoclonal anti-
human (clone UCHT1), dilution 1:100
BioLegend Cat# 300406,
RRID:AB_314060
IF: CD31 sheep polyclonal anti-human,
dilution 1:100
R&D Systems Cat# AF806, RRID:AB_355617
IF: CD4 rabbit monoclonal anti-human
(clone EPR6855), dilution 1:100
Abcam Cat# ab133616, RRID:AB_2750883
IF: CD45 goat polyclonal anti-human,
dilution 1:100
LSBio Cat# LS-B14248-300, RRID:AB_2889893
IF: CD45 AF647 mouse monoclonal
anti-human (clone HI30), dilution 1:100
BioLegend Cat# 304018, RRID:AB_389336
IF: CD49D rat monoclonal anti-human
(clone PS/2), dilution 1:100
Bio X Cell Cat# BE0071, RRID:AB_1107657
IF: CD68 mouse monoclonal anti-human
(clone KP1), dilution 1:100
Abcam Cat# ab955, RRID:AB_307338
IF: CD8 mouse monoclonal anti-human
(clone 4B11), dilution 1:100
Bio-rad Cat# MCA1817T, RRID:AB_323534
IF: FOXP3 rabbit monoclonal anti-human
(clone SP97), dilution 1:100
Thermo Fisher Scientific Cat# MA5-16365, RRID:AB_2537884
IF: HIF1a goat polyclonal anti-human,
dilution 1:100
R&D Systems Cat# AF1935, RRID:AB_355064
IF: HIF2a rabbit polyclonal anti-human,
dilution 1:100
Abcam Cat# ab109616, RRID:AB_11156727
IF: Ki67 AF647 rat monoclonal anti-human
(clone SolA15), dilution 1:100
Thermo Fisher Scientific Cat# 17-5698-82, RRID:AB_2688057
IF: P2RY12 rabbit polyclonal anti-human,
dilution 1:600
Sigma-Aldrich Cat# HPA014518, RRID:AB_2669027
IF: Pan-cadherin rabbit polyclonal
anti-human, dilution 1:100
Abcam Cat# ab16505, RRID:AB_443397
IF: PD-L1 rabbit monoclonal anti-human
(clone SP142), dilution 1:50
Abcam Cat# ab228462, RRID:AB_2827816
IF: CD279 (PD-1) AF647 mouse monoclonal
anti-human (clone NAT105), dilution 1:100
BioLegend Cat# 367419, RRID:AB_2721353
In vitro: Human TNF-a Neutralizing
(clone D1B4) Rabbit mAb
Cell Signaling Technology Cat# 7321S, RRID:AB_10925386
In vitro: Human/Primate IL-6 Antibody (clone 6708) R&D Systems Cat# MAB206, RRID:AB_2127617
In vitro: Human IL-9 Antibody, polyclonal R&D Systems Cat# AF209, RRID:AB_2296123
Biological samples
Non-tumor, glioma, and brain metastasis tissue,
along with matched patient blood
Centre Hospitalier
Universitaire Vaudois,
Lausanne, Switzerland
N/A
(Continued on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566.e1–e15, October 12, 2023 e3
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Non-tumor, glioma and brain metastasis tissue Memorial Sloan Kettering
Cancer Center, New York,
NY, USA
N/A
Buffy coats from healthy donors Transfusion Interre´gionale
Croix-Rouge Suisse,
Epalinges, Switzerland
N/A
Healthy donor blood Universite´ de Lausanne,
Lausanne, Switzerland
N/A
Healthy donor blood New York Blood Bank,
New York, NY, USA
N/A
Chemicals, peptides, and recombinant proteins
DMEM-F12 (1:1), GlutaMAX Gibco, Thermo Fisher Cat# 31331028
HBSS Gibco, Thermo Fisher Cat# 14175-095
DMEM-F12 w/o phenol red Gibco, Thermo Fisher Cat# 21041-025
Penicillin/Streptomycin Gibco, Thermo Fisher Cat# 15140122
Fetal Bovine Serum (FBS) Gibco, Thermo Fisher Cat# 31331028
Trypsin-EDTA (0.05%), phenol red Gibco, Thermo Fisher Cat# 25300062
PBS Gibco, Thermo Fisher Cat# 20012027
RBC Lysis Buffer (10X) Biolegend Cat# 00-4333-57
Ultrapure distilled water Invitrogen Cat# 10977-035
Ficoll-Paque Premium GE Cat# 17-5442-02
Tissue-Tek O.C.T. Compound Tissue-Tek, Sakura Finetek Cat# 4583
CS&T beads BD Bioscience Cat# 656504
2-methylbutane Sigma-Aldrich Cat# 50970
Methanol, 99.9% Thermo Fisher Cat# 176840025
Neutral Buffered Formalin Fisher scientific, EprediaTM 5701 Cat# 22-050-104
Trizol LS Thermo Fisher Scientific Cat# 10296028
Tween 20 Applied Chemicals Cat# A4974
Triton X-100 Applied Chemicals Cat# A4975
Blocking Reagent (PNB) Perkin Elmer Cat# FP1020
SlowFade Diamond Antifade Mountant Invitrogen Cat# S36972
Fluorescence Mounting Medium Dako, Agilent Cat# S302380
TCEP tris(2-carboxyethyl)phosphine Sigma Cat# C4706
Urea Panreac Applichem ITW reagents Cat# A1360
Guanidium Chloride Carl Roth Cat# 0037.1
Glycine Panreac Applichem ITW reagents Cat# A1067
Bovine Serum Albumin Jackson ImmunoResearch Cat# 001-000-162
UltraPure 0.5M EDTA Invitrogen Cat# 15575020
Normal donkey serum SigmaAldrich Cat# S30-M
Luminol Sigma Cat# A8511-5G
Diphenyleneiodonium chloride (DPI) Sigma Cat# D2926-10mg
Phorbol-12-myristat-13-acetate (PMA) Abcam Cat# abb120297
Endothelial Cell Medium ScienCell Cat# 1001
Astrocyte Medium ScienCell Cat# 1801
Trypsin Neutralization Solution ScienCell Cat# 0113
Recombinant human CSF-1 R&D Systems Cat# 216-MC-025
RLT buffer Qiagen DNAeasy Blood&Tissue kit Cat# 69504
Brilliant Stain Buffer BD Bioscience Cat# 563794
(Continued on next page)
ll
OPEN ACCESS
e4 Cell 186, 4546–4566.e1–e15, October 12, 2023
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Corning Matrigel Growth Factor
Reduced (GFR) Basement Membrane
Matrix, Phenol Red-free, LDEV-free
Corning Cat# 356231
Attane Isoflurane Attane N/A
2% lidocaı¨ne Streuli Pharma Cat# 784734
Bupivacaine 0.5% Carbostesin; Aspen Pharma Schweiz Cat# 972430
Buprenorphine 0.3 mg/ml (Temgesic) Eumedica Pharmaceuticals AG Cat# 6664853
Bepanthen cream Bayer N/A
HEPES (1M) Gibco, Thermo Fisher Cat# 15630056
Collagenase/Dispase Sigma-Aldrich Cat# 10269638001
DNase I Sigma-Aldric Cat# 11284932001
Pentobarbital CHUV Hospital,
Lausanne, Switzerland
NA
Tamoxifen,R99% Sigma-Aldrich Cat# T5648
BrBzGCp2 (GLO-1 inhibitor) MedChemExpress Cat# HY-136684
Recombinant Human TNF-a PreproTech Cat# 300-01A
Recombinant Human Glyoxalase I Protein, CF R&D Systems Cat# 4959-GL
Interleukin-6 human Sigma-Aldrich Cat# SRP3096
Recombinant Human IL-9 Protein R&D Systems Cat# 209-ILB
Ceruloplasmin human Sigma-Aldrich Cat# C4519-100UN
Critical commercial assays
Brain Tumor Dissociation Kit (P) Miltenyi Cat# 130-095-942
Tumor Dissociation Kit, human Miltenyi Cat# 130-095-929
Tumor Dissociation Kit, mouse Miltenyi Cat# 130-096-730
Myelin Removal Beads II kit Miltenyi Cat# 130-096-433
CD14 MicroBeads, human Miltenyi Cat# 130-050-201
Human TruStain FcX BioLegend Cat# 422302
Zombie NIR Fixable Viability Kit BioLegend Cat# 423106
eBioscienceTM Foxp3 / Transcription
Factor Staining Buffer Set
Thermo Fischer Scientific Cat# 00-5523-00
CellROX Green Reagent, for oxidative
stress detection
Thermo Fischer Scientific Cat# C10444
4’, 6- Diamidino-2-Phenylindole,
Dihydrochloride (DAPI)
Life Technologies Cat# D1306
MACSxpress Whole Blood
Neutrophil Isolation Kit
Miltenyi Cat# 130-104-434
CD66b Antibody, anti-human,
Biotin, REAfinity
Miltenyi Cat# 130-118-983
Anti-Biotin MicroBeads UltraPure Miltenyi Cat# 130-105-637
Anti-Ly-6G MicroBeads UltraPure Miltenyi Cat# 130-120-337
Easy HLA Chimerism Whole Blood
CD3 positive selection kit
Stemcell Cat# 17871
DNeasy Blood & Tissue Kit Qiagen Cat# 69504
eBioscience Annexin V Apoptosis Detection Kits Thermo Fischer Scientific Cat# 88-102-72
Assay Diluent MSD Cat# K151AEL-1
U-PLEX Immuno-Oncology Group 1 (hu) assay MSD Cat# K151AEL-1
L1000 Glass Slide Human Antibody Array RayBiotech Life Cat# AAH-BLG-1000-4
Deposited data
Raw data This paper https://joycelab.shinyapps.io/braintime/
(Continued on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566.e1–e15, October 12, 2023 e5
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Human reference genome hg38, GENCODE v36 Gencode https://www.gencodegenes.org/
human/release_36.html
scRNA-seq NSCLC TAN dataset Zilionis et al.36 https://singlecell.broadinstitute.
org/single_cell/study/SCP739/
single-cell-transcriptomics-of-
human-and-mouse-lung-
cancers-reveals-conserved-
myeloid-populations-across-
individuals-and-species
scRNA-seq NSCLC TAN dataset Salcher et al.37 https://zenodo.org/record/7227571
scRNA-seq Melanoma-BrM TAN dataset Alvarez-Breckenridge et al.38 https://singlecell.broadinstitute.
org/single_cell/study/SCP1493/
microenvironmental-correlates-
of-immune-checkpoint-inhibitor-
response-in-human-melanoma-
brain-metastases-revealed-by-t-
cell-receptor-and-single-cell-
rna-sequencing
STRING Protein-Protein-Interaction
database, version 11.5
Szklarczyk et al.77 https://string-db.org
Experimental models: Cell lines
Human: human brain microvascular
endothelial cells (HBMEC)
ScienCell Cat# 1000
Human: Primary human astrocytes (HA) ScienCell Cat# 1800
Human: human umbilical vein
endothelial cells (HUVEC)
Prof. T. Petrova N/A
MMTV-PyMT (murine mammary tumor virus;
Polyoma middle T antigen) 99LN BrM3
Croci et al.78
Prof. J.A. Joyce
Available upon request
PDGFB-HA: DF1 chicken fibroblasts (ATCC)
with RCAS viral vector expressing PDGFB-HA
Dr. Tatsuya Ozawa and
Dr. Eric Holland
N/A
shP53 DF1: DF1 chicken fibroblasts (ATCC)
with a short hairpin against murine p53 (shP53)
Dr. Tatsuya Ozawa and
Dr. Eric Holland
N/A
Experimental models: Organisms/strains
C57BL/6J mice The Jackson Laboratory RRID:IMSR_JAX:000664
Ly6GCreERT2 mice (C57BL/6J background) Ballesteros et al.20 N/A
Nestin-Tv-a mice (C57BL/6J background) Holland et al.79 N/A
Software and algorithms
FlowJo, version 10.8.2 BD bioscience https://www.flowjo.com/
GraphPad Prism v9.4.1 GraphPad Software https://www.graphpad.com/
scientific-software/prism/
ZEN software Zeiss https://www.zeiss.com/
microscopy/en/products/
software.html
VIS Image Analysis, version 2022.04 Visiopharm https://www.visiopharm.com/
QuPath version 0.3.2 Bankhead et al.80 https://qupath.github.io/
StarDist Schmidt et al.81 https://github.com/stardist/stardist
Python version 3.9 Python https://www.pyhton.org
R environment, version 4.1.1 The R Foundation https://www.r-project.org/
RStudio, version 2023.03.0 RStudio https://www.posit.co/
Matlab (R2022a) MathWorks https://www.mathworks.
com/products/matlab.html
STAR aligner, version 2.7.7a Dobin et al.82 https://github.com/alexdobin/STAR
(Continued on next page)
ll
OPEN ACCESS
e6 Cell 186, 4546–4566.e1–e15, October 12, 2023
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
RSEM version 1.3.3 Li and Dewey83 https://deweylab.github.io/RSEM/
Spatstat R package version 2.3-4 Baddeley et al.84 http://spatstat.org
edgeR version 3.32.1 Robinson et al.85 https://bioconductor.org/
packages/release/bioc/
html/edgeR.html
limma version 3.46.0 Ritchie et al.86 https://bioconductor.org/
packages/release/bioc/
html/limma.html
sva version 3.38.0 Leek and Storey87 https://bioconductor.org/
packages/release/bioc/
html/sva.html
ClusterProfiler R package version 3.18.1 Wu et al.88 https://bioconductor.org/
packages/release/bioc/
html/clusterProfiler.html
scDataviz R package Blighe 89 https://github.com/kevinblighe/scDataviz
ggplot2 R package version 3.3.6 Wickham90 https://ggplot2.tidyverse.org
Monocle3 R package version 1.0.0 Trapnell et al.91 https://cole-trapnell-lab.
github.io/monocle3/
Slingshot R package version 2.2.0 Street et al.92 https://github.com/kstreet13/slingshot
Cutadapt version 2.3 Martin93 https://github.com/marcelm/
cutadapt/blob/v2.3/doc/index.rst
Burrows-Wheeler aligner version 0.7.17 Li and Durbin94 https://github.com/lh3/bwa
SAMtools version 1.8 Danecek et al.95 https://github.com/samtools/samtools/
Picard tools version 2.9.0 Broad Institute https://broadinstitute.github.io/picard/
ENSEMBL Variant Effect Predictor version 96 McLaren et al.96 https://www.ensembl.org/info/
docs/tools/vep/index.html
ComplexHeatmap R package version 2.10.0 Gu et al.97 https://jokergoo.github.io/
ComplexHeatmap-reference/book/
iGraph R package version 1.2.6 Csardi and Nepush98 https://igraph.org
lme4 R package version 1.1-32 Bates et al.99 https://github.com/lme4/lme4/
lmerTest R package version 3.1-3 Kuznetsova et al.100 https://github.com/runehaubo/lmerTestR
Other
gentleMACS Octo Dissociator Miltenyi Cat# 130-095-937
gentleMACS C Tubes Miltenyi Cat# 130-096-334
LS Columns Miltenyi Cat# 130-042-401
SepMate-50 tubes StemCell Cat# 85450
PermaLife Cell Culture Bags OriGen Biomedical Cat# PL30-2G
LSR II flow cytometer BD Bioscience N/A
Fortessa flow cytometer BD Bioscience N/A
FACSAria III, flow cytometer & cell sorter BD Bioscience N/A
Axio Scan.Z1 slide scanner Zeiss N/A
Omni Tissue Homogenizer (TH) Omni International Cat# TH220
Tissue-Tek Cryomold Sakura Cat# 4557
Hydrophobic pen Daido Sangyo, Plano Cat# 22304
Coverslip ThermoFisher Cat# BB02400500A113FST0
Microscope slide Fisherbrand Cat# 12-550-15
Illumina NovaSec 6000 sequencer Illumina N/A
MESO QuickPlex SQ120 instrument Meso Scale Discovery N/A
Nunc MicroWell 96-Well, Nunclon
Delta-Treated, White Flat-Bottom Microplate
ThermoFisher Cat# 136101
Microplate reader Infinite 200 PRO Tecan Tecan N/A
(Continued on next page)
ll
OPEN ACCESS
Cell 186, 4546–4566.e1–e15, October 12, 2023 e7
Article
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Trucount Absolute Counting Tubes BD Bioscience Cat# 340334
Corning 96-well Clear Round Bottom
TC-treated Microplate, Individually
Wrapped, with Lid, Sterile
Corning Cat# 3799
Corning 96-well Clear Flat Bottom
Polystyrene TC-treated Microplates,
Individually Wrapped, with Lid, Sterile
Corning Cat# 3596
BD Vacutainer BD Bioscience Cat# 368861
Vetbond tissue adhesive 3M Cat# 1469SB
ll
OPEN ACCESS ArticleRESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Prof. Jo-
hanna Joyce (johanna.joyce@unil.ch).
Materials availability
This study did not generate new unique reagents.
Data and code availability
Bulk RNA-seq count data generated and used for this study can be queried and downloaded at: https://joycelab.shinyapps.io/
braintime/
Due to strict privacy protection, the raw RNA-seq data will be made available when possible. Future users can contact the corre-
sponding author for access to the raw, unprocessed RNA-seq data, and those requests will then be individually reviewed by the rele-
vant institutional committees. This paper does not report original code.
EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS
Human subjects
All procedures in this study involving the use of tissues derived from human participants were in accordance with the ethical stan-
dards of the national research committees and the declaration of Helsinki.
For detailed information related to the cohort of human participants (e.g. sex, age, and treatment), please see Tables S1A–S1C.
Informed consent was obtained for all human subjects participating in this study. The processing of non-tumor and tumor tissue
at the Biobank of the Brain and Spine Tumor Center (BB_031_BBLBGT) at the Centre Hospitalier Universitaire Vaudois (CHUV, Lau-
sanne, Switzerland), was approved by the Commission cantonale d’e´thique de la recherche sur l’eˆtre humain (CER-VD, protocol PB
2017-00240, F25 / 99). Similarly, the processing of tumor tissue collected at Memorial Sloan Kettering Cancer Center (MSKCC, New
York, NY, USA) was approved by the institutional review board (IRB, protocols #IRB #06-107, #14-230). Non-tumor brain surgeries
were performed on patients with therapy-resistant focal epilepsy, as part of the clinical management. Brain tumor resections were
performed to either aid in diagnosing the origin of the lesion or as part of the therapeutic strategy. Diagnostic identification of the brain
(tumor) tissue was performed by a trained pathologist as part of the standard of care at the CHUV or MSKCC.
Matched blood of brain tumor-bearing patients was collected at the time of the brain surgery as part of the ethical permit stated
above (CER-VD, protocol PB 2017-00240, F25 / 99). Whole blood collection from identified healthy donor volunteers was approved
by CER-VD (protocol 2018-00492). Buffy coats from anonymous voluntary donors were obtained from the Transfusion Interre´gionale,
Croix-Rouge Suisse (Epalinges/Lausanne, Switzerland).
All tissue specimens were coded before further handling in accordance with patient privacy regulations.
Primary cell cultures
Human
Brain tumor-derived cell lines were generated by plating single-cell suspensions of dissociated tumors in a cell culture flask at a den-
sity of 2 x 106 cells per ml in DMEM-F12 (1:1) +Glutamax (Gibco, cat. no. 31331028) +10% FCS (Gibco, ThermoFisher, cat. no.
10270106) +1% penicillin/streptomycin (P/S, Gibco, cat. no. 15140122) in an incubator at 37C and 5% CO2. Cells were passaged
using 0.05% Trypsin-EDTA (Gibco, cat. no. 25300062) once the cells reached 70% confluency in the cell culture flask. Cell lines were
validated using whole exome sequencing (WES) of both snap frozen tumor tissue and matched tumor cell lines. For detailed infor-
mation (e.g. sex, age, treatment) related to brain tumor-derived cell lines, please see Table S1C.e8 Cell 186, 4546–4566.e1–e15, October 12, 2023
ll
OPEN ACCESSArticlePrimary human astrocytes (HA; ScienCell, cat. no. #1800) and human brain microvascular endothelial cells (HBMEC; ScienCell,
cat. no. #1000) were cultured according to the vendor’s instructions. Human umbilical vein endothelial cells (HUVECs) were kindly
provided by the laboratory of Prof. T. Petrova, and were cultured in DMEM-F12 (1:1) +GlutaMAX +10% FBS +1%P/S. Healthy donor
monocyte-derived macrophages (MDMs) were derived from peripheral blood mononuclear cells (PBMCs) from buffy coats, isolated
using a Ficoll (GE Healthcare, cat. no. 17-1440-02) gradient and SepMate tubes (StemCell, cat. no. 85450). Monocytes were selected
using MACsorting by CD14 MicroBeads (Miltenyi, cat. no. 130-050-201) and differentiated into macrophages by culture in Teflon-
coated bags (OriGen Biomedical, cat. no. PL30-2G) for 7 days in DMEM-F12 (1:1) +GlutaMAX +10% FBS +1% P/S with the addition
of 10 ng/ml recombinant human CSF-1 (R&D Systems, cat. no. 216-MC-025).
Murine
To generate murine breast-to-brain metastasis (BrM) a brain-homing breast tumor cell line, MMTV-PyMT (murine mammary tumor
virus; Polyoma middle T antigen)-BrM3, was used (C57BL/6J background).78 This cell line was initially derived from a murine breast-
to-lymph nodemetastasis and consequently passaged three times in vivo in female C57BL/6JWTmice for brain-homing capacity via
intracardiac injections. The PyMT-BrM3 cell line was maintained in DMEM-F12 (1:1) +GlutaMAX medium +10% FCS +1% P/S and
passaged using 0.05% Trypsin-EDTA once the cells reached 70% confluency.
Transfected DF1 chicken fibroblasts
DF1 chicken fibroblasts (ATCC), which were previously transfected with an RCAS viral vector expressing PDGFB-HA or a short
hairpin against murine p53 (shP53) were kindly provided by Dr. Tatsuya Ozawa and Dr. Eric Holland.101,102 The PDGFB-HA and
p53 DF1 cells were cultured in DMEM-F12 (1:1) +GlutaMAX medium +10% FCS +1% P/S under standard conditions.
Mouse models
All experiments performed in this study were approved by the local Institutional Animal Care and Use Committees of the University of
Lausanne and the Canton Vaud, Switzerland (protocol numbers VD3314, VD3444, VD3688). Experiments involving the development
of breast(-BrM) tumors were performed in 6-10-week-old female Ly6GCreERT2 mice in a C57BL/6J background20 (generously pro-
vided by Prof. Andre´s Hidalgo) or in female C57BL/6J wildtype mice. Experiments involving the development of gliomas were per-
formed in 4.5–7-week-old male and female Ly6GCreERT C57BL/6J mice bred to Nestin-Tv-a mice or in Nestin-Tv-a wildtype mice79
(generously provided by Dr. Eric Holland). All mice were housed in the Agora In Vivo Center (AIVC) animal facility in individually venti-
lated cages. Mice were housed under a 12-hour light/dark schedule at 22C, in the presence of 2-5 cagemates. Standard autoclaved
lab diet and water were provided.
METHOD DETAILS
Human brain (tumor) enzymatic digestion
Single cell suspensions of human non-tumor and brain tumor tissue were obtained after enzymatic digestion as described in Maas
et al.23 (Module 2). In brief, enzymatic digestion was performed using the ‘‘Brain Tumor Dissociation Kit (P)’’ (BTDK, Miltenyi, cat. no.
130-095-942) for non-tumor brain tissue and gliomas, for BrMs the ‘‘Tumor Dissociation Kit, human’’ (Miltenyi, cat. no. 130-095-929)
was used according to the manufacturer’s protocol. For non-tumor brain and glioma samples, enzymatic digestion was followed by
myelin removal using Myelin Removal Beads II according to the manufacturer’s protocol (Miltenyi, cat. no. 130-096-433). Both tumor
single cell suspensions and whole blood were RBC-lysed prior to downstream processing using 1x RBC lysis buffer (Biolegend, cat.
no. 00-4333-57, diluted to 1X in deionized water) according to the manufacturer’s protocol. Single-cell suspensions were used for
flow cytometric (FCM) analysis, bulk-RNA sequencing (RNA-seq) after fluorescence-activated cell (FAC)-sorting, ex vivo functional
analysis as well as the generation of tumor-derived cell lines.
Flow cytometry and FACS
FCM and FACS were performed as described in Maas et al.23 (Module 2a, see key resources table for the antibodies used). In brief,
cells were stained in 100ul ZombieNIR Fixable Viability dye (Biolegend, cat. no. 423106) for 20 min at RT in the dark, followed by
10 min incubation at RT in the dark with Fc receptor blocking solution (5 ul added per sample, BioLegend, cat. no. 422302). Next,
1-5 ul of 2x antibody staining mix in Brilliant Stain Buffer (BD; cat. no. 563794) was added and incubated for 15 min at 4 C in the
dark. After washing the cells with FACS buffer (PBS (Gibco, ThermoFisher, cat. no. 20012027), 0.5% BSA (Jackson
ImmunoResearch, cat. no. 001-000-162), 2mM EDTA (Thermo Fischer Scientific, cat. no. 15575020)) samples were either directly
acquired using a flow cytometer or fixed according to themanufacturer’s protocol (eBioscience Foxp3 / Transcription Factor Staining
Buffer Set, Thermo Fischer Scientific, cat. no. 00-5523-00). The desired cell populations were directly purified using FACS in 750 ul
Trizol LS (Sigma-Aldrich, cat. no. T3934) and snap-frozen in liquid nitrogen. Validation of the sorting purity (>97%) was assessed by
re-acquisition of the sorted populations by FCM.23 Technical consistency of the flow cytometer setup, allowing for comparison of
data across different timepoints, was ensured by saving a baselinemeasurement on the flow cytometer used on the first day of acqui-
sition. On consecutive dates, the baselinewas reloaded and adapted based on the read-out of the acquiredCS&T beads (BD, cat. no.
656504) in the different channels, maximizing a consistent read-out. For the measurement of ROS by FCM CellROX green (Thermo
Fischer Scientific, cat. no. C10444) was used according to the manufacturer’s protocol.Cell 186, 4546–4566.e1–e15, October 12, 2023 e9
ll
OPEN ACCESS ArticleImmunofluorescence staining of tissue sections
Human tissue was frozen in optimal cutting temperature (OCT) compound (Tissue-Tek, Sakura Finetek, cat. no. 4583) as described in
Maas et al.23 (Module 1). In short, a 2-methylbutane (Sigma-Aldrich, cat. no. 50970) container was placed on dry ice to ensure a tem-
perature of -80
C. Tumor tissue (minimum size of 8mm3) was placed in a cryomold with OCT and dipped in chilled 2-methylbutane for
2-3min. OCT blocks were stored at -80
C until they were sectioned using a cryostat. 10-um thick frozen sections were used for both
standard single-round immunofluorescence (IF) staining, as well as sequential IF. Standard single-round IF was performed on non-
tumor brain tissue, IDHmut andWT gliomas, and BrMs, and described inMaas et al.23 (Module 1, and below) and for sequential IF up
to 7 rounds, frozen tissue slides from IDHWT gliomas and BrMs were used. Sections were air-dried for 20-30 min in a laminar flow
hood. Slides for single-round IF were fixed for 10 min in 99.9% methanol (Thermo Fisher, cat. no. 176840025), prechilled to -20
C.
Slides used for sequential IF were fixed in 10%Neutral Buffered Formalin (Fisher Scientific, EprediaTM 5701, Cat. No. 22-050-104) for
40 min at RT. For both single round and sequential IF the slides were then rehydrated and washed 3x 5 min in PBS while gently rock-
ing. A ring was drawn around the tissue using a hydrophobic pen (Daido Sangyo, Plano, cat. no. 22304). Any potential autofluores-
cence signal was quenched by adding 100 ul glycine 10mM (Panreac Applichem ITW reagents, A1067) diluted in PBS for 10 min at
RT. Slides were washed 2x 5 min in PBS-0.2%Tween (PanReac AppliChem cat. no. A4974). Tissue was permeabilized with
PBS-0.2%Triton (PanReac AppliChem, cat. no. A4975) for 3h at RT if non-tumor brain and IDH mut gliomas slides were included
for the study, otherwise a 10 min permeabilization sufficed for IDH WT and BrM slides. Next, slides were washed 2x 5 min with
PBS-0.2%Tween. The slides were incubated with blocking buffer (1x Blocking Reagent (PerkinElmer, cat. no. FP1012; ‘PNB’) +
0.5% Tween + 10% normal donkey serum (SigmaAldrich, cat.no. S30-M) + 2% BSA (Jackson ImmunoResearch, cat. no. 001-
000-162)) filtered through a 0.22-um filter, for 1 hour in a humidified chamber. Primary antibodies (see key resources table for the
antibodies used) were diluted in antibody dilution buffer (1x PNB + 0.5% Tween + 10%normal donkey serum; 0.22-um filtered before
use) for 3h in a humidified chamber on a shaker at RT. Slides were then washed 3x 5 min, before adding fluorophore-conjugated
secondary antibodies and a nuclear detection marker 4’, 6- Diamidino-2-Phenylindole, Dihydrochloride (DAPI, Life technologies,
D1306) in antibody dilution buffer to the slides (see key resources table for the antibodies used). Slides were placed in a humidified
chamber for 1 hour at RT. Staining with secondary antibodies alone was used as the control. The slides were washed 6x 10 min with
PBS-0.2%Tween, and a final wash was performed with PBS.To mount the slides for single-round IF, 2 drops of fluorescence
mounting media (Dako, cat. no. S302380) were added to the tissue. Sequential IF slides were mounted by adding 20 ul SlowFade
Diamond Antifade Mountant (Thermo Fischer Scientific, cat. no. S36972) to each tissue. Then a coverslip (ThermoFisher, cat. no.
BB02400500A113FST0) was carefully placed on top of the slides. The fluorescent signal was acquired using the Axio Scan.Z1 slide
scanner (Zeiss), with a Colibri 7 LED light source (Zeiss) and a Plan-Apochromat 20x/0.8 DIC M27 cover slip-corrected objective
(Zeiss). For sequential IF, the coverslip was carefully removed after acquisition by leaving the slide upright in PBS to let the coverslip
slide off. The slides were washed 3x 5 min in PBS, prior to elution in freshly prepared tris(2-carboxyethyl)phosphine (TCEP)-based
elution buffer (0.5 M glycine + 3 M guanidium chloride (Carl Roth, cat. no. 0037.1) + 2 M urea (Panreac Applichem ITW reagents, cat.
no. A1360) + 40 mM TCEP (Sigma, cat. no. C4706) in ddH2O) for 5 min on a rocker at RT.
103 Finally, slides were washed 3x 5 min in
PBS-Tween, before restarting the staining process by returning to adding blocking buffer to the slides, and following all the consec-
utive steps detailed above.
Image analysis and cell type identification
Standard single-round IF
Image quantificationwas performed using the VIS Image Analysis software (Visiopharm, v2022.04) on preprocessed images applying
ZEN software (stitching and z-stacking of the images). Cellular identification, as well as perivascular niche (PVN) characterization,
were performed as described in Klemmet al.10 Briefly, the tissue outline was detected by applying a 21 pixel meanDAPI filter. Nuclear
detection was based on the DAPI watershed signal and filtered by area to exclude anomalies. The nuclear label was expanded by 5
pixels to capture nuclear and cytoplasmic fluorescent signal. A hierarchical decision tree with manually set thresholds was applied to
identify the different cell types. Vessels were identified using a separate mask based on pixel intensity of the CD31 signal. Nuclear
classifiers were excluded unless their signal exceeded the threshold for CD31. By generating an ROI around the vessels at a 20 um
distance, a perivascular niche (PVN) region was created.
Sequential IF
Image preprocessing was performed using ZEN software. The images of different rounds of staining were stitched, and background
subtraction was performed using the rolling ball method with a radius of 75.104 Using the signal obtained from single staining of the
tissue with DAPI alone allowed for the subsequent subtraction of autofluorescence signal. Autofluorescence subtraction and align-
ment of the final image made from the sequential rounds of staining were performed in a Python (v3.9) library script (Watson, Joyce
et al., unpublished data).
Image quantification was performed using QuPath (v0.3.2) open-source image analysis software.80 Tumor tissue was detected
within each sample using the ‘‘Pixel classifier’’ command. First, a training image made of single regions of interest (ROIs) drawn in
each sample was generated. The ‘‘Pixel classifier’’ was trained on these images, using a positive detection class for tissue detection.
Then, the trained pixel classifier was applied to the entire image dataset, followed by cleanup of the resulting binary image by
morphological operations. Nucleus detection was performed by StarDist, a deep-learning-based method from Schmidt et al., using
the dsb2018_heavy_augment.pb model.81 A cell expansion value of 3 was used and subsequently, area identification (PVN ande10 Cell 186, 4546–4566.e1–e15, October 12, 2023
ll
OPEN ACCESSArticlevessels) was performed with the ‘‘Pixel classifier’’ command according to the same method described above. Finally, cell identifica-
tion was performed using QuPath’s ‘‘Object classifier’’ command. Training was performed on 40% of the ROIs and validated on the
remaining 60%. A classifier was generated from one or several mutually exclusive markers. Training of the classifier was based on
annotations (minimum 10 annotations per class). Selected features for training included all marker measurements. Composite clas-
sifiers generated from sequentially added classifiers were created to allow final cell identification. These classifiers were applied to
the entire project and data export in.csv format for subsequent analysis in R (v4.4.1). Filtering of cells using a diameter size of > 4 um
and < 12.5 um, detected nuclei probability > 0.65, as well as filtering of the relevant final cell populations, was performed in R.
Nearest neighbor distance and neighborhood analysis of sequential IF data
Nearest neighbor distance from different T cell populations andCD8+ T cells stratified by PD-1 expression to TANs in IDHWTgliomas
and BrMs, was assessed using the spatstat R package (v2.3-4).105 Neighborhood analysis of PD-1+ CD8+ T cells and TANs was per-
formed based on the occurrence of cells within a 20mm radius of the border of the nuclei, calculated using the radius of individual
nuclei.
RNA-seq analysis
Isolation and sequencing of RNA was performed by Genewiz (South Plainfield, New Jersey, USA) using chloroform extraction and
isopropanol precipitations as previously described.10,23 The SMART-Seq preparation kit (CloneTech) was used to generate RNA-
seq libraries and fragmented using the Nextera XT kit (Illumina). Sequencing of paired end, 100 or 150 base pair and single end
100 base pair, was performed on an Illumina HiSeq 2500 (Illumina).
Alignment of RNA-seq reads was performed using STAR (v2.7.7a)82 and quantified using RSEM (v1.3.3).83 Human genome v38
was used with GENCODE v36 annotation. Raw counts of transcripts with the same gene symbol were pooled. All genes were hier-
archically clustered using Ward distance, and the expression heatmap was visually examined to identify a major branch containing a
non-informative expression pattern. Non-protein coding genes, pseudogenes, predicted genes, genes with less than 1 FPKM in at
least one sample in a sample group (tissue/disease group), and genes with low variance were filtered out. A total of 16,052 genes
were retained and their expression was normalized between samples using the TMMmethod (edgeR v3.32.185) and log2 transformed
with voom (limma v3.46.086). Batch effect was corrected for with ComBat (sva v3.38.087). Differential expressionwas computed using
limma (v3.46.0).86 To reduce the potential bias introduced by any contaminating tumor cells (CD45
 cells), genes with higher expres-
sion in the CD45
 cells vs. neutrophils in the leading comparison group were filtered out if they additionally were found to be signif-
icantly differentially expressed in CD45
 cells, CD4+ T cells or CD8+ T cells within the same comparison. Geneswith p.adj value <0.05
and -1>LFC>1 are called significantly differentially expressed in all comparisons except when non-tumor brain (n=5) was one of the
comparison groups, then genes with p-value <0.005 and -1>LFC>1 were called significantly differentially expressed. Pathway ana-
lyses (ORA and GSEA) were performed using clusterProfiler R package (v3.18.1).88 For analysis across the major TME populations,
T cells contain both CD4+ T cells and CD8+ T cells.
Mouse tumor models
Breast-BrM model
For the development of breast-BrMs in immunocompetent mice, 6-10-week-old female Ly6GCreERT2mice or WTmice were anesthe-
tized using isoflurane inhalation (O2 + 2% isoflurane) and shaved on their chest. Mice were placed with their back down on a heating
pad, and 1 x 105 PyMT-BrM3 cells resuspended in 100ul HBSS (Gibco, cat. no. 14175-095) were injected in the left cardiac ventricle.
Mice were monitored weekly by magnetic resonance imaging (MRI).
Primary breast tumor model
6-10-week-old female Ly6GCreERT2 mice or WT mice were anesthetized using isoflurane inhalation (O2 + 2% isoflurane) and placed
with their back on a heating pad. 7 x 103 PyMT-BrM3 cells were resuspended in HBSS andmixed 1:1 with Matrigel (Corning, cat. no.
356231) in 100 ul total volume. Cells were injected in the right abdominal mammary fat pad (MFP). Mice were monitored 3x per week
to assess the tumor size by calipers measurement.
Glioma tumor model
4.5 to 7 week-old male and female immunocompetent Ly6GCreERT;Nestin-Tv-a mice or Nestin-Tv-a mice were used to generate pri-
mary brain tumors using the RCAS system as previously described.106,107 In brief, mice were anesthetized using isoflurane inhalation
(O2 + 2% isoflurane), and a mixture of 0.5% bupivacaine (Carbostesin; Aspen Pharma Schweiz, cat. no. 972430) and 2% lidocaı¨ne
(Streuli Pharma, cat. no. 784734) was applied as a local analgesic (50 ul per mouse), and 0.3 mg/ml buprenorphine (Temgesic;
Eumedica Pharmaceuticals AG, cat. no. 6664853) was injected subcutaneously as a systemic analgesic (100 ul per mouse). Using
a stereotactic apparatus, transfected DF1 cells were injected into the right frontal cortex (2 mm frontal, 1.5 mm lateral from bregma,
2 mm deep). Mice were injected with a 1:1 mixture of PDGFB-HA and shP53 DF1 cells, for a total of 3 x 105 DF1 cells intracranially.
The skin of the skull was sealed using Vetbond tissue adhesive (3M, cat. no. 1469SB). The mouse was placed on a heating pad and
monitored until fully recovered from anesthesia. Finally, Bepanthen cream (Bayer) was applied on the site of the incision prior to
placing the animal back in the cage. Mice were carefully monitored, and tumor development was assessed by weekly MRI.
Tumor-bearing Ly6GCreERT and Ly6GCreERT;Nestin-Tv-a mice were injected intraperitoneally with 400 ul of 10 mg/ml
Tamoxifen (Sigma-Aldrich, cat. no. T5648), 0-8 days prior to sacrificing to follow the decay of iLy6G TdTomato+ neutrophils in
different organs.Cell 186, 4546–4566.e1–e15, October 12, 2023 e11
ll
OPEN ACCESS ArticleMouse tissue analysis
Whole blood was collected from mice prior to sacrifice via the submandibular vein. Mice were euthanized by terminal anesthesia by
intraperitoneal injection of pentobarbital (150mg/kg; CHUVHospital, Lausanne, Switzerland), followed by transcardial perfusion with
PBS. Spleen, femur, lungs, mammary fat pad and brain were resected from eachmouse and placed on ice. Single cell suspensions of
the different organswere obtained using the followingmethods: i) The spleen was smashed through a 70-um cell strainer. ii) The bone
marrow was flushed out of the femur by cutting off the crown of the bone on one end, placing the bone upside-down into a pierced
0.2 mL vial inside a 1.5 mL vial and centrifuging for 10s at maximum velocity in an Eppendorf centrifuge. iii) Lung tissue was cut in
1 mm3 pieces, resuspended in digestion mix (HEPES buffer (Gibco, cat. no. 15630056) + 2mg/ml Collagenase D (Sigma-Aldrich, cat.
no. 10269638001) + 80U/ml DNase I (Sigma-Aldrich, cat. no. 11284932001)) and transferred to a gentleMACSC tube (Miltenyi, cat.
no. 130-096-334). The C tube was placed in the gentleMACS Dissociator (Miltenyi, cat. no. 130-096-427) and following programs
were run consecutively: ‘‘m_lung_02_01’’ (165 rounds per run (rpr) for 0.36 s at RT), 1500 rpr for 30 min at 37C and finally 2079
rpr for 37s at RT. The digested tissue was filtered through a 70-um cell strainer. iv) Brain (tumor) and MFP (tumor) tissue was cut
in 1 mm3 pieces, placed in a C-tube and resuspended in Tumor Dissociation Kit, mouse (TDK, Miltenyi, cat. no. 130-096-730), ac-
cording to themanufacturer’s instructions. The C tubewas placed in the GentleMACS dissociator and the ‘‘37C_m_TDK_1’’ (1081 rpr
at 37C for 41 min) digestion program was run. The digested (tumor) tissue was filtered through a 40-um cell strainer. For healthy
brain tissue, subsequent myelin removal using the Myelin Removal Beads II kit (Miltenyi, cat no. 130-096-433) was performed ac-
cording to the manufacturer’s instructions. Apart from healthy brain, all tissues were subjected to red blood cell lysis (Biolegend,
cat. no. 00-4333-57, diluted to 1X in deionized water) prior to FCM staining.
Mathematical modeling of mean neutrophil lifetime
To assess neutrophil mean lifetimes, the proportion of tdTomato+ neutrophils was normalized to the maximum proportion measured
0-8 days post tamoxifen injection. To track neutrophils in each tissue, an age-structured mathematical model was developed that
incorporated the phenotypic heterogeneity of neutrophils. Let u = uðt; aÞdenote the density of neutrophils which, at time t, have
an age a. The age varies in the intervala˛ ½0;amax, where amaxis the maximum age (or maximum life span) that a neutrophil could
have. To describe the temporal dynamics of their age distribution, the following age-structured model, via a linear first-order partial
differential equation, was used:
vu
vt
+
vu
va
= 
 uðt; aÞ
tðaÞ +4ðt; aÞ: (Equation 1)
The left-hand side of (Equation 1) represents the temporal change in the number of neutrophils and their age. The first term on the
right-hand side accounts for neutrophil death. The death time, tðaÞ, generally depends on neutrophil age. The addition of the flux 4ðt;
aÞaddresses the net effect of neutrophils entering and/or leaving the tissue. The total number of neutrophils at time t, irrespective of
age, is given by the following integral:
nðtÞ =
Z amax
0
uðt; aÞ da: (Equation 2)
When calculating the mean lifetime, tLT, of neutrophils in the context of one synchronous wave of neutrophils after tamoxifen
administration, the survival function SðaÞ, with a˛ ½0; amax can be applied, yielding the fraction of neutrophils that survive to an exact
age a:
tLT =
Z amax
0
SðaÞ da: (Equation 3)
As SðaÞis not available, an effective probability density function has been introduced here based on the solution to (Equation 1):
tLTðtÞ =
R amax
0
a uðt; aÞ da
nðtÞ : (Equation 4)
Since (Equation 4) depends on time t, computing a representative mean lifetime requires the temporal average of tLTðtÞ leading to
the following final formula:
CtLTD =
RN
0
tLTðtÞ nðtÞ dtRN
0
nðtÞ dt : (Equation 5)
Assuming that at time t = 0 no neutrophils of any age have yet arrived at tissue i, we have the initial condition uð0;aÞ = 0. Thus, the
exact solution to (Equation 1) is:
uðt; aÞ = e

R t
0
dx
tða
 xÞ
Z t
0
4ðx; a 
 t + xÞ e
R x
0
dh
tða
 t+hÞ dx: (Equation 6)e12 Cell 186, 4546–4566.e1–e15, October 12, 2023
ll
OPEN ACCESSArticleIn the nonlinear regressions shown in Figures 5D, 5E, and S5E, we used for the neutrophil death time tðaÞ = ti. Furthermore, to
describe the synchronous wave of neutrophils recruited at tissue i, a bivariate Gaussian profile was assumed for the flux function in
(Equation 1):
4ðt; aÞ = 40 exp
"

 1
2ð1 
 r2Þ
 
ðt 
 tiÞ2
s2t

 2rðt 
 tiÞða 
 aiÞ
stsa
+
ða 
 aiÞ2
s2a
!#
: (Equation 7)
The parameter set, comprising ti, ti, ai, st, sa and r (notice that 40was not needed), was estimated from the measured data for the
proportion of tdTomato+ neutrophils in both non-tumor and PyMT-BrM3-bearing mouse brains, together with themodel Equations 1,
2, 3, 4, 5, 6, and 7 described above. The standard deviation dtLT of the mean lifetime CtLTD was computed via
dtLT
2 =
RN
0
ðtLTðtÞ 
 CtLTDÞ2 nðtÞ dtRN
0
nðtÞ dt : (Equation 8)
To perform the regression analysis, we defined a mean squared error function in terms of the normalized total number of neutro-
phils for the different datasets and the values predicted by our model Equation 2 at the specific measured times. When using the
neutrophil death time tðaÞ = ti, exact closed-form expressions for (Equation 6) were obtained. All remaining integrals entering in
Equations 2, 5 and 8 were computed numerically using Matlab (R2022a) global adaptive quadrature function integral.
Murine PBN and TAN isolation
Whole blood from healthy and tumor-bearing mice was collected by intracardiac aspiration into a BD Vacutainer tube (BD, cat. no.
368861) after post-terminal anesthesia was administered. Whole blood and enzymatically digested brain tumors were lysed via red
blood cell lysis (Biolegend, cat. no. 00-4333-57, diluted to 1X in deionized water) for 10min at RT. Neutrophils were isolated using
Magnetic-activated cell sorting (MACS) where samples were subjected to positive selection via anti-Ly-6G MicroBeads UltraPure
magnetic bead isolation (Miltenyi, cat. no. 130-120-337) according to the manufacturer’s instructions. This allowed to extract
both peripheral blood neutrophils (PBNs) from whole blood and tumor-associated neutrophils (TANs) from the dissociated brain
tumors.
Human PBN and TAN isolation
Whole blood from healthy donors (HD) and patients was collected in BD Vacutainer tubes (BD 368861). PBNs were isolated using the
MACSxpress Whole Blood Neutrophil Isolation Kit (Miltenyi, 130-104-434) according to the manufacturer’s instructions, followed by
red blood cell lysis for 10 min at RT (Biolegend, 00-4333-57, diluted to 1X in deionized water). TANs were isolated from enzymatically
digested human brain tumors using Magnetic-activated cell sorting (MACS) as described in Maas et al.23 (Module 3). Briefly, a pos-
itive selection method was applied by adding 1 ul anti-human CD66B-Biotin antibody (Miltenyi, cat. no. 130-118-983) to 100 ul of the
enzymatically digested brain tumor suspension and incubated for 10 min at 4
C. After washing with FACS buffer 1 x 108 cells were
resuspended in 80 ul FACS buffer. 20 ul Anti-Biotin MicroBeads UltraPure (Miltenyi, cat. no. 130-105-637) were added per 1 x 107
cells, mixed well, and incubated for 15 min at 4
C. The cell suspension was washed in 1-2 ml of FACS buffer and up to 1 x 108 cells
resuspended in 500 ul FACS buffer. MACS of TANs, defined as CD66B+ cells, was performed with LS columns (Miltenyi, cat. no. 130-
042-401) according to the manufacturer’s protocol.
Reactive oxygen species (ROS) detection
Isolated PBNs were resuspended at 5 x 106 cells/ml in starved media (DMEM-F12 w/o phenol red, Gibco, 21041-025) and 1 x 104
murine PBN/TANs were plated in 100ul serum-starved media in a white flat bottom 96-well plate (ThermoFisher, 136101). For human
HD and patient PBNs 5 x 105 cells were plated in 100 ul per well. For human PBN/TAN co-cultures, 5 x 104 neutrophils were plated per
well and CD66B
 TME cells were added at a ratio 5:1 in 100 ul final volume. Luminol (Sigma, A8511-5G) was added to a final con-
centration of 100 uM. Inhibition of ROS production by PBNs was achieved by adding diphenyleneiodonium chloride (DPI, Sigma,
D2926-10mg) at a final concentration of 50 uM; stimulation of ROS production was achieved by adding phorbol-12-myristat-13-ac-
etate (PMA, abcam, abb120297) at a final concentration of 4 uM. The plate was gently shaken prior to read-out by measuring all
wavelengths in a luminescence plate-reader (Microplate reader Infinite 200 PRO Tecan) at an interval of 5min for 2h.
T cell cytokine array
Peripheral CD3+ T cells were isolated from patient whole blood using the Easy HLAChimerismWhole BloodCD3 positive selection kit
(Stemcell, cat. no. 17871) according to themanufacturer’s protocol. 1 x 105 viable peripheral CD3+ T cells or 2.5 x 105 viable CD66B

TME cells were plated in complete media in a flat bottom 96-well plate (Corning, cat. no. 3596), in the presence or absence of 1 x 105
PBNs/TANs in a final volume of 100ul. After 96 h, the plate was spun at 300G for 10 min, and the supernatant collected and frozen at
-20C until further use. Supernatants were thawed at 4C and diluted 2x in Assay Diluent (MSD, cat. no K151AEL-1). A customized
cytokine panel assay containing Granzyme B, IFN-y and IL-12p70 (U-PLEX Immuno-Oncology Group 1 (hu) assay, MSD, Cat. no.
K151AEL-1) was performed according to the manufacturer’s instructions. Cytokine detection was performed using the MESO
QuickPlex SQ120 instrument (Meso Scale Discovery).Cell 186, 4546–4566.e1–e15, October 12, 2023 e13
ll
OPEN ACCESS ArticleAnnexin V apoptosis analysis
PBNs were cultured with whole tumor MEC at a ratio of 1:1 in complete media (DMEM-F12 + 10% FBS + 1% P/S) at 37C and 5%
CO2. Cells were collected after 24h or 48h and incubated in PBS with Human TruStain FcX Fc receptor blocking solution (FC block;
1:100) (BioLegend, cat. no. 422302) for 10min at RT. The 2x concentrated FCM antibodymix was added to the cells for 15min at 4C
in the dark (see key resources table for the antibodies used). Cells were washed with 1x binding buffer (Thermo Fischer Scientific, Cat
no. 88-102-72) and stained with a 1:100 dilution of Annexin V-PE (Thermo Fischer Scientific, 88-102-72) in binding buffer. The cells
were incubated for 15 min at RT in the dark, before washing in binding buffer. Cells were resuspended in binding buffer containing
2 ug/ml DAPI (Life Technologies, cat. no. D1306). Flow cytometric acquisition was performed within 4hrs after the staining.
UMAP and trajectory analysis
The manually gated PBN/TAN FCM population was exported as a separate FCS file. Variables with low variance were removed, data
was randomly down-sampled to 50’000 events, and transformed (asinh(x)) using the R package scDataviz.89 Dimensional reduction
with UMAP as well as clustering analysis were performed using the monocle3 R package (v1.0.0).91 Normalization was performed by
size factor using principal component analysis. Unsupervised clustering was performed using the Leiden community detection
method with the number of nearest neighbours (k) set to 100. Trajectory analysis was performed using the slingshot R package
(v2.2.0).92 Cluster 2 was selected as the origin. Multiple, disjoint trajectories were permitted by setting the omega parameter to TRUE.
Microenvironmental culture (MEC)-conditioned media (CM) collection and analysis
The generation and collection of whole-tumor microenvironmental culture-conditioned media (MEC-CM) from murine and patient
tumors was performed as described in Maas et al.23 (Module 3a). In brief, enzymatically dissociated tumor tissue was resuspended
in complete media at a concentration of 2 x 106 cells/ml in a tissue culture flask and incubated at 37
C, 5% CO2 and 95% humidity.
CM was collected after 24h, spun for 10 min at 300 g, and stored at -80 C until further use. For the generation of MEC-CM for ROS
detection, media without phenol red was used. The immune landscape of MECs was characterized by FCM at 0h, 24h and 48h after
the start of culture. Survival of immune populations over time was assessed by absolute viable cell counts using Trucount Absolute
Counting Tubes (BD Biosciences, cat. no. 340334) by FCM. For detailed clinical information please see Table S1C.
Conditioned media (CM) generation from cell lines and primary cells
Tumor cell lines were derived from brain tumor MEC by passaging the cells at least 3 times prior to submission for validation byWES.
Validated tumor cells were plated at 2 x 106 cells/ml in a tissue flask for the creation of CM. For detailed clinical information please see
Table S1C.
HBMEC (ScienCell, Cat. no 1000) were cultured in Endothelial Cell Medium (ScienCell, 1001) andHA (ScienCell, Cat. no 1800) were
cultured in Astrocyte Medium (ScienCell, 1801) in accordance with vendor’s recommendations. HUVECs were cultured in DMEM-
F12 (1:1) +GlutaMAX +10% FBS +1% P/S. Cells were passaged when 70% confluency was reached. 0.25% Trypsin was added
to the plates and trypsinization was stopped by adding Trypsin Neutralization Solution (ScienCell, 0113) according to manufacturer’s
instructions.
Differentiated MDMs from healthy donors were generated as described above and harvested from Teflon-coated bags at day 7,
followed by plating at a density of 1 3 106 cells/well of a 6-well plate in DMEM +10% FBS +1% P/S.
For all cell lines, media was changed to complete media (DMEM +10% FBS + 1% P/S) when 70% confluency was reached. The
supernatant of tumor cell lines (TCM) and non-tumor-associated primary cell lines (non-TCM) was collected after 24h, spun at 300G
for 10 min and stored at -80C until further use.
CM education experiments
MEC-CM, TCMand non-TCMwere thawed at RT. Isolated humanPBNs fromHDwere resuspended in CMand plated at 1 x 106 cells/
ml in U-bottom 96-well plates (Corning, Cat no. 3799). Isolated murine PBNs were resuspended in MEC-CM and plated at 5 x 105
cells/ml. HD PBN cultured in MEC-CM were treated with human TNF-a neutralizing antibody (10ug/ml, Cell Signaling Technologies,
cat. no. 7321S), GLO-1 inhibitor (10mM, MedChemExpress, cat. no. HY-136684), anti-IL-6 (10ug/ml, R&D, cat. no. MAB206), or anti-
IL-9 (10ug/ml, R&D, cat. no. AF209). Cells were incubated at 37C and 5% CO2, until they were collected for phenotypic character-
ization by FCM (after 5–6 h) and cell survival assessment at 24h and 48h. For the assessment of ROS detection by FCM, the readout
was performed immediately.
Whole exome sequencing (WES)
WES of snap-frozen and OCT-embedded frozen tumor tissue was performed as described in Maas et al.23 and A´lvarez-Prado et al.39
(Module 1b). Briefly, tumor tissue (minimum size 3 mm3) was placed on aluminum foil and snap-frozen by directly dipping it in pre-
cooled 2-methylbutane (Sigma-Aldrich, cat. no. 50970) placed in bucket containing liquid nitrogen. Frozen tissues were directly
transferred in a cryovial and stored at -80
C until further use. Tumor pieces (1-2 mm3) were placed into a 2 ml round-bottommicro-
centrifuge tube and 600 ul RLT buffer (Qiagen DNAeasy Blood&Tissue kit, cat. no. 69504) was added. The tissue was disrupted and
homogenized using an Omni Tissue Homogenizer with a 5-7 mm probe for 1-3 min. The lysate was incubated at 55 C in a heat block
for 1 h. For the validation of tumor cell lines byWES, a cell pellet of 1 x 106 tumor cells, as well asmatched PBMCs, were resuspendede14 Cell 186, 4546–4566.e1–e15, October 12, 2023
ll
OPEN ACCESSArticlein 600 ul RLT buffer and incubated at 55C in a heating block for 1h. Genomic DNA was extracted from frozen tumor tissue, PBMCs,
and tumor cells followingmanufacturer’s instructions (Qiagen DNAeasy Blood&Tissue kit, cat. no. 69504). Sequencing of paired end,
150 base pair (2 x 150), was performed by Genewiz (South Plainfield, New Jersey, USA) on an Illumina NovaSeq 6000 sequencer
(Illumina). Somatic variant calling was performed fromWES data of paired tumor/cell line and PBMC samples. Raw reads (fastq files)
were quality-control checked by FastQC and sequencing adaptors were removed by cutadapt (v2.3).93 The resulting reads were then
aligned to the hg38 (GRCh38.95) reference genome using the Burrows-Wheeler aligner (v0.7.17).94 Aligned reads were sorted and
written into BAM alignment files using SAMtools (v1.8)95 andmarked for duplicates using Picard tools (v2.9.0). Somatic mutation call-
ing was performed usingMuTect2 from the GenomeAnalysisTK-4.1.0.0 (GATK4) andGATKResource Bundle GRCh38 following best
practices for somatic variant calling as described by The Broad Institute (https://gatk.broadinstitute.org/hc/en-us/articles/
360035894731-Somatic-short-variant-discovery-SNVs-Indels-). Somatic variants were filtered to only retain confident calls by using
the GenomeAnalysisTK FilterMutectCalls function. VCF files containing only ‘‘PASS’’ (filtered) variants were generated by using
GenomeAnalysisTK SelectVariants (parameters: –exclude-filtered) and filtered variants were annotated using ENSEMBL Variant Ef-
fect Predictor (v96).96
Protein array
Frozen CM was submitted to RayBiotech Life and analyzed by the manufacturer’s in-house service using L1000 Glass Slide Human
Antibody Array (RayBiotech Life, cat. no. AAH-BLG-1000-4). The threshold for protein detection was set to a two-fold standard de-
viation of the negative control, and expression values were normalized to those from complete media (DMEM-F12 + 10% FBS +
1% P/S).
Recombinant protein treatment
Isolated HDPBNswere resuspended at a concentration of 1 x 106 cells/ml in complete media (DMEM+ 10%FBS + 1%P/S), and 100
ul of the cell suspension was placed in a U-bottom 96 well plate (Corning, Cat no. 3799). Cells were treated with Ceruloplasmin
(0.5uM, Sigma-Aldrich, cat. no. C4519-100UN), TNF-a (10 ng/ml, PeproTech, cat. no. 300-01A), GLO-1 (50mg/ml, R&D, cat. no.
4959-GL), IL-6 (5ng/ml, Sigma-Aldrich, cat. no. SRP3096), and IL-9 (5ng/ml, R&D, cat. no. 209-ILB). For the measurement of pheno-
typic alterations using FCM, cells were incubated for 5-6h at 37C and 5% CO2. For the measurement of ROS by FCM cells were
incubated for 30min at 37C and 5% CO2.
Software and visualization
FCM data was analyzed using FlowJo (v10.8.2). Graphs and plots were generated using either the ggplot2 (v3.3.6)90 R package or
GraphPad Prism (v9.4.1). Heatmaps were created using the ComplexHeatmap (v2.10.0)97 R package. The protein interaction
network was generated using the STRING platform (https://string-db.org)77 and visualized with iGraph (v1.2.6)98 R package.
QUANTIFICATION AND STATISTICAL ANALYSIS
All biostatistics analysis were performed within R (v4.1.1). Data was analyzed with the statistical test described in the corresponding
figure legend. The ‘n’ value and the use of median and standard deviation (SD) values are defined in the accompanying figure
legend. Mixed effect model calculations were performed with the R packages lme4 (v1.1-32)99 and lmerTest (v3.1-3).100
Statistical significancewas depicted as the p-value for single comparisons and p.adj value formultiple comparisons unless otherwise
stated: * <0.05, ** <0.01, *** <0.001, **** <0.0001, ns >0.05. P.adj values were obtained by correcting for multiple testing using the
Benjamini-Hochberg method, unless otherwise stated.Cell 186, 4546–4566.e1–e15, October 12, 2023 e15
Supplemental figures
(legend on next page)
ll
OPEN ACCESSArticle
Figure S1. Neutrophil abundance in blood and brain tumor microenvironment and their phenotypic alterations, related to Figure 1
(A) Mean proportions of all CD45+ immune cell populations in non-tumor (n = 11) and tumor tissue (nIDH mut = 31, nIDH WT = 70, nBreast-BrM = 15, nLung-BrM = 41,
nOther-BrM = 24) based on flow cytometry (FCM). Microglia (MG), monocyte-derived macrophages (MDMs), immature monocytes (iMCs), dendritic cells (DCs), T
regulatory cells (Tregs), double negative T cells (DNT cells).
(B) Mean proportions of CD45+ immune cells in whole blood from healthy donors (HD; n = 12) and brain tumor patients (nIDH mut = 20, nIDH WT = 43, nBreast-BrM = 8,
nLung-BrM = 25, nOther-BrM = 17) based on FCM.
(C) Cell-type identification of TANs (CD45+,CD15+), non-TAN immune cells (CD45+,CD15
), non-immune (CD45
) and vessels (CD31+) on tissues as shown in
Figure 1E. Dashed yellow line indicates the border of the perivascular niche (PVN), corresponding to a 20 mm distance surrounding the vessel.
(D) Fold-change median fluorescence intensity (MFI) with SD of the indicated markers in matched PBNs and TANs from tumor-bearing patients (nIDH mut > 10,
nIDH WT = 19, nBrM > 31) normalized to HD PBNs (n > 14). Wilcoxon signed-rank test of matched patient neutrophils only. p.adj values: *p < 0.05, **p < 0.01,
***p < 0.001, ****p < 0.0001.
(E) Representative IF images of immune cells (CD45+,CD15
), neutrophils (CD45+,CD15+), and vessels (CD31+) and their expression of S100A9 in IDHWT glioma
and lung-BrM tissue. Scale bars: 100 mm.
ll
OPEN ACCESS Article
(legend on next page)
ll
OPEN ACCESSArticle
Figure S2. Neutrophils in human brain (tumor) tissue differ substantially from PBNs, related to Figure 2
(A) Principal component (PC) plot of neutrophil transcriptional profiles in blood (empty circles) and brain (tumor) tissue (filled circles), based on 1% most vari-
able genes.
(B) Euler diagram showing the (absence of) overlap of up- and downregulated differentially expressed genes (DEGs) in glioma and BrM vs. HD PBNs from
Figure S2D (cutoff: p.adj < 0. 05; LFC > 1 or < 
1).
(C) Euler diagram depicting the intersect of up- and downregulated DEGs in BrM TANs and PBNs vs. HD PBNs (cutoff: p.adj < 0.05; LFC > 1 or < 
1).
(D) Over-representation analysis (ORA) of pathways from KEGG and Reactome databases on upregulated DEGs shared between BrM PBNs and TANs vs.
HD PBNs.
(E) Euler diagram showing the intersect of downregulated DEGs in BrM and glioma TANs, and non-tumor brain neutrophils vs. HD PBNs (cutoff:
p.adj < 0.05; LFC < 
1).
(F) Rank plot of gene set enrichment analysis (GSEA) using KEGG and Hallmark databases of non-tumor brain neutrophils vs. HD PBNs from (Figure 2C) (cutoff:
p.adj < 0.05; NES > 0).
(G) Euler diagrams depicting the intersect of BrM vs. glioma TANs and BrM TANs vs. matched PBNs (cutoff: p.adj < 0.05; LFC > 1 or < 
1).
(H) Euler diagrams depicting the intersect of glioma vs. BrM TANs and glioma TANs vs. matched PBNs (cutoff: p.adj < 0.05; LFC > 1 or < 
1).
(I) Top 10 significant GSEA pathways using Hallmark, KEGG, and Reactome databases on DEGs in lung- vs. breast-BrM TANs. (cutoff: p.adj < 0.05;

1.5 > NES > 1.5). Highlighted in light green are lung-BrM TAN specific pro-inflammatory pathways, highlighted in dark green are breast-BrM TAN specific cell-
cycle-associated pathways.
(J and K) (J) Heatmap depicting non-tumor neutrophils and brain TANs aligned to the top 5most variable genes per cluster as defined by the Salcher37 and (K) the
top 10 most-variable genes per cluster as defined by the Alvarez-Breckenridge single cell RNA-seq dataset.38 Genes of interest are indicated on the right.
Samples were hierarchically clustered.
(L) GSEA using Hallmark, KEGG and Reactome databases to analyze DEGs in TANs isolated from KRAS mut vs. WT and, TP53 mut vs. WT lung-BrM (cutoff:
p.adj < 0.05; 
1.5 > NES > 1.5). Pathways of interest are highlighted in green.
ll
OPEN ACCESS Article
Figure S3. ROS levels are altered in human BrM PBNs and brain-TAN, related to Figure 3
(A) Reactive oxygen species (ROS) detected in PBNs from HD (n = 18), and BrM-bearing patients (nBreast-BrM = 11, nLung-BrM = 24, nOther-BrM = 16). Wilcoxon rank-
sum test. p.adj value as indicated.
(B) GSEA based on Hallmark database in glioma and BrM PBNs vs. HD PBNs. ROS-related pathways are highlighted in dark gray (cutoff: p value < 0.05).
(C) Detected ROS inmatched PBNs and TANs, stratified by the absence ‘‘No’’ (nIDHmut = 5, nIDHWT = 5, nBrM = 1) or presence ‘‘Yes’’ (nIDHmut = 1, nIDHWT = 4, nBrM =
8) of previous oncologic treatment (e.g., immunotherapy, radiotherapy, chemotherapy, targeted therapy, hormone therapy, surgery, or corticosteroids). Wilcoxon
signed-rank test, p value: *p < 0.05, **p < 0.01.
(D) Neutrophil ROS production in a titrated coculture with CD66B
 TME population (nIDH mut = 6, nIDH WT = 9, nBrM = 9). Wilcoxon signed-rank test. p.adj value:
*p < 0.05, ***p < 0.001, ****p < 0.0001. Data in (A), (C), and (D) are represented as mean ± SD.
ll
OPEN ACCESSArticle
(legend on next page)
ll
OPEN ACCESS Article
Figure S4. Brain TANs do not have evident T cell stimulatory capacity and reside in close proximity to deformed vessels, related to Figure 4
(A) Proportion of cells expressing markers associated with antigen presentation in matched PBNs and TANs (nIDH WT = 4, nBrM = 2). Wilcoxon signed-rank test.
(B) Cytokine concentration (pg/mL) in supernatant of CD66B
 cells cocultured with either PBNs or TANs for 96 h (nIDH mut = 1, nIDH WT = 7, nBrM = 6). Wilcoxon
signed-rank test, p.adj values as shown.
(C) Representative IF images and cell type quantification of T cells and TANs in a lung-BrM, with cell type-specific identification of TANs (CD45+,CD15+), CD8+
T cells (CD45+,CD3+,CD4
,CD8+), CD4+ T cells (CD45+,CD3+,CD4+,CD8
,FOXP3
) and Tregs (CD45+,CD3+,CD4+,CD8
,FOXP3+).
(D) Representative IF images and cell type quantification of CD8+ T cells (CD45+,CD3+,CD8+), TANs (CD45+,CD15+), PD-L1 and PD-1 in a lung-BrM sample,
corresponding to the quantification shown in Figure 4E. Scale bars in (C) and (D): 20 mm.
(E) Distance of individual CD8+ T cells(n = 54,495) to TANs stratified by their respective expression of PD-1 and PD-L1 (nBrM = 20, nIDH WT = 7 analyzed together).
Wilcoxon rank-sum test.
(F) Representative lung-BrM IF images of vessels (CD31+); inserts on the lower left of each image show the form factor value of the highlighted vessel. Scale
bars: 50 mm.
(G) Form factor of individual vessels grouped by absence (Neg, nNon-tumor = 35,078, nIDH mut = 54,015, nIDH WT = 62,261, nBrM = 79,225) or presence (Pos,
nNon-tumor = 67, nIDHmut = 1,630, nIDHWT = 2,745, nBrM = 37,242) of neutrophils in the PVN from non-tumor (n = 5) and tumor tissue (nIDHmut = 12, nIDHWT = 15, nBrM =
27). Wilcoxon signed-rank test.
(H) Mean vessel area in mm2 stratified by the presence of neutrophils in the PVN on the same dataset shown in (G). Wilcoxon signed-rank test.
(I) Normalized log2-transformed gene expression of pro-angiogenic genes across the major TME populations.
(J) Correlation between mean vessel form factor vs. normalized gene counts of angiogenesis-associated genes in non-tumor, glioma, and BrM MG and MDMs
using the Pearsonmethod. p.adj values in (A), (E), (G), and (H): **p < 0.01, ***p < 0.001, ****p < 0.0001. Data in (A), (B), (E), and (G)–(I) are represented asmean ±SD.
ll
OPEN ACCESSArticle
(legend on next page)
ll
OPEN ACCESS Article
Figure S5. Phenotypic and lifespan alterations of TANs result from a combination of cell-intrinsic and microenvironment-mediated effects,
related to Figure 5
(A) Density plots of the age distribution of mouse neutrophils in Ntv-a;iLy6GtdTomato non-tumor brain and GBM as well as iLy6GtdTomato non-tumor bearing brain
and BrM mouse models over time.
(B) The proportion of tdTom+ neutrophils using iLy6GtdTomato reporter mice normalized to the maximum in primary tumor-bearing mice (mammary fat pad [MFP]
injection of the PyMT-BrM3 cell line) and BrM-bearing mice over an 8 day time course. A minimum of 3 mice from 6 independent experiments are shown per time
point. Mean lifetime in days is shown on the right.
(C) Proportion of tdTom+ neutrophils normalized to the maximum in a series of different non-brain tissues isolated from non-tumor (Ctrl) and BrM-bearing
iLy6GtdTomato mice over 8 consecutive days. A minimum of 3 mice are shown per time point from 6 individual experiments.
(D) 24 h survival of patient TANs (nIDH mut = 5, nIDH WT = 7, nBrM = 6), PBNs from HD (n = 10) and patient PBNs (nIDH mut = 5, nIDH WT = 8, nBrM = 9). Wilcoxon rank-
sum test.
(E) Proportion of live neutrophils after 24 h in whole tumor, PBNs and PBNs cocultured with CD66B
 TME stra PBNs and TANs, stratified by the absence ‘‘No’’
(nIDH mut = 4, nIDH WT = 3, nBrM = 1) or presence ‘‘Yes’’ (nIDH mut = 1, nIDH WT = 2, nBrM = 3) of previous oncologic treatment (e.g., immunotherapy, radiotherapy,
chemotherapy, targeted therapy, hormone therapy, surgery, or corticosteroids). Wilcoxon signed-rank test, p.adj values as shown.
(F) Proportion of live TANs in matched whole tumor, isolated TANs and in coculture with CD66B
 TME populations after 24 and 48 h. Wilcoxon signed-rank test.
(G) MFI of neutrophil markers in PBNs in whole blood vs. TANs over time and in the presence of the CD66B
 TME (nIDH mut = 4, nIDH WT = 5, nBrM = 4). Wilcoxon
signed-rank test.
(H) Detected ROS in murine PBNs (n = 7) cultured in regular media (Ctrl) vs. MEC-CM from mouse GBMs or BrMs. Wilcoxon signed-rank test.
(I) Proportion of livemurine PBNs after 24 h culture in regularmedia (Ctrl) vs. MEC-CM frommouseGBMs (n = 10) and BrM (n = 9).Wilcoxon signed-rank test. Data
in (B)–(I) are represented as mean ± SD. p.adj values in (D), (F), and (G): ns > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001. p value in (H) and (I): *p < 0.05, **p < 0.01.
ll
OPEN ACCESSArticle
(legend on next page)
ll
OPEN ACCESS Article
Figure S6. The brain TME is a rich source of soluble factors which alter neutrophil phenotype and lifespan, related to Figure 6
(A) Venn diagrams of somatic mutations identified by whole exome sequencing (WES) in matched BrM MEC and tumor cell lines (n = 2).
(B) Venn diagrams of somatic mutations identified by WES of IDH mut (n = 1) and IDH WT gliomas (n = 3) matched MEC and tumor cell lines.
(C) Fold-change in 24 h survival of HD PBNs (n = 25) cultured in conditioned media (CM) of tumor naive cell lines (non-TCM; nHBMEC, HUVEC, HA = 1, nMDM = 3) vs.
control medium and MEC-CM (nMEC = 6). Mixed-effect model using CM type as fixed effect and HD ID as random effect: ANOVA p value = 2.941 3 10
7.
(D) Fold-change MFI of the indicated proteins in HD PBN (n = 12) cultured for 6 h in CM (nMEC = 6, nHBMEC, HUVEC, HA = 1, nMDM = 3) vs. control medium and MEC-
CM. Mixed-effect model as in (C).
(E) Unsupervised heatmap of proteins detected in brain tumorMEC-CM, tumor cell-conditionedmedia (TCM), non-TCM, and control media. The sample outlier is
breast-BrM LAU_BrM_0092, which was not characterized by any evident difference compared with other samples based on known clinical parameters, see also
Table S1C.
(F) Visualization of differentially detected proteins in glioma and BrM MEC-CM vs. TCM and non-TCM. The unique nproteins found in individual groups and in-
tersects are indicated on top of the bars (cutoff: p < 0.05).
(G) GSEA of TNF-a signaling pathways in glioma and BrM TANs compared with matched PBNs.
(H) Fold-change MFI of indicated markers measured in HD PBNs (n = 5) cultured in MEC-CM or treated with MEC-CM specific soluble factors. Wilcoxon signed-
rank test. Data in (C), (D), and (H) are represented as mean ± SD, p.adj values: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
ll
OPEN ACCESSArticle
Figure S7. Survival of MEC populations in ex vivo culture, related to Figure 7
Normalized survival in% (compared with 0 h, mean ± SD) of microenvironmental culture (MEC) populations over time in culture (nIDH mut = 1, nIDHWT = 3, nBrM = 3).
ll
OPEN ACCESS Article