Article

Remodeling of the immune and stromal cell

compartment by PD-1 blockade in mismatch repair-
deficient colorectal cancer
Graphical abstract Authors
Jianxia Li, Cheng Wu, Huabin Hu, ...,
Yaoxu Chen, Zhi Xie, Yanhong Deng

Correspondence
[email protected]

In brief
Li et al. use scRNA-seq and multiplex
immunohistochemistry to investigate the
cellular dynamics in patients with d-
MMR/MSI-H CRC treated with
neoadjuvant toripalimab, revealing
immune and stromal features of complete
responders and non-complete
responders and highlighting the role of
inflammatory conditions in regulating the
balance between immune clearance and
immune escape.

Highlights
d Immune and stromal cell alterations occur in d-MMR/MSI-H
CRC following PD-1 blockade

d Changes in the CD8+ T cell cytotoxic and proliferation

programs are response-related

d ICI decreases CD4+ Tregs and increases CD40+ B cells in

complete responsive tumors

d Resolution of tumor-promoting inflammation correlates with

ICI response

Li et al., 2023, Cancer Cell 41, 1–18

June 12, 2023 ª 2023 Elsevier Inc.
https://doi.org/10.1016/j.ccell.2023.04.011 ll
 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll

Article
Remodeling of the immune and stromal cell
compartment by PD-1 blockade in mismatch
repair-deficient colorectal cancer
Jianxia Li,1,3,6 Cheng Wu,4,6 Huabin Hu,1,3,6 Ge Qin,1,3,6 Xueqian Wu,1,3 Fan Bai,1,3 Jianwei Zhang,1,3 Yue Cai,1,3
Yan Huang,2,3 Chao Wang,2,3 Jiaqi Yang,4 Yizhao Luan,4 Zehang Jiang,4 Jiayu Ling,1,3 Zehua Wu,1,3 Yaoxu Chen,5 Zhi Xie,4
and Yanhong Deng1,3,7,*
1Department of Medical Oncology, Department of General Surgery, The Sixth Affiliated Hospital, Sun Yat-sen University, Guangzhou
510655, China
2Department of Pathology, Department of General Surgery, The Sixth Affiliated Hospital, Sun Yat-sen University, Guangzhou 510655, China
3Guangdong Provincial Key Laboratory of Colorectal and Pelvic Floor Disease, The Sixth Affiliated Hospital, Sun Yat-sen University,

Guangzhou 510655, China

4State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangzhou 510627, China
5Medical Affairs, 3D Medicines Inc., Shanghai 201114, China
6These authors contributed equally
7Lead contact

*Correspondence: [email protected]
https://doi.org/10.1016/j.ccell.2023.04.011

SUMMARY

Immune checkpoint inhibitor (ICI) therapy can induce complete responses in mismatch repair-deficient and
microsatellite instability-high (d-MMR/MSI-H) colorectal cancers (CRCs). However, the underlying mecha-
nism for pathological complete response (pCR) to immunotherapy has not been completely understood.
We utilize single-cell RNA sequencing (scRNA-seq) to investigate the dynamics of immune and stromal cells
in 19 patients with d-MMR/MSI-H CRC who received neoadjuvant PD-1 blockade. We found that in tumors
with pCR, there is a concerted decrease in CD8+ Trm-mitotic, CD4+ Tregs, proinflammatory IL1B+ Mono
and CCL2+ Fibroblast following treatment, while the proportions of CD8+ Tem, CD4+ Th, CD20+ B, and
HLA-DRA+ Endothelial cells increase. Proinflammatory features in the tumor microenvironment mediate
the persistence of residual tumors by modulating CD8+ T cells and other response-associated immune
cell populations. Our study provides valuable resources and biological insights into the mechanism of suc-
cessful ICI therapy and potential targets for improving treatment efficacy.

INTRODUCTION lying mechanisms of complete response of ICI treatment has not

been completely understood, which has hindered the improve-
Mismatch repair-deficient and microsatellite instability-high (d- ment of current therapeutic efficacy and regimens for d-MMR/
MMR/MSI-H) represent a distinct biomarker-defined population MSI-H CRC. Emerging evidence from gene expression signa-
of cancers accounting for approximately 15% of all colorectal tures using bulk RNA sequencing (RNA-seq) and multiple
cancers (CRCs).1 Of all CRCs, d-MMR/MSI-H CRC is associated immunofluorescence suggests that the efficacy of ICIs largely
with a higher tumor neoantigen load and denser immune cell infil- depends on the tumor immune microenvironment (TIME). It is re-
tration than mismatch repair-proficient and microsatellite-stable ported that the response to PD-1 blockade in d-MMR/MSI-H
(pMMR/MSS) tumors,2,3 suggesting benefit from immune CRC is not associated with tumor mutation burden but with
checkpoint inhibitor (ICI) therapy. ICI treatment targeting the high clonally expanded T cells.11 In the NICHE study, d-MMR
PD-1/PD-L1 pathway is highly effective for first-line treatment colorectal tumors showed significant increases in CD3+ and
in patients with metastatic d-MMR/MSI-H CRC4–7 and for CD8+ T cell infiltration as well as interferon (IFN)-g scores after
neoadjuvant treatment in patients with nonmetastatic d-MMR/ ICI treatment.8 However, these studies mainly rely on profiling
MSI-H CRC.8–10 technologies that measure tumors in bulk, which limited their
However, challenges remain associated with ICI treatment ability to capture cell heterogeneity and explore changes in im-
that need to be addressed in order to improve the efficacy of cur- mune components or cellular interactions comprehensively.
rent treatment regimens, since up to 50% of patients with meta- Recent advances in single-cell RNA-seq (scRNA-seq) have pro-
static d-MMR/MSI-H CRC are resistant to immunotherapy with vided an avenue to explore cell type-specific patterns and inter-
subsequent progression and disease recurrence.5–7 The under- actions in the response to ICI treatment at a cellular resolution.

Cancer Cell 41, 1–18, June 12, 2023 ª 2023 Elsevier Inc. 1
 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Single-cell analyses of the TIME in solid tumors, such as breast separately (STAR Methods) and revealed heterogeneous
cancer, melanoma, following immunotherapy have been immune and stromal cell subtypes that exhibited distinct
recently published,12–15 but it is not clear how such findings molecular signatures indicative of their unique cellular identities
can be applied to the study of immunotherapy in CRC. Particu- (Figures 1E, S2C–S2G).
larly, single-cell analysis before and after immunotherapy in
CRC have not yet been reported. ICI-related changes in the CD8 T cell cytotoxicity and
Here, we present the first in-depth cellular and molecular anal- proliferation program
ysis of cell populations in patients with d-MMR/MSI-H CRC It is well known that d-MMR/MSI-H tumors demonstrate high
treated with PD-1 blockade (toripalimab) or toripalimab with infiltration of cytotoxic CD8+ T cells and upregulated expression
the COX-2 inhibitor (celecoxib) from a prospective cohort.9 The of multiple immune checkpoint molecules, which is also the
study aims to uncover the underpinnings of resistance and reason for their sensitivity to immunotherapy.3 We therefore first
sensitivity to ICI treatment and to suggest the potential therapeu- focused on the CD8+ T clusters, which further categorized into
tic targets of ICI treatment for residual tumor patients. We intraepithelial lymphocyte (IEL), mucosal-associated invariant T
compare the cell type distributions and functional changes (MAIT), effector memory (Tem), and two tissue resident memory
before and after ICI exposure in patients who achieved patholog- (TRMs) subsets (Figures 2A and 2B; see Table S2 for marker
ical complete response (pCR) to elucidate the mechanism genes). We first used gene set enrichment analysis (GSEA) to
related to successful ICI therapy (
ICI vs. +ICI/pCR). We also explore the effect of ICI on overall CD8+ T cell transcriptome. It
analyze the differences in treatment dynamics between pCR was found that the trends on the distribution of cytotoxic, ex-
and non-pCR response to explore the mechanism of resistance hausted, and proliferated gene signature were correlated with
to immunotherapy. ICI treatment; the cytotoxic signature was enriched in the +ICI/
pCR group compared with the
ICI group, while the exhausted
RESULTS and proliferated signatures were enriched in the +ICI/non-pCR
group compared with the
ICI group (Figure 2F).
Single-cell expression atlas of neoadjuvant We next analyzed the expression of cytotoxic, exhausted, and
immunotherapy-treated d-MMR/MSI-H CRC proliferated genes in CD8+ T clusters to explore the key cell type
The patient cohorts were from a randomized phase 2 study associated with response to ICI. Among these CD8+ T clusters,
(NCT03926338) with locally advanced primary invasive d- the CD8+ Tem (CD8-C3) and CD8+ Trm (CD8-C4) subsets ex-
MMR/MSI-H carcinoma of the colon and rectum treated with tor- pressed granzymes (GZMB, GZMK, GZMA), perforin 1 (PRF1),
ipalimab with or without celecoxib for six cycles before curative IFN-g (IFNG), and Fas Ligand (FASLG), and were enriched in
surgical resection (PICC study).9 We performed scRNA-seq on the CD8 cytotoxic gene signature,16,17 which was suggestive
40 samples of tumor and adjacent normal tissues from 19 pa- of effector functions (Figures 2C and 2E). CD8+ Trm (CD8-C4)
tients to characterize cellular and molecular of immune and stro- and CD8+ Trm-mitotic (CD8-C5) cells expressed markers of co-
mal cells and dynamics during ICI treatment (Figures 1A and inhibitory receptors (PDCD1, LAG3, TIGIT, HAVCR2) and
S1B; Table S1). Most of the patients (15 of 19) achieved pCR exhaustion (TOX, ENTPD1, BATF, PRDM1), and were enriched
with no residual tumor after neoadjuvant ICI treatment in the CD8 exhaustion signature (Figures 2C and 2E). Further
(Figures S1A and S1B; Table S1). In total, we obtained a high- analysis found that CD8+ Tem cells showed profound similarities
quality single-cell transcriptome atlas for a total of 155,397 cells to previously published CD8+ T effector memory cells18,19
from 40 samples (Figures S1C and S1E; STAR Methods). Inte- (Figures 2E and S3A), while CD8+ Trm (CD8-C4) and CD8+
gration of all the samples for cluster analysis revealed clusters Trm-mitotic (CD8-C5) cells showed features consistent with hu-
representing six broad cell types, including T/I/NK cells (T manTRMs by expressing significantly higher levels of TRMs
cells/innate lymphocytes/NK cells), B cells, myeloid cells, markers (ENTPD1 [CD39], ITGAE) (Figures 2C and 2G) and en-
epithelial cells, endothelial cells, and fibroblasts (Figure 1B), riched TRMs signature19–21 (Figures 2E and S3A). These results
which expressed known markers accordingly (Figures S1D and are consistent with conclusions derived from previous studies
S1F). In tumor and normal samples, immune cells (T/I/NK cells, that high PD-1 expression is a tissue-residency feature of tumor
B cells, and myeloid cells) accounted for a large proportion of to- TRM cells.19,21,22
tal cells (29.70% on average in the
ICI group and 27.60% on Having identified candidate subsets of CD8+ T cells within d-
average in the +ICI/N group) (Figures 1D and S2A). Massive im- MMR/MSI-H CRC that bore some similarities to the previously
mune cell infiltration was confirmed by multiplex fluorescence described cytotoxic and exhausted population, we further
immunohistochemical staining (mIHC) (Figure S2B), which was analyzed changes in frequencies of CD8+ T clusters to address
consistent with previous findings in d-MMR/MSI-H CRC.3 whether tumor regression was mainly driven by CD8+ T cells,
By comparing pretreatment and posttreatment samples, we as shown in other cancer types.23–25 We observed that ICI treat-
observed that the frequencies of T/I/NK cells, fibroblasts, and ment induced significant changes in the proportions of CD8+
endothelial cells significantly increased in the +ICI/pCR group Tem cells (p = 0.006, Wilcoxon test) and CD8+ Trm-mitotic cells
(p = 0.030 for T/I/NK cells, p = 0.049 for fibroblasts, p = 0.004 (p < 0.001, Wilcoxon test) in pCR response, while the proportion
for endothelial cells, Wilcoxon test) but not in the +ICI/non- of CD8+ Trm cells remained relatively unchanged (p = 0.648, Wil-
pCR group (p = 0.635 for T/I/NK cells, p = 0.188 for fibroblasts, coxon test, Figure 2D). A similar trend was observed in both the
p = 0.945 for endothelial cells, Wilcoxon test) (Figure 1C), sug- anti-PD-1-only regimen- and combination regimen-treated
gesting that T/I/NK cells and stromal cells play important roles groups compared with the
ICI group (Figure S3B). The distribu-
in ICI treatment. We further reclustered the broad cell types tions of these subsets and their association with pCR suggest

2 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Figure 1. Global analysis of immune and stromal cell populations in d-MMR/MSI-H CRC patients receiving ICI treatment
(A) Overview of the study design.
(B) Uniform Manifold Approximation and Projection (UMAP) plot of broad cell types from all samples (n = 40).
(C) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
(D) Cell cluster frequency shown as a fraction of total cells for each sample.
(E) UMAP plots of T/I/NK cells, B cells, myeloid cells, endothelial cells, and fibroblasts showing transcriptionally distinct clusters.
Also see Figures S1 and S2.

Cancer Cell 41, 1–18, June 12, 2023 3

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

(legend on next page)

4 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

the important role of CD8+ T cell cytotoxicity and proliferation For transient CD8+ Trm cells, antitumor immunity programs
programs in the neoadjuvant ICI response. were upregulated after ICI treatment, including programs for
the antigen receptor-mediated signaling pathway and lympho-
cyte-mediated immunity (Figure 3C). In addition, we found that
Shared characteristics between ICI-related changes
GZMK, GZMM, and IFN-g response genes (HLA-DQA2, CD74)
and CD8+ Trm transcriptomic alterations
were upregulated, while CD8+ Trm marker genes (CXCL13,
We next sought to address the process of fluctuation in CD8+
GZMB, GNLY) and immune inhibited receptors (HAVCR2,
subset proportions. For pCR tumors, ICI not only increased the
TNFRSF18) were downregulated (Figure 3A; Table S3). Overall,
proportions of CD8+ Tem cells, but also increased antigen pre-
the CD8+ Tem signature of CD8+ Trm cells increased and the
sentation processes and characteristic IFN-g response gene
CD8+ Trm-mitotic signature decreased after ICI treatment (Fig-
(HLA-DQA1, HLA-DRA, and CD74) expression in CD8+ Tem cells
ure 3D), which indicates that ICI may promote the transformation
(Figure 3A; Table S3) on a per-cell basis. This was consistent with
of CD8+ Trm cells to CD8+ Tem cells and block the transforma-
previous clinical observations that PD-1 blockade increased pre-
tion of a portion of CD8+ Trm cells into CD8+ Trm-mitotic cells,
cursor exhausted and effector memory CD8+ T cells in human
resulting in an increase in CD8+ Tem cells and a decrease in
cancer23,26 and in mouse models.27,28 The expression of HLA-
CD8+ Trm-mitotic cells and eventually resulting in the enhance-
DR genes, which indicate activated CD8+ T cell status, was
ment of overall antitumor immunity.
increased mainly in CD8+ Tem cells after treatment (Figures 3B
and S3E), and the activated HLA-DR+ CD8+ cells were observed
to be improved after treatment via mIHC (Figure S3D). In CD4+ T helper cells and CD20+ B cells concertedly
contrast, CD8+ Trm-mitotic cells marked by proliferation marker expanded in tumors completely responsive to the ICI
gene (MKI67, TOP2A, STMN1) dramatically decreased in relative treatment
frequencies (Figure 2D) after treatment. The proportion of Among CD4+ T clusters (Figures 4A and 4B), we found a signifi-
MKI67+ CD8+ cells was confirmed to be significantly decreased cant decrease in CD4 regulatory cell (Tregs) fraction in pCR sam-
after treatment in pCR response via mIHC (Figure S3D). Although ples after ICI treatment (p = 0.002, Wilcoxon test, Figure 4D),
both CD8+ Trm-mitotic and CD8+ Trm display CD8 exhaustion which was confirmed by mIHC (Figure 4C). This is consistent
features and enriched in exhausted signature, only CD8+ Trm- with a previous study on lung cancer showing that anti-PD-1
mitotic cells displayed mitotic features with high expression of decreases Tregs in responsive tumors.23 Intratumoral Tregs
genes associated with proliferation (Figure 2C), aligning with express PD-1(PDCD1) and CTLA-4(CTLA4), which can suppress
the previously defined CD8+ TRM-like cells actively undergoing antitumor immunity.29,30 We also found that Treg cells expressed
cell division in breast cancer.19 These results indicate that ICI several immune suppressive receptors (IRs), including
mainly caused the reduction of proliferated subsets of TRMs, TNFRSF18, LAG3, HAVCR2(TIM3), CTLA4, and TIGIT
not all TRMs. (Figures S4B and S4C). Consistently, the fraction of IR-ex-
By calculating the cytotoxicity and exhaustion scores of CD8+ pressed CD4+ Tregs decreased after ICI treatment in the pCR
T clusters (STAR Methods; Table S3), we found that CD8+ Trm group but was more abundant in the non-pCR group (Figure S4D;
cells had higher cytotoxicity and exhaustion scores than CD8+ Table S4). Together, these results indicate that successful ICI
Tem cells but lower exhaustion scores than CD8+ Trm-mitotic treatment reduces the CD4+ T cell immunosuppressive program
cells (Figure S3C). Previous study indicated that pre-exhausted by decreasing Tregs.
and exhausted CD8+ T cells form a continuum of cell states In contrast to Tregs, CD4+ T helper cells (CD4+ Th) marked by
with different intensities in proliferating, exhausted, and cyto- high CD40LG and IL7R gene expression were significantly
toxic programs.16 Together with the evidence that the CD8+ increased in the pCR group after ICI treatment (p = 0.008, Wil-
Trm cells shared cytotoxic characteristics with CD8+ Tem cells coxon test, Figure 4D). Previous studies have suggested that
and exhausted characteristics with CD8+ Trm-mitotic, CD8+ CD4 helps improve the intrinsic capacity of cytotoxic T lympho-
Tem, CD8+ Trm, and CD8+ Trm-mitotic cells indicated the exis- cytes (CTLs) to kill target cells and that helpless CTLs express
tence of a pool with different degrees of CD8+ T cell differentia- high levels of coinhibitory receptors.31 In this study, CD40LG
tion in the tumor environment. Trajectory analysis confirmed the was mainly expressed in CD4+ T helper subsets, including follic-
potential connections among these three CD8 subsets ular helper (Tfh), Th1-like, and Th subsets (Figure S4E). We there-
(Figure 3E). fore explored the impact of the CD40-CD40LG-helper signature

Figure 2. ICI-related changes in the CD8 T cell cytotoxic and proliferation program
(A) UMAP plot of CD8+ T cell clusters.
(B) Dot plots showing marker genes across CD8+ T cell clusters. Dot size indicates fraction of expressing cells, colored according to Z score normalized
expression levels. See Table S2 for all marker genes.
(C) Heatmap of scaled normalized expression for genes in CD8+ T cell clusters.
(D) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
(E) Dot plot of normalized enrichment scores (NES) for gene sets significantly enriched in each CD8+ T subset compared with other CD8+ T cells. The p values
were determined by fast preranked gene set enrichment analysis (GSEA).
(F) GSEA enrichment for gene sets in all CD8+ T cells comparing pretreatment (
ICI) vs. ICI-treated pCR (+ICI/pCR) and
ICI vs. ICI-treated non-pCR (+ICI/non-
pCR). NES, normalize enrichment score. The p values were determined by fast preranked GSEA.
(G) UMAP plots showing the expression of selected marker genes in CD8+ T cells.
Also see Figure S3.

Cancer Cell 41, 1–18, June 12, 2023 5

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Figure 3. Shared characteristics between ICI-related changes and the CD8+ Trm transcriptomic alterations
(A) Differentially expressed genes (DEGs) between the
ICI vs. +ICI/pCR groups among CD8+ T clusters. The x axis and y axis values were calculated by the
Seurat method. Red points indicate significantly upregulated genes in +ICI/pCR, and blue points indicate significantly downregulated genes in the +ICI/pCR
group (STAR Methods).
(B) Venn diagram showing the intersecting and unique DEGs between the
ICI and +ICI/pCR groups in CD8-C3, CD8-C4, and CD8-C5 cells. See Table S3 for
all DEGs.
(C) Pathways enriched in CD8+ Trm cells in
ICI and +ICI/pCR groups. Hypergeometric test. Benjamini-Hochberg (BH) adjusted p value <0.05. Odds ratio = Gene
Ratio/Background Ratio.
(D) Signature scores computed for the CD8-C4 (CD8+ Trm) cells and compared between patient groups with a two-sided Wilcoxon test. Signature scores were
defined by the average expression of CD8-C3-specific and CD8-C5-specific marker genes (fold change >1.5, adjusted p < 0.05, percent expr. >30%, see
Table S2 for marker genes).
(E) Pseudotime-ordered analysis of CD8+ T cell clusters from all samples. CD8+ T cell subsets are labeled by color.
Also see Figure S3.

on ICI-induced tumor regression. Overall, CD40LG-positive ety of immune and nonimmune cells as a part of humoral and
CD4 cells (CD40LG+ CD4+ Th) were found to be significantly cell-mediated immunity.32,33 In this study, CD40 was found to
increased in the pCR group after treatment (Figure S4F; be expressed in both myeloid cells and B cells (Figure S4H).
Table S4). Correlation analysis found that CD40LG+ CD4+ Th Clustering of B cells showed clear separation of CD20
cells were negatively correlated with CD8+ Trm-mitotic cells (Fig- (MS4A1)+ B cells and plasma B cells, with CD40 mainly ex-
ure 4H). These results suggest that an effective helper signature pressed in CD20+ B clusters (Figures S2E and 4E). By comparing
expressed by CD40LG+CD4+ T helper cells may impair the CD8+ the
ICI and +ICI/pCR groups, we found that fraction of CD20+ B
T cell exhaustion program. cells were significantly increased (Figure 4F), in line with their
The interaction of CD40 present on antigen-presenting cells expression of CD40. Increased CD40+CD20+ B cells after ICI
with CD40LG is an essential stimulus for the activation of a vari- treatment were validated by mIHC (Figure S4N). Using The

6 Cancer Cell 41, 1–18, June 12, 2023

Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

(legend on next page)

Cancer Cell 41, 1–18, June 12, 2023 7

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Cancer Genome Atlas (TCGA) CRC cohort data, we found that interactions of CD40-CD40LG between Bgc and CD4+ Th cells
only CD20+ B cells were significantly enriched in MSI samples, were shown in the +ICI/pCR group (Figure 4J). These results indi-
while plasma B cells showed no group enrichment (Figure S4J). cate that interaction between CD4+ Th and Bgc cells may attribute
These data suggest that B cells may participate in TIME remod- to antitumor immunity during ICI treatment.
eling caused by PD-1 blockade independent of B cell antibody-
driven immunity. By analyzing the transcriptome differences of B Decreased proinflammatory program in myeloid cells
cells before and after treatment, we found that the +ICI/pCR after ICI treatment
group exhibited enrichment for genes related to the antigen re- Furthermore, re-clustering of 8,469 myeloid cells revealed 10 clus-
ceptor-mediated signaling pathway and the T cell activation ters including macrophages, dendritic cells, and monocytes
pathway (Figure 4I). Compared with tumors with low abundance (Figures 5A and S5A). By comparing the +ICI/pCR and
ICI groups
of CD20+ B cells, tumors containing high abundance of CD20+ B of macrophages and dendritic cells, we found that upregulated
cells had higher CD8+ Tem and CD8+ Trm cell abundance and genes were associated with phagocytosis, endocytosis, antigen
lower CD8+ Trm-mitotic cell abundance (Figure S4G), suggest- receptor-mediated signaling pathway, and T cell activation (Fig-
ing that CD20+ B cells are related to CD8+ T cell status, which ure S5F). Notably, we identified two mono-like clusters (IL1B+
may contribute to the state transition toward pCR. Mono, FCN1+ Mono) based on their high expression of monocyte
CD20+ B cells were composed of heterogeneous subsets of markers (FCN1, S100A8, S100A12), low expression of both
naive B (Bn), memory B (Bmem), follicular B (Bfoc), and germinal macrophage markers (CD68, CD163), and HLA genes
center (Bgc) subsets with distinct transcriptional profiles (Figures 5B and S5A). A previous study revealed the precursor
(Figures 1E and S2D). In these CD40-expressing CD20+ B subsets, characteristics of FCN1 monocytes.38 However, the function of
the most striking increase was observed in Bgc cells (p = 0.004, these monocytes and their involvement with PD-1 blockade-medi-
Wilcoxon test) (Figure 4G). Bgc cells are characterized by high ated modulation of the TIME have not yet been described. We
expression of genes important for the geminal-center dark zone further analyzed the remodeling of myeloid cell subsets after
(CXCR4), pronounced germinal center reaction phenotype ICI treatment and found that the proportion of IL1B+ Mono cells
(CD69), and geminal-center selection (NR4A1, NR4A2) (Fig- significantly decreased (p < 0.001, Wilcoxon test, Figure 5C).
ure S2D; Table S2), displaying the characteristics of mature tertiary IL1B+CD14+CD68
mono-like cells were confirmed to be infil-
lymphoid structures (TLS).34–36 A study on a preclinical glioma trated in d-MMR/MSI-H CRC tumors, with VEGFA cytokines scat-
model has demonstrated that systemic delivery of CD40 induces tered around the cells (Figure 5E). By comparing these cells with
the formation of TLS near meningeal tissue and that the presence other myeloid cells, we observed that IL1B+ Mono cells showed
of TLS is associated with increased T cell infiltration.37 We also the characteristics of inflammatory monocytes by highly express-
found that a mature TLS signature was highly correlated with ing the proinflammatory factors IL6, IL1A, IL and the neutrophil
CD40-CD40LG expression (Figure S4M), but only the mature chemokines CXCL8, CXCL2, CXCL1, CCL2, and CCL4 (Figure 5D).
TLS signature rather than CD40-CD40LG expression was signifi- Inflammation is known to be correlated with cancer development
cantly enriched in MSI tumors (Figure S4L). This connection sug- and its response to therapy.39 GSEA showed that differentially ex-
gests that CD40-CD40LG may contribute to antitumor immunity pressed genes of IL1B+ Mono cells were significantly enriched in
by promoting TLS maturation. Next, we studied the relationship the positive regulation of the epithelial-mesenchymal transforma-
between Bgc cells and other immune cells. Bgc cells showed a sig- tion pathway, suggesting the tumor-promoting function of IL1B+
nificant correlation with CD8+ T cells, especially CD8+ Trm-mitotic Mono cells (Figure S5E).
cells (Figures 4H and S4K). Cellchat analysis revealed increased IL1B+ Mono cells remaining after therapy displayed decreased
number and strength of communications between CD4+ Th and proinflammatory factors than the pretreatment population,
Bgc in +ICI/pCR compared with
ICI group (Figure 4K). Stronger including IL1B and CXCL8 (Figure S5C), suggesting their

Figure 4. CD4+ Th and CD20+ B cells concertedly expanded after ICI treatment
(A) UMAP plot of CD4+ T cell clusters.
(B) Dot plots showing marker genes across CD4+ T cell clusters. Dot size indicates fraction of expressing cells, colored according to Z score normalized
expression levels. See Table S2 for all marker genes.
(C) Representative images of mIHC staining indicating CD4+FOXP3+ cells, in
ICI and +ICI/pCR samples. Each dot in the boxplot presents the fraction of cells in
each sample based on IHC staining image. Two-sided Wilcoxon test.
(D) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
(E) UMAP plots of B cell clusters showing expression of CD40.
(F and G) Quantification of the cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group.
Two-sided Wilcoxon test.
(H) Correlations of immune cell subsets with CD8+ Trm-mitotic in cellular proportions in d-MMR/MSI-H CRCs, n = 40. PCC, Pearson correlation coefficient.
*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
(I) Pathways enriched in B cells in
ICI and +ICI/pCR groups. Hypergeometric test. Benjamini-Hochberg (BH) adjusted p value <0.05. Odds ratio = Gene Ratio/
Background Ratio.
(J) Receptor–ligand pairs that differ significantly between
ICI and +ICI/pCR based on CD4+ Th and other cell clusters. Dot size indicates p value, colored
according to the communication probability of pathways.
(K) Differences in the number and strength of communications of various cell types between +ICI/pCR and
ICI group. The blue line indicates decreased
communications in +ICI/pCR, and the red line indicates increased communications in +ICI/pCR. The thicker the line, the greater the difference.
Also see Figure S4.

8 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

A B

C D

Figure 5. Decreased proinflammatory program in myeloid cells after ICI treatment

(A) UMAP plot of myeloid cells and cell cluster frequency shown as a fraction of total myeloid cells for each group.
(B) Heatmap of scaled normalized expression for HLA-I and HLA-II genes in myeloid cell clusters.
(C) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
(D) Heatmap of scaled normalized expression for proinflammatory genes in myeloid cell clusters.
(E) Representative images of mIHC staining indicating CD68
IL1B+CD14+ cells.
Also see Figure S5.

reduced proinflammatory ability on a per-cell basis. In addition, we next sought to explore the dynamics of stromal cell clusters
we also found significant decreases in serum proinflammatory during ICI treatment. Re-clustering of stromal cells identified
factors after ICI treatment, including IL1A, IL1B, IL6, CCL20, three fibroblast subsets and six endothelial cell subsets
and CXCL2 (Figure S5D), which further confirmed that ICI in d- (Figures 1E, S2F, and S2G; see Marker genes in Table S2).
MMR/MSI-H CRC modulates the inflammatory program. As reported in previous study, fibroblasts (Fibro) expressed
markers that are colon tissue-specific.44 CXCL12+ Fibro ex-
Remodeling of stromal cells by ICI treatment pressed markers (DCN, SLIT2, CXCL12) of mesenchymal cells
Based on changes in endothelial and fibroblast proportion be- distributed throughout the lamina propria, while CCL2+ Fibro
tween
ICI and +ICI/pCR (Figure 1C) and increased literature expressed markers (F3, WNT5A, BMP2, POSTN, HSD17B2)
highlighting the importance of the cross talk between immune of a smaller sub-population in close proximity to the epithelial
and stromal cells in determining TIME and ICI response,40–43 monolayer (Figures S2G and S6A; Table S2). Myofibroblast

Cancer Cell 41, 1–18, June 12, 2023 9

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

A B

C D

F
E

Figure 6. Remodeling of stromal cells by ICI treatment

(A) Dot plots showing selected genes across fibroblast clusters. Dot size indicates fraction of expressing cells, colored according to Z score normalized
expression levels.
(legend continued on next page)
10 Cancer Cell 41, 1–18, June 12, 2023
 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

and CXCL12+ Fibro increased significantly in pCR samples af- and associated with tumorigenesis.57 In summary, these results
ter ICI treatment (Figure 6B). CXCL12+ Fibro enriched for elastic highlight the important roles of stromal cells in lymphocyte
fibers (FBLN1, FBLN2, FBLN5, EFEMP1), while myofibroblast chemotaxis and tumor-promoting inflammation regression dur-
showed specific expression of sheet collagens (COL4A1, ing immune remodeling by ICI.
COL4A2) that are key constituents of the epithelial basement
membrane (Figure 6A). In line with increasing CXCL12+ Fibro, Resolution of pro-tumor inflammation correlated with
expressing of CXCR4 was found upregulated in CD4+ T, response
CD8+ T, and B cells after treatment (Figures 6C and S6D). Cell- Inflammation has been recognized as being closely involved in
chat analysis revealed that the interactions between CXCL12 the development and progression of malignancies via driving
expressed by fibroblast and CXCR4 expressed by lymphocyte chronic non-resolving inflammation.58–60 ICI induced significant
populations were enriched specifically in +ICI/pCR (Figures 6F decrease of proinflammatory myeloid subsets (IL1B+ Mono)
and S6E), indicating CXCL12+ Fibro may participate in lympho- and proinflammatory stromal subsets (CCL2+ Fibro) (Figures 7A
cyte recruitment via the CXC12-CXCR4 axis. Although fibro- and 7D), suggesting a resolution of inflammation.
blasts were reported to adopt the CXCL12-CXCR4 axis to CD8 TRMs were reported to attract neutrophils and mono-
restrain T cells in pancreatic cancer,45 it has also been cytes by secreting CCL3 and CCL4 to trigger innate and adaptive
shown that CXCL12-producing fibroblasts amass lymphocyte immune responses.61,62 In addition, inflammatory stimulation
migration.46,47 promotes the maturation of TRMs.63 We observed that CD8
Among endothelial cells (EC), HLA-DRA+ EC increased signif- TRMs clusters including CD8+ Trm and CD8+ Trm-mitotic ex-
icantly after treatment in patients who achieved pCR response pressed proinflammatory factors (CCL3, CCL4) and damage-
(Figure 6H). HLA-DRA+ EC expressed chemokine receptor D6 associated molecular patterns (DAMPs) such as HMGB1 and
(ACKR1), SELP, SELE, and VCAM1 (Figure 6G), which regulate HMGB2 (Figure S7A), and have higher inflammatory response
the migration of leukocytes through the vessel wall.48 High scores than other CD8+ T cells (Figure S7B). What’s more,
expression of major histocompatibility complex class II (MHC- IL1B+ Mono cells were positively correlated with the abundance
II) surface molecule HLA-DR was found in HLA-DRA+ EC cluster of CD8+ Trm-mitotic cells and, consistently, with the proliferation
(Figure 6D), which allow antigen presentation to CD4+ and exhaustion process (Figure 7C). Consistently, CD8 inflam-
T cells.49–51 To investigate global cell-cell interactions in d- matory response program was related to the exhaustion and
MMR/MSI-H CRC, we performed computational modeling by proliferation signature of CD8 TRMs (Figure S7C). These results
combining scRNA-seq and TCGA bulk RNA-seq dataset anal- suggest a positive feedback loop between CD8 TRMs and the in-
ysis, a method previously used in cancer studies.38,52,53 We flammatory signaling pathway and that inflammation was asso-
thereby inferred the correlative cell-cell interactions of specific ciated with the exhaustion program of CD8+ T cells.
cell types. We observed that HLA-DRA+ EC harbored more con- Next, we sought to explore whether remodeling of CD4+
nections with B cells and CD4+ T cells (Figure S6F) than other T cells and B cells also has conspicuous overlap with resolution
stromal cells, indicating an important role in assisting immune of inflammation. In this study, Tregs have a transcriptional profile
activation. similar to inflammation-derived Tregs64 characterized by upre-
In contrast to HLA-DRA+ EC and CXCL12+ Fibro, PLVAP+ EC gulation of both a core Treg (FOXP3, CTLA4, TIGIT) and effector
and CCL2+ Fibro significantly decreased after treatment program (TNFRSF18, PRDM1, BATF) (Figure S4C). We observed
(Figures 6B and 6H). By analyzing characteristics of these two increased expression of IL1b receptors (IL1R1, IL1R2) in Treg
cell clusters, we found they both expressed IL32 (Figure 6E), cells comparing other CD4+ T cells (Figure S7D). Of note, IL1b re-
which was reported to be involved in proinflammatory and tumor ceptor positive (IL1R+) Tregs had higher Tregs core gene expres-
promotion processes.54 CCL2+ Fibro also displayed characteris- sion than negative Tregs (Figure S7E). These results reveal the
tics of cancer-associated fibroblasts (FAP, TWIST1,WNT5A)55,56 association between the inflammatory environment and Tregs,
and cytokines that recruit inflammatory monocytes (CCL2, and suggest the possible mechanism of CD4 remodeling during
CCL11) (Figure 6A). In addition, PLVAP+ EC highly expressed inflammatory resolution in ICI-induced pCR response. For B
markers associated with previously reported PLPP3+, cells, we observed ICI treatment induced a significant pheno-
IGFBP3+, and PLVAP+ EC (Figure S2F) that located with tumor typic switch from plasma B to CD20+ B cells in pCR response
associated macrophages, Tregs, and tumor epithelial cells (Figures 4F and 7A). It has been reported that immunoglobulin

(B) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
(C) Heatmap of scaled normalized expression for CXCR4, CXCR5, CXCR6, and CXCR3 genes in immune cell clusters. Significant DEGs between
ICI and +ICI/
pCR, as well as
ICI and +ICI/non-pCR are marked with asterisks.
(D) Dot plots showing selected genes across endothelial cell clusters. Dot size indicates fraction of expressing cells, colored according to Z score normalized
expression levels.
(E) UMAP plots and Dot plots showing IL32 gene expression across endothelial cell and fibroblast clusters.
(F) Circle plots showing the inferred CXCL12-CXCR4 signaling among cell types under
ICI, +ICI/pCR, and +ICI/non-pCR, respectively.
(G) Dot plots showing selected genes across endothelial cell clusters. Dot size indicates fraction of expressing cells, colored according to Z score normalized
expression levels.
(H) Quantification of cell cluster frequency representation among patient groups. The median and interquartile range are shown for each patient group. Two-sided
Wilcoxon test.
Also see Figure S6.

Cancer Cell 41, 1–18, June 12, 2023 11

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

(Ig)A-producing and IgG-producing B cells favor tumor growth we do not know the specific correlation between inflammation
by contribution to chronic inflammation.65–69 In this study, we and processes of proliferation and exhaustion in CD8+ T cells.
found a significant decrease of Ig producing B cells after treat- In this study, we used an in vivo model and found that IL1b cyto-
ment, especially for IgA plasma B cells (Figure 4G). kine increased the proportion of PD-1+ exhausted CD8+ T cells;
Taken together, remodeling of immune and stromal cells by ICI however, due to low flow gating of PD-1+CD8+ cells in the IL1b--
in pCR response has striking overlap with pro-tumor inflamma- treated group, the association between IL1b and exhausted
tory resolution (Figure 7B). However, unlike pCR, non-pCR did CD8+ T cells needs to be further validated. Yet we observed a
not significantly decrease proinflammatory cells (IL1B+ Mono, significant decrease of CD8+ T cells and CD40+ B cells in the
CCL2+ Fibro) after ICI treatment (Figure 7A). Proinflammatory IL1b-treated group, which provides valuable insight on the ther-
IL1B+ Mono cells were enriched in +ICI/non-pCR compared apeutic target in immunotherapy.
with +ICI/pCR (Figure 7A). Among the ICI efficacy associated Despite significant heterogeneity, similar tumor microenviron-
cells, inflammatory associated Tregs, CD8+ Trm-mitotic, and ment remodeling has been observed in a variety of tumors, such
IgA plasma B cells were also enriched in +ICI/non-pCR as reduction of effector T cells, yet an accumulation of exhausted
compared with +ICI/pCR (Figure 7A). In addition, proinflamma- CD8+ T cells and suppressive Tregs, reprograming of myeloid cells
tory factors (IL1A, IL1B, CCL2/3/4, IL6, IL32, CXCL1, CXCL8, to harbor immunosuppressive phenotypes.74,75 Tumors engage
etc.) from myeloid cells, endothelial cells, and fibroblasts were common patterns of the immune archetypes and modulate
significantly more highly expressed in the +ICI/non-pCR group response to ICI.76,77 Dynamics of immune cells during ICI treat-
(Figure S7F). Together, these results suggesting that the mech- ment in d-MMR/MSI-H CRC found in this study is similar to that
anism of non-pCR may be a failure of inflammation resolution of other tumors. For example, in the treatment of non-small cell
(Figure 7B). We next sought to elucidate the effects of the inflam- lung cancer with PD-1 blockade, Liu et al. observed that Tregs
matory environment on immune cell populations in an in vivo decreased only in the responders, and the non-exhausted
model. We found IL-1b cytokine treated MC38 mice significantly precursor GZMK+ CD8+ T cells increased only in the responders.23
decreased CD8+ T cells and CD40+ B cells (Figure S7G). In melanoma treated with ICI therapy (anti-PD-1 and/or anti-
Although proportions of PD-1+ exhausted CD8+ cells and Treg CTLA4), Sade et al. found that proliferative CD8+ T cells and
cells were low, the populations of these cells were potentially myeloid cells were enriched in the non-responder group, and B
increased after IL-b treatment (Figure S7G). These data cells rather than plasma cells were enriched in the responder
confirmed correlation between proinflammatory molecules and group.26 In advanced renal cell carcinoma, myeloid inflammation
immunosuppressive TIME and suggested potential targets to is associated with poor overall survival, while the lymphatic system
improve ICI treatment for d-MMR/MSI-H CRC. is associated with good overall survival.13 Collectively, these re-
sults indicate that effective responses to ICIs require non-ex-
DISCUSSION hausted T cells as well as depletion of immunosuppressive cells
including Tregs and myeloid subsets. Although it has been found
We investigated the cellular dynamics during complete responses that stromal cells can affect the therapeutic effect of ICIs,78,79 inno-
to ICI treatment in locally advanced d-MMR/MSI-H CRC to ICI vative single-cell approach-based cohort study on the relationship
treatment. The nonmetastatic neoadjuvant treatment cohort al- between stromal cells and ICI treatment are relatively lacking. The
lowed us to clearly observe the specific changes induced by suc- correlation between stromal cells and the ICI therapeutic effect
cessful immunotherapy. Therefore, we were able to identify the mentioned in this study provides deeper insights into the functional
molecular mechanism of immunotherapy in d-MMR/MSI-H CRC states and dynamics of stromal cells.
and the mechanism of immune escape of residual tumor cells. Although we have observed the similarity of tumor immune re-
In this study, the overall Trm signature was enriched in the modeling in some solid tumors and the importance of relieving
non-pCR group (Figure 2F). However, Trm signature has been the immunosuppressive effect on the response to immuno-
previously associated with better prognosis in lung cancer21 therapy, it may not be applicable to other diseases called
and breast cancer.19 A possible explanation was that the gene "cold" tumors. While these diseases have not been approved
set we use is based on features that distinguish Trm from non- for immunotherapy, relevant research is lacking. These tumors
Trm. Therefore, it includes the feature of Trm-mitotic as well, lack immune cell infiltration, and inflammatory response might
which is related to non-pCR. Our results suggest the importance be needed to recruit immune cells and trigger immune response
of identification for Trm subgroups. The funding that proliferating either by ICI or conventional chemotherapeutic agents and radi-
CD8+ Trm-mitotic cells decreased after ICI treatment was ation therapy.39 Increased treatment efficacy provided by block-
consistent with the result in a study investigated ICI treatment ing inflammatory signaling pathways (such as IL-1b, IL-6, COX2/
of melanoma.26 Proliferating CD8+ T cells has been reported to PGE2) may be more likely to occur in ‘‘hot’’ tumors. In this study,
be associated with T cell exhaustion in a variety of tumors70–72; we concluded that non-pCR samples displayed higher expres-
however, the mechanism of their formation and maintenance is sion of proinflammatory factors compared with pCR samples.
still unclear. We proposed that the inflammatory environment However, frequencies of IL1B+CD14+CD68
were compared
promotes the transformation from CD8+ Trm cells to CD8+ between pCR and non-pCR samples according to mIHC stain-
Trm-mitotic cells. Consistently, a study performed global anal- ing. It may be due to a small sample size in the validation cohort,
ysis of immune cell populations in ICI-induced colitis and found or the sub-optimal combination of markers that defined IL1B+
a shift from Trm to proliferating T cells in CD8+ T cells.73 Mono cells. Future study is needed to validate the association
Together, these results highlight the association between inflam- in a larger cohort and optimize the detection methods for the
mation and the formation of proliferative CD8+ T cells. However, key inflammatory factors and their source cells.

12 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Figure 7. Proinflammatory IL1B+ Mono cells were enriched in non-pCR group

(A) Characteristics and summary of key immune and stromal cell cluster dynamics of ICI treatment. *Red (or green) arrows represent significant increase (or
decrease) of cell clusters in +ICI/pCR and +ICI/non-pCR compared with
ICI, dashed lines represent no significant difference. yRed arrows represent significantly
higher cell cluster frequency in either +ICI/pCR or +ICI/non-pCR.
(B) Difference of cell cluster between pCR and non-pCR and their association with inflammation.
(C) Quantification of cell cluster frequency representation among the IL1B+ Mono-positive and IL1B+ Mono-negative sample groups. The median and interquartile
range are shown for each patient group. Two-sided Wilcoxon test.
(D) Representative images of mIHC staining indicating IL1B+CD14+CD68
cells, in +ICI/pCR and +ICI/non-pCR samples. Each dot in boxplot presents the
fractions of IL1B+CD14+CD68
cells in each sample based on the IHC staining image. Two-sided Wilcoxon test.
Also see Figure S7.

Cancer Cell 41, 1–18, June 12, 2023 13

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

We reviewed the changes of all cell clusters in different treat- d EXPERIMENTAL MODEL AND SUBJECT DETAILS
ment regimens and found that among all cell types, only B Patient population and clinical study
CXCL12+ EC were significantly less in the celecoxib combination B In vivo mouse models
group than the anti-PD-1-only group, as shown in the supple- d METHOD DETAILS
mentary figures (Figures S3B, S4A, S4I, S5B, S6B, and S6C). B Sample collection and dissociation for scRNA-seq
However, CXCL12+ EC was not pCR associated (Figure 6F). B Single-cell RNA-seq library preparation and
Consistent with this biological phenomenon, the results from sequencing
the PICC study showed that although celecoxib combination B Multiplex fluorescence immunohistochemistry
regimen improved the pCR rate, the difference between PD-1 B Luminex-based cytokine/chemokine detection for
blockade monotherapy and celecoxib combination was not sta- serum samples
tistically significant (88% in the monotherapy group vs. 65% in B Tissue digestion and flow cytometry analysis for
the combination therapy group, p = 0.23).9 The possible expla- mouse samples
nation is that PD-1 blockade achieved a very high pCR rate in B Serum cytokine/chemokine data analysis
the neoadjuvant setting, which may obsecure some contribution B Pre-processing of single cell RNA-seq data
from celecoxib. Anyhow, our results provide clues about immune B Unsupervised clustering analysis and identification of
cell subsets that are most likely to be affected by celecoxib, broad cell type
which may be helpful for future study. B Removing cells with high expression level of mitochon-
Limitations of our study include the small sample size and lack drial genes
of pretreatment non-pCR samples. Further studies are needed B Dimensionality reduction using UMAP
to explore the mechanism of immune modulation by PD-1 B Re-clustering of broad cell types
blockade. In addition, due to the lack of pretreatment non-pCR B Identification of marker genes
samples, we were unable to identify cell clusters that are associ- B Differential expression and functional annotation
ated with pCR response in the baseline setting, which limited the B Definition of gene signature score
significance of efficacy prediction. By comparing the differences B Gene set enrichment analysis
between pCR and non-pCR after treatment, we identified poten- B Correlation analysis
tial targets to improve efficacy of ICI. However, how these results B Cell-to-cell communication of scRNA-seq data
may apply to intrinsically resistant tumors (e.g., in stage IV dis- B Cell developmental trajectory
ease) needs to be further explored. Since the actual tumor B Cell subtype abundance estimated from TCGA bulk
regression grade has reached more than 90% in non-pCR in gene expression profiles
this study, which is not completely the same as the concept of B Correlative cell-cell interactions inferred by combined
resistance in stage IV disease, which is defined as that the tumor scRNA-seq and TCGA MSI-H CRC datasets
volume does not shrink or even increase after treatment. It d QUANTIFICATION AND STATISTICAL ANALYSIS
should be noted that the research aimed at improving pCR has
important clinical significance, which is not only to prolong sur- SUPPLEMENTAL INFORMATION
vival, but more importantly, to exempt surgery to protect organ
function. The phenomenon found in sustaining non-pCR signa- Supplemental information can be found online at https://doi.org/10.1016/j.
ccell.2023.04.011.
ture contributes to understanding the knowledge of complete
response in neoadjuvant ICI treatment, and thus to identify pa-
ACKNOWLEDGMENTS
tients who may be suitable for sphincter/organ-preserving or
watch-and-wait strategies. This project was supported by National Natural Science Foundation of China,
In conclusion, our study illuminates key immune and stromal China (82272800, 81974369, 82102952), Science and Technology Program of
cell subsets during ICI treatment and provides new insights Guangzhou, China (202206080011), National Natural Science Foundation of
into the cellular underpinnings of distinct responses to PD-1 Guangdong Province, China (2021A1515010568), National Science Fund for
Distinguished Young Scholars, China (82002944), and Key Technologies
blockade in patients with d-MMR/MSI-H CRC. Our study also
Research and Development Program of China, China (2019YFC1316003).
provides evidence of an association between inflammation and
ICI-induced treatment responses. These data also provide a
AUTHOR CONTRIBUTIONS
rich resource for the identification of therapeutic targets for
d-MMR/MSI-H CRC, which requires further investigation. Conceptualization: Y.H.D., J.X.L., and H.B.H.; Methodology: J.X.L., C.Wu,
H.B.H., G.Q., and Z.X.; Investigation: X.Q.W., F.B., and J.Q.Y.; Data Curation:
Y.Z.L., J.X.L., C.Wu, and Z.H.J.; Resources: G.Q., Y.H., C.Wang, J.W.Z., Y.C.,
STAR+METHODS J.Y.L., Z.H.W., and Y.X.C.; Writing – Original Draft: Y.H.D., J.X.L., C.Wu.,
H.B.H., and Z.X.; Writing – Review & Editing: Y.H.D., J.X.L., and G.Q.; Funding
Detailed methods are provided in the online version of this paper Acquisition: Y.H.D., J.W.Z., and G.Q.; Supervision: Y.H.D.
and include the following:
DECLARATION OF INTERESTS
d KEY RESOURCES TABLE
The authors declare no competing interests.
d RESOURCE AVAILABILITY
B Lead contact
INCLUSION AND DIVERSITY
B Materials availability
B Data and code availability We support inclusive, diverse, and equitable conduct of research.

14 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article
Received: July 8, 2022 with breast cancer. Nat. Med. 27, 820–832. https://doi.org/10.1038/
Revised: January 6, 2023 s41591-021-01323-8.
Accepted: April 18, 2023 13. Bi, K., He, M.X., Bakouny, Z., Kanodia, A., Napolitano, S., Wu, J., Grimaldi,
Published: May 11, 2023 G., Braun, D.A., Cuoco, M.S., Mayorga, A., et al. (2021). Tumor and im-
mune reprogramming during immunotherapy in advanced renal cell carci-
REFERENCES noma. Cancer Cell 39, 649–661.e5. https://doi.org/10.1016/j.ccell.2021.
02.015.
1. Vilar, E., and Gruber, S.B. (2010). Microsatellite instability in colorectal
14. Jerby-Arnon, L., Shah, P., Cuoco, M.S., Rodman, C., Su, M.J., Melms,
cancer-the stable evidence. Nat. Rev. Clin. Oncol. 7, 153–162. https://
J.C., Leeson, R., Kanodia, A., Mei, S., Lin, J.R., et al. (2018). A cancer
doi.org/10.1038/nrclinonc.2009.237.
cell program promotes T cell exclusion and resistance to checkpoint
2. Giannakis, M., Mu, X.J., Shukla, S.A., Qian, Z.R., Cohen, O., Nishihara, R., blockade. Cell 175, 984–997.e24. https://doi.org/10.1016/j.cell.2018.
Bahl, S., Cao, Y., Amin-Mansour, A., Yamauchi, M., et al. (2016). Genomic 09.006.
correlates of immune-cell infiltrates in colorectal carcinoma. Cell Rep. 15,
15. Zhang, Y., Chen, H., Mo, H., Hu, X., Gao, R., Zhao, Y., Liu, B., Niu, L., Sun,
857–865. https://doi.org/10.1016/j.celrep.2016.03.075.
X., Yu, X., et al. (2021). Single-cell analyses reveal key immune cell subsets
3. Llosa, N.J., Cruise, M., Tam, A., Wicks, E.C., Hechenbleikner, E.M., associated with response to PD-L1 blockade in triple-negative breast can-
Taube, J.M., Blosser, R.L., Fan, H., Wang, H., Luber, B.S., et al. (2015). cer. Cancer Cell 39, 1578–1593.e8. https://doi.org/10.1016/j.ccell.2021.
The vigorous immune microenvironment of microsatellite instable colon 09.010.
cancer is balanced by multiple counter-inhibitory checkpoints. Cancer
16. Li, H., van der Leun, A.M., Yofe, I., Lubling, Y., Gelbard-Solodkin, D., van
Discov. 5, 43–51. https://doi.org/10.1158/2159-8290.Cd-14-0863.
Akkooi, A.C.J., van den Braber, M., Rozeman, E.A., Haanen, J.B., Blank,
4. André, T., Shiu, K.K., Kim, T.W., Jensen, B.V., Jensen, L.H., Punt, C.,
C.U., et al. (2019). Dysfunctional CD8 T cells form a proliferative, dynam-
Smith, D., Garcia-Carbonero, R., Benavides, M., Gibbs, P., et al. (2020).
ically regulated compartment within human melanoma. Cell 176, 775–
Pembrolizumab in microsatellite-instability-high advanced colorectal can-
789.e18. https://doi.org/10.1016/j.cell.2018.11.043.
cer. N. Engl. J. Med. 383, 2207–2218. https://doi.org/10.1056/
17. Zheng, Y., Chen, Z., Han, Y., Han, L., Zou, X., Zhou, B., Hu, R., Hao, J., Bai,
NEJMoa2017699.
S., Xiao, H., et al. (2020). Immune suppressive landscape in the human
5. Le, D.T., Kim, T.W., Van Cutsem, E., Geva, R., Ja€ger, D., Hara, H., Burge,
esophageal squamous cell carcinoma microenvironment. Nat. Commun.
M., O’Neil, B., Kavan, P., Yoshino, T., et al. (2020). Phase II open-label
11, 6268. https://doi.org/10.1038/s41467-020-20019-0.
study of pembrolizumab in treatment-refractory, microsatellite insta-
18. Milner, J.J., Toma, C., Yu, B., Zhang, K., Omilusik, K., Phan, A.T., Wang,
bility-high/mismatch repair-deficient metastatic colorectal cancer:
D., Getzler, A.J., Nguyen, T., Crotty, S., et al. (2017). Runx3 programs
KEYNOTE-164. J. Clin. Oncol. 38, 11–19. https://doi.org/10.1200/jco.
CD8(+) T cell residency in non-lymphoid tissues and tumours. Nature
19.02107.
552, 253–257. https://doi.org/10.1038/nature24993.
6. Le, D.T., Uram, J.N., Wang, H., Bartlett, B.R., Kemberling, H., Eyring, A.D.,
Skora, A.D., Luber, B.S., Azad, N.S., Laheru, D., et al. (2015). PD-1 19. Savas, P., Virassamy, B., Ye, C., Salim, A., Mintoff, C.P., Caramia, F.,
blockade in tumors with mismatch-repair deficiency. N. Engl. J. Med. Salgado, R., Byrne, D.J., Teo, Z.L., Dushyanthen, S., et al. (2018).
372, 2509–2520. https://doi.org/10.1056/NEJMoa1500596. Single-cell profiling of breast cancer T cells reveals a tissue-resident mem-
ory subset associated with improved prognosis. Nat. Med. 24, 986–993.
7. Overman, M.J., McDermott, R., Leach, J.L., Lonardi, S., Lenz, H.J., Morse,
https://doi.org/10.1038/s41591-018-0078-7.
M.A., Desai, J., Hill, A., Axelson, M., Moss, R.A., et al. (2017). Nivolumab in
patients with metastatic DNA mismatch repair-deficient or microsatellite 20. Duhen, T., Duhen, R., Montler, R., Moses, J., Moudgil, T., de Miranda,
instability-high colorectal cancer (CheckMate 142): an open-label, multi- N.F., Goodall, C.P., Blair, T.C., Fox, B.A., McDermott, J.E., et al. (2018).
centre, phase 2 study. Lancet Oncol. 18, 1182–1191. https://doi.org/10. Co-expression of CD39 and CD103 identifies tumor-reactive CD8 T cells
1016/s1470-2045(17)30422-9. in human solid tumors. Nat. Commun. 9, 2724. https://doi.org/10.1038/
s41467-018-05072-0.
8. Chalabi, M., Fanchi, L.F., Dijkstra, K.K., Van den Berg, J.G., Aalbers, A.G.,
Sikorska, K., Lopez-Yurda, M., Grootscholten, C., Beets, G.L., 21. Clarke, J., Panwar, B., Madrigal, A., Singh, D., Gujar, R., Wood, O., Chee,
Snaebjornsson, P., et al. (2020). Neoadjuvant immunotherapy leads to S.J., Eschweiler, S., King, E.V., Awad, A.S., et al. (2019). Single-cell tran-
pathological responses in MMR-proficient and MMR-deficient early-stage scriptomic analysis of tissue-resident memory T cells in human lung can-
colon cancers. Nat. Med. 26, 566–576. https://doi.org/10.1038/s41591- cer. J. Exp. Med. 216, 2128–2149. https://doi.org/10.1084/jem.20190249.
020-0805-8. 22. Prasad, S., Hu, S., Sheng, W.S., Chauhan, P., Singh, A., and Lokensgard,
9. Hu, H., Kang, L., Zhang, J., Wu, Z., Wang, H., Huang, M., Lan, P., Wu, X., J.R. (2017). The PD-1: PD-L1 pathway promotes development of
Wang, C., Cao, W., et al. (2022). Neoadjuvant PD-1 blockade with toripa- brain-resident memory T cells following acute viral encephalitis.
limab, with or without celecoxib, in mismatch repair-deficient or microsat- J. Neuroinflammation 14, 82. https://doi.org/10.1186/s12974-017-
ellite instability-high, locally advanced, colorectal cancer (PICC): a single- 0860-3.
centre, parallel-group, non-comparative, randomised, phase 2 trial. 23. Liu, B., Hu, X., Feng, K., Gao, R., Xue, Z., Zhang, S., Zhang, Y., Corse, E.,
Lancet. Gastroenterol. Hepatol. 7, 38–48. https://doi.org/10.1016/ Hu, Y., Han, W., and Zhang, Z. (2022). Temporal single-cell tracing reveals
s2468-1253(21)00348-4. clonal revival and expansion of precursor exhausted T cells during anti-
10. Cercek, A., Lumish, M., Sinopoli, J., Weiss, J., Shia, J., Lamendola-Essel, M., PD-1 therapy in lung cancer. Nat. Cancer 3, 108–121. https://doi.org/10.
El Dika, I.H., Segal, N., Shcherba, M., Sugarman, R., et al. (2022). PD-1 1038/s43018-021-00292-8.
blockade in mismatch repair-deficient, locally advanced rectal cancer. 24. Wei, S.C., Levine, J.H., Cogdill, A.P., Zhao, Y., Anang, N.A.S., Andrews,
N. Engl. J. Med. 386, 2363–2376. https://doi.org/10.1056/NEJMoa2201445. M.C., Sharma, P., Wang, J., Wargo, J.A., Pe’er, D., and Allison, J.P.
11. Bortolomeazzi, M., Keddar, M.R., Montorsi, L., Acha-Sagredo, A., (2017). Distinct cellular mechanisms underlie anti-CTLA-4 and anti-PD-1
Benedetti, L., Temelkovski, D., Choi, S., Petrov, N., Todd, K., Wai, P., checkpoint blockade. Cell 170, 1120–1133.e17. https://doi.org/10.1016/
et al. (2021). Immunogenomics of colorectal cancer response to check- j.cell.2017.07.024.
point blockade: analysis of the KEYNOTE 177 trial and validation cohorts. 25. Yost, K.E., Satpathy, A.T., Wells, D.K., Qi, Y., Wang, C., Kageyama, R.,
Gastroenterology 161, 1179–1193. https://doi.org/10.1053/j.gastro.2021. McNamara, K.L., Granja, J.M., Sarin, K.Y., Brown, R.A., et al. (2019).
06.064. Clonal replacement of tumor-specific T cells following PD-1 blockade.
12. Bassez, A., Vos, H., Van Dyck, L., Floris, G., Arijs, I., Desmedt, C., Boeckx, Nat. Med. 25, 1251–1259. https://doi.org/10.1038/s41591-019-0522-3.
B., Vanden Bempt, M., Nevelsteen, I., Lambein, K., et al. (2021). A single- 26. Sade-Feldman, M., Yizhak, K., Bjorgaard, S.L., Ray, J.P., de Boer, C.G.,
cell map of intratumoral changes during anti-PD1 treatment of patients Jenkins, R.W., Lieb, D.J., Chen, J.H., Frederick, D.T., Barzily-Rokni, M.,

Cancer Cell 41, 1–18, June 12, 2023 15

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article
et al. (2018). Defining T cell states associated with response to checkpoint 41. Asrir, A., Tardiveau, C., Coudert, J., Laffont, R., Blanchard, L., Bellard, E.,
immunotherapy in melanoma. Cell 175, 998–1013.e20. https://doi.org/10. Veerman, K., Bettini, S., Lafouresse, F., Vina, E., et al. (2022). Tumor-asso-
1016/j.cell.2018.10.038. ciated high endothelial venules mediate lymphocyte entry into tumors and
27. Miller, B.C., Sen, D.R., Al Abosy, R., Bi, K., Virkud, Y.V., LaFleur, M.W., predict response to PD-1 plus CTLA-4 combination immunotherapy.
Yates, K.B., Lako, A., Felt, K., Naik, G.S., et al. (2019). Subsets of ex- Cancer Cell 40, 318–334.e9. https://doi.org/10.1016/j.ccell.2022.01.002.
hausted CD8(+) T cells differentially mediate tumor control and respond 42. Bagaev, A., Kotlov, N., Nomie, K., Svekolkin, V., Gafurov, A., Isaeva, O.,
to checkpoint blockade. Nat. Immunol. 20, 326–336. https://doi.org/10. Osokin, N., Kozlov, I., Frenkel, F., Gancharova, O., et al. (2021).
1038/s41590-019-0312-6. Conserved pan-cancer microenvironment subtypes predict response to
28. Siddiqui, I., Schaeuble, K., Chennupati, V., Fuertes Marraco, S.A., immunotherapy. Cancer Cell 39, 845–865.e7. https://doi.org/10.1016/j.
Calderon-Copete, S., Pais Ferreira, D., Carmona, S.J., Scarpellino, L., ccell.2021.04.014.
Gfeller, D., Pradervand, S., et al. (2019). Intratumoral Tcf1(+)PD-1(+) 43. Mariathasan, S., Turley, S.J., Nickles, D., Castiglioni, A., Yuen, K., Wang,
CD8(+) T cells with stem-like properties promote tumor control in response Y., Kadel, E.E., III, Koeppen, H., Astarita, J.L., Cubas, R., et al. (2018).
to vaccination and checkpoint blockade immunotherapy. Immunity 50, TGFb attenuates tumour response to PD-L1 blockade by contributing to
195–211.e10. https://doi.org/10.1016/j.immuni.2018.12.021. exclusion of T cells. Nature 554, 544–548. https://doi.org/10.1038/
29. Gubin, M.M., Esaulova, E., Ward, J.P., Malkova, O.N., Runci, D., Wong, P., nature25501.
Noguchi, T., Arthur, C.D., Meng, W., Alspach, E., et al. (2018). High-dimen- 44. Kinchen, J., Chen, H.H., Parikh, K., Antanaviciute, A., Jagielowicz, M.,
sional analysis delineates myeloid and lymphoid compartment remodeling Fawkner-Corbett, D., Ashley, N., Cubitt, L., Mellado-Gomez, E., Attar,
during successful immune-checkpoint cancer therapy. Cell 175, 1014– M., et al. (2018). Structural remodeling of the human colonic mesenchyme
1030.e19. https://doi.org/10.1016/j.cell.2018.09.030. in inflammatory bowel disease. Cell 175, 372–386.e17. https://doi.org/10.
30. Tanaka, A., and Sakaguchi, S. (2017). Regulatory T cells in cancer immu- 1016/j.cell.2018.08.067.
notherapy. Cell Res. 27, 109–118. https://doi.org/10.1038/cr.2016.151. 45. Feig, C., Jones, J.O., Kraman, M., Wells, R.J.B., Deonarine, A., Chan, D.S.,
€ller, W. Connell, C.M., Roberts, E.W., Zhao, Q., Caballero, O.L., et al. (2013).
31. Borst, J., Ahrends, T., Ba˛ba1a, N., Melief, C.J.M., and Kastenmu
Targeting CXCL12 from FAP-expressing carcinoma-associated fibro-
(2018). CD4(+) T cell help in cancer immunology and immunotherapy.
blasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer.
Nat. Rev. Immunol. 18, 635–647. https://doi.org/10.1038/s41577-018-
Proc. Natl. Acad. Sci. USA 110, 20212–20217. https://doi.org/10.1073/
0044-0.
pnas.1320318110.
32. Byrne, K.T., and Vonderheide, R.H. (2016). CD40 stimulation obviates
46. Krishnamurty, A.T., and Turley, S.J. (2020). Lymph node stromal cells: car-
innate sensors and drives T cell immunity in cancer. Cell Rep. 15, 2719–
tographers of the immune system. Nat. Immunol. 21, 369–380. https://doi.
2732. https://doi.org/10.1016/j.celrep.2016.05.058.
org/10.1038/s41590-020-0635-3.
33. Elgueta, R., Benson, M.J., de Vries, V.C., Wasiuk, A., Guo, Y., and Noelle,
47. Rodda, L.B., Bannard, O., Ludewig, B., Nagasawa, T., and Cyster, J.G.
R.J. (2009). Molecular mechanism and function of CD40/CD40L engage-
(2015). Phenotypic and morphological properties of germinal center dark
ment in the immune system. Immunol. Rev. 229, 152–172. https://doi.
zone Cxcl12-expressing reticular cells. J. Immunol. 195, 4781–4791.
org/10.1111/j.1600-065X.2009.00782.x.
https://doi.org/10.4049/jimmunol.1501191.
34. Brooks, J.F., Tan, C., Mueller, J.L., Hibiya, K., Hiwa, R., Vykunta, V., and
48. Vestweber, D. (2015). How leukocytes cross the vascular endothelium.
Zikherman, J. (2021). Negative feedback by NUR77/Nr4a1 restrains B
Nat. Rev. Immunol. 15, 692–704. https://doi.org/10.1038/nri3908.
cell clonal dominance during early T-dependent immune responses. Cell
Rep. 36, 109645. https://doi.org/10.1016/j.celrep.2021.109645. 49. Carman, C.V., and Martinelli, R. (2015). T lymphocyte-endothelial interac-
tions: emerging understanding of Trafficking and antigen-specific immu-
35. Cabrita, R., Lauss, M., Sanna, A., Donia, M., Skaarup Larsen, M., Mitra, S.,
nity. Front. Immunol. 6, 603. https://doi.org/10.3389/fimmu.2015.00603.
Johansson, I., Phung, B., Harbst, K., Vallon-Christersson, J., et al. (2020).
Tertiary lymphoid structures improve immunotherapy and survival in mel- 50. Savage, C.O., Brooks, C.J., Harcourt, G.C., Picard, J.K., King, W.,
anoma. Nature 577, 561–565. https://doi.org/10.1038/s41586-019- Sansom, D.M., and Willcox, N. (1995). Human vascular endothelial cells
1914-8. process and present autoantigen to human T cell lines. Int. Immunol. 7,
471–479. https://doi.org/10.1093/intimm/7.3.471.
36. Meednu, N., Rangel-Moreno, J., Zhang, F., Escalera-Rivera, K., Corsiero,
E., Prediletto, E., DiCarlo, E., Goodman, S., Donlin, L.T., Raychauduri, S., 51. Ma, W., and Pober, J.S. (1998). Human endothelial cells effectively costi-
et al. (2022). Dynamic spectrum of ectopic lymphoid B cell activation and mulate cytokine production by, but not differentiation of, naive CD4+
hypermutation in the RA synovium characterized by NR4A nuclear recep- T cells. J. Immunol. 161, 2158–2167.
tor expression. Cell Rep. 39, 110766. https://doi.org/10.1016/j.celrep. 52. Puram, S.V., Tirosh, I., Parikh, A.S., Patel, A.P., Yizhak, K., Gillespie, S.,
2022.110766. Rodman, C., Luo, C.L., Mroz, E.A., Emerick, K.S., et al. (2017). Single-
37. van Hooren, L., Vaccaro, A., Ramachandran, M., Vazaios, K., Libard, S., cell transcriptomic analysis of primary and metastatic tumor ecosystems
van de Walle, T., Georganaki, M., Huang, H., Pietila€, I., Lau, J., et al. in head and neck cancer. Cell 171, 1611–1624.e24. https://doi.org/10.
(2021). Agonistic CD40 therapy induces tertiary lymphoid structures but 1016/j.cell.2017.10.044.
impairs responses to checkpoint blockade in glioma. Nat. Commun. 12, 53. Tirosh, I., Izar, B., Prakadan, S.M., Wadsworth, M.H., 2nd, Treacy, D.,
4127. https://doi.org/10.1038/s41467-021-24347-7. Trombetta, J.J., Rotem, A., Rodman, C., Lian, C., Murphy, G., et al.
38. Zhang, L., Li, Z., Skrzypczynska, K.M., Fang, Q., Zhang, W., O’Brien, S.A., (2016). Dissecting the multicellular ecosystem of metastatic melanoma
He, Y., Wang, L., Zhang, Q., Kim, A., et al. (2020). Single-cell analyses by single-cell RNA-seq. Science 352, 189–196. https://doi.org/10.1126/
inform mechanisms of myeloid-targeted Therapies in colon cancer. Cell science.aad0501.
181, 442–459.e29. https://doi.org/10.1016/j.cell.2020.03.048. 54. Hong, J.T., Son, D.J., Lee, C.K., Yoon, D.Y., Lee, D.H., and Park, M.H.
39. Zhao, H., Wu, L., Yan, G., Chen, Y., Zhou, M., Wu, Y., and Li, Y. (2021). (2017). Interleukin 32, inflammation and cancer. Pharmacol. Ther. 174,
Inflammation and tumor progression: signaling pathways and targeted 127–137. https://doi.org/10.1016/j.pharmthera.2017.02.025.
intervention. Signal Transduct. Target. Ther. 6, 263. https://doi.org/10. 55. Erez, N., Truitt, M., Olson, P., Arron, S.T., and Hanahan, D. (2010). Cancer-
1038/s41392-021-00658-5. associated fibroblasts are activated in incipient neoplasia to orchestrate
40. Pelka, K., Hofree, M., Chen, J.H., Sarkizova, S., Pirl, J.D., Jorgji, V., tumor-promoting inflammation in an NF-kappaB-Dependent manner.
Bejnood, A., Dionne, D., Ge, W.H., Xu, K.H., et al. (2021). Spatially orga- Cancer Cell 17, 135–147. https://doi.org/10.1016/j.ccr.2009.12.041.
nized multicellular immune hubs in human colorectal cancer. Cell 184, 56. Ma, Z., Li, X., Mao, Y., Wei, C., Huang, Z., Li, G., Yin, J., Liang, X., and Liu,
4734–4752.e20. https://doi.org/10.1016/j.cell.2021.08.003. Z. (2022). Interferon-dependent SLC14A1(+) cancer-associated

16 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article
fibroblasts promote cancer stemness via WNT5A in bladder cancer. 72. Zhang, L., Yu, X., Zheng, L., Zhang, Y., Li, Y., Fang, Q., Gao, R., Kang, B.,
Cancer Cell 40, 1550–1565.e1557. https://doi.org/10.1016/j.ccell.2022. Zhang, Q., Huang, J.Y., et al. (2018). Lineage tracking reveals dynamic re-
11.005. lationships of T cells in colorectal cancer. Nature 564, 268–272. https://
57. Sharma, A., Seow, J.J.W., Dutertre, C.A., Pai, R., Blériot, C., Mishra, A., doi.org/10.1038/s41586-018-0694-x.
Wong, R.M.M., Singh, G.S.N., Sudhagar, S., Khalilnezhad, S., et al. 73. Luoma, A.M., Suo, S., Williams, H.L., Sharova, T., Sullivan, K., Manos, M.,
(2020). Onco-fetal reprogramming of endothelial cells drives immunosup- Bowling, P., Hodi, F.S., Rahma, O., Sullivan, R.J., et al. (2020). Molecular
pressive macrophages in hepatocellular carcinoma. Cell 183, 377– pathways of colon inflammation induced by cancer immunotherapy. Cell
394.e21. https://doi.org/10.1016/j.cell.2020.08.040. 182, 655–671.e22. https://doi.org/10.1016/j.cell.2020.06.001.
58. Headland, S.E., and Norling, L.V. (2015). The resolution of inflammation: 74. Lavin, Y., Kobayashi, S., Leader, A., Amir, E.A.D., Elefant, N., Bigenwald,
principles and challenges. Semin. Immunol. 27, 149–160. https://doi. C., Remark, R., Sweeney, R., Becker, C.D., Levine, J.H., et al. (2017).
org/10.1016/j.smim.2015.03.014. Innate immune landscape in early lung adenocarcinoma by paired sin-
gle-cell analyses. Cell 169, 750–765.e17. https://doi.org/10.1016/j.cell.
59. Mantovani, A., Barajon, I., and Garlanda, C. (2018). IL-1 and IL-1 regulatory
2017.04.014.
pathways in cancer progression and therapy. Immunol. Rev. 281, 57–61.
https://doi.org/10.1111/imr.12614. 75. Davidson, S., Efremova, M., Riedel, A., Mahata, B., Pramanik, J.,
Huuhtanen, J., Kar, G., Vento-Tormo, R., Hagai, T., Chen, X., et al.
60. Zhang, Q., Zhu, B., and Li, Y. (2017). Resolution of cancer-promoting
(2020). Single-cell RNA sequencing reveals a dynamic stromal niche that
inflammation: a new approach for anticancer therapy. Front. Immunol. 8,
supports tumor growth. Cell Rep. 31, 107628. https://doi.org/10.1016/j.
71. https://doi.org/10.3389/fimmu.2017.00071.
celrep.2020.107628.
61. Schenkel, J.M., Fraser, K.A., Vezys, V., and Masopust, D. (2013). Sensing
76. Combes, A.J., Samad, B., Tsui, J., Chew, N.W., Yan, P., Reeder, G.C.,
and alarm function of resident memory CD8⁺ T cells. Nat. Immunol. 14,
Kushnoor, D., Shen, A., Davidson, B., Barczak, A.J., et al. (2022).
509–513. https://doi.org/10.1038/ni.2568.
Discovering dominant tumor immune archetypes in a pan-cancer census.
62. Schenkel, J.M., Fraser, K.A., Beura, L.K., Pauken, K.E., Vezys, V., and Cell 185, 184–203.e19. https://doi.org/10.1016/j.cell.2021.12.004.
Masopust, D. (2014). T cell memory. Resident memory CD8 T cells trigger
77. Thorsson, V., Gibbs, D.L., Brown, S.D., Wolf, D., Bortone, D.S., Ou Yang,
protective innate and adaptive immune responses. Science 346, 98–101.
T.H., Porta-Pardo, E., Gao, G.F., Plaisier, C.L., Eddy, J.A., et al. (2018). The
https://doi.org/10.1126/science.1254536.
immune landscape of cancer. Immunity 48, 812–830.e14. https://doi.org/
63. Mackay, L.K., Stock, A.T., Ma, J.Z., Jones, C.M., Kent, S.J., Mueller, S.N., 10.1016/j.immuni.2018.03.023.
Heath, W.R., Carbone, F.R., and Gebhardt, T. (2012). Long-lived epithelial
78. Kieffer, Y., Hocine, H.R., Gentric, G., Pelon, F., Bernard, C., Bourachot, B.,
immunity by tissue-resident memory T (TRM) cells in the absence of per-
Lameiras, S., Albergante, L., Bonneau, C., Guyard, A., et al. (2020). Single-
sisting local antigen presentation. Proc. Natl. Acad. Sci. USA 109, 7037–
cell analysis reveals fibroblast clusters linked to immunotherapy resis-
7042. https://doi.org/10.1073/pnas.1202288109.
tance in cancer. Cancer Discov. 10, 1330–1351. https://doi.org/10.1158/
64. Mijnheer, G., Lutter, L., Mokry, M., van der Wal, M., Scholman, R., 2159-8290.Cd-19-1384.
Fleskens, V., Pandit, A., Tao, W., Wekking, M., Vervoort, S., et al. (2021). 79. Dominguez, C.X., Mu €ller, S., Keerthivasan, S., Koeppen, H., Hung, J.,
Conserved human effector Treg cell transcriptomic and epigenetic signa- Gierke, S., Breart, B., Foreman, O., Bainbridge, T.W., Castiglioni, A.,
ture in arthritic joint inflammation. Nat. Commun. 12, 2710. https://doi.org/ et al. (2020). Single-cell RNA sequencing reveals stromal evolution into
10.1038/s41467-021-22975-7. LRRC15(+) myofibroblasts as a determinant of patient response to cancer
65. Shalapour, S., Lin, X.J., Bastian, I.N., Brain, J., Burt, A.D., Aksenov, A.A., immunotherapy. Cancer Discov. 10, 232–253. https://doi.org/10.1158/
Vrbanac, A.F., Li, W., Perkins, A., Matsutani, T., et al. (2017). Inflammation- 2159-8290.Cd-19-0644.
induced IgA+ cells dismantle anti-liver cancer immunity. Nature 551, 80. Hao, Y., Hao, S., Andersen-Nissen, E., Mauck, W.M., 3rd, Zheng, S.,
340–345. https://doi.org/10.1038/nature24302. Butler, A., Lee, M.J., Wilk, A.J., Darby, C., Zager, M., et al. (2021).
66. de Visser, K.E., Korets, L.V., and Coussens, L.M. (2005). De novo carcino- Integrated analysis of multimodal single-cell data. Cell 184, 3573–
genesis promoted by chronic inflammation is B lymphocyte dependent. 3587.e29. https://doi.org/10.1016/j.cell.2021.04.048.
Cancer Cell 7, 411–423. https://doi.org/10.1016/j.ccr.2005.04.014. 81. Korotkevich, G., Sukhov, V., Budin, N., Shpak, B., Artyomov, M.N., and
67. Roumenina, L.T., Daugan, M.V., Noé, R., Petitprez, F., Vano, Y.A., Sergushichev, A. (2019). Fast gene set enrichment analysis. Preprint at
Sanchez-Salas, R., Becht, E., Meilleroux, J., Clec’h, B.L., Giraldo, N.A., bioRxiv. https://doi.org/10.1101/060012.
et al. (2019). Tumor cells hijack macrophage-produced complement 82. Jin, S., Guerrero-Juarez, C.F., Zhang, L., Chang, I., Ramos, R., Kuan, C.H.,
C1q to promote tumor growth. Cancer Immunol. Res. 7, 1091–1105. Myung, P., Plikus, M.V., and Nie, Q. (2021). Inference and analysis of cell-
https://doi.org/10.1158/2326-6066.Cir-18-0891. cell communication using CellChat. Nat. Commun. 12, 1088. https://doi.
68. Pilette, C., Detry, B., Guisset, A., Gabriels, J., and Sibille, Y. (2010). org/10.1038/s41467-021-21246-9.
Induction of interleukin-10 expression through Fcalpha receptor in human 83. Qiu, X., Mao, Q., Tang, Y., Wang, L., Chawla, R., Pliner, H.A., and Trapnell,
monocytes and monocyte-derived dendritic cells: role of p38 MAPKinase. C. (2017). Reversed graph embedding resolves complex single-cell trajec-
Immunol. Cell Biol. 88, 486–493. https://doi.org/10.1038/icb.2009.120. tories. Nat. Methods 14, 979–982. https://doi.org/10.1038/nmeth.4402.
69. de Jonge, K., Tillé, L., Lourenco, J., Maby-El Hajjami, H., Nassiri, S., Racle, 84. Fawkner-Corbett, D., Antanaviciute, A., Parikh, K., Jagielowicz, M., Gerós,
J., Gfeller, D., Delorenzi, M., Verdeil, G., Baumgaertner, P., and Speiser, A.S., Gupta, T., Ashley, N., Khamis, D., Fowler, D., Morrissey, E., et al.
D.E. (2021). Inflammatory B cells correlate with failure to checkpoint (2021). Spatiotemporal analysis of human intestinal development at sin-
blockade in melanoma patients. OncoImmunology 10, 1873585. https:// gle-cell resolution. Cell 184, 810–826.e23. https://doi.org/10.1016/j.cell.
doi.org/10.1080/2162402x.2021.1873585. 2020.12.016.
70. Guo, X., Zhang, Y., Zheng, L., Zheng, C., Song, J., Zhang, Q., Kang, B., 85. Pijuan-Sala, B., Griffiths, J.A., Guibentif, C., Hiscock, T.W., Jawaid, W.,
Liu, Z., Jin, L., Xing, R., et al. (2018). Global characterization of T cells in Calero-Nieto, F.J., Mulas, C., Ibarra-Soria, X., Tyser, R.C.V., Ho, D.L.L.,
non-small-cell lung cancer by single-cell sequencing. Nat. Med. 24, et al. (2019). A single-cell molecular map of mouse gastrulation and early
978–985. https://doi.org/10.1038/s41591-018-0045-3. organogenesis. Nature 566, 490–495. https://doi.org/10.1038/s41586-
71. Thommen, D.S., Koelzer, V.H., Herzig, P., Roller, A., Trefny, M., Dimeloe, 019-0933-9.
S., Kiialainen, A., Hanhart, J., Schill, C., Hess, C., et al. (2018). A transcrip- 86. Squair, J.W., Gautier, M., Kathe, C., Anderson, M.A., James, N.D., Hutson,
tionally and functionally distinct PD-1(+) CD8(+) T cell pool with predictive T.H., Hudelle, R., Qaiser, T., Matson, K.J.E., Barraud, Q., et al. (2021).
potential in non-small-cell lung cancer treated with PD-1 blockade. Nat. Confronting false discoveries in single-cell differential expression. Nat.
Med. 24, 994–1004. https://doi.org/10.1038/s41591-018-0057-z. Commun. 12, 5692. https://doi.org/10.1038/s41467-021-25960-2.

Cancer Cell 41, 1–18, June 12, 2023 17

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article
87. Wu, T., Hu, E., Xu, S., Chen, M., Guo, P., Dai, Z., Feng, T., Zhou, L., Tang, 89. Liberzon, A., Birger, C., Thorvaldsdóttir, H., Ghandi, M., Mesirov, J.P., and
W., Zhan, L., et al. (2021). clusterProfiler 4.0: a universal enrichment tool for Tamayo, P. (2015). The Molecular Signatures Database (MSigDB) hallmark
interpreting omics data. Innovation 2, 100141. https://doi.org/10.1016/j. gene set collection. Cell Syst. 1, 417–425. https://doi.org/10.1016/j.cels.
xinn.2021.100141. 2015.12.004.
90. Bonneville, R., Krook, M.A., Kautto, E.A., Miya, J., Wing, M.R., Chen, H.Z.,
88. Whitfield, M.L., George, L.K., Grant, G.D., and Perou, C.M. (2006). Reeser, J.W., Yu, L., and Roychowdhury, S. (2017). Landscape of micro-
Common markers of proliferation. Nat. Rev. Cancer 6, 99–106. https:// satellite instability across 39 cancer types. JCO Precis. Oncol. 2017, 1–15.
doi.org/10.1038/nrc1802. https://doi.org/10.1200/po.17.00073.

18 Cancer Cell 41, 1–18, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Antibodies
Rabbit anti-human CD4 Abcam Cat# ab133616; RRID: AB_2750883
Mouse anti-human FOXP3 Abcam Cat# ab20034; RRID: AB_445284
Rabbit anti-human CTLA4 Abcam Cat# ab237712; RRID: AB_2905652
Rabbit anti-human TIM3 Abcam Cat# ab241332; RRID: AB_2888936
Rabbit anti-human CD68 Abcam Cat# ab213363; RRID: AB_2801637
Rabbit anti-human IL1B Novus Cat# NB600-633; RRID: AB_10001060
Rabbit anti-human CD14 Abcam Cat# ab133335; RRID: AB_2889158
Mouse anti-human VEGFA (VG-1) Abcam Cat# ab1316; RRID: AB_299738
Mouse anti-human KI-67 CST Cat# 9449; RRID: AB_2797703
Rabbit anti-human HLA-DR Abcam Cat# ab92511; RRID: AB_10563656
Rabbit anti-human CD3 Dako Cat# A0452; RRID: AB_2335677
Rabbit anti-human CD8 Abcam Cat# ab178089;
RRID: AB_2756374
Mouse anti-human CD20 Proteintech Cat# 60271-1-Ig; RRID: AB_2881391
Rabbit anti-human CD40 Abcam Cat# ab224639; RRID: AB_2883981
anti-mouse CD8 BioLegend Cat# 100712; RRID: AB_312751
anti-mouse CD45-FITC BioLegend Cat# 103108; RRID: AB_312973
anti-mouse CD45-PE/Cyanine7 BioLegend Cat# 103114; RRID: AB_312979
anti-mouse CD279 (PD-1) BioLegend Cat# 135218; RRID: AB_2561447
anti-mouse CD4 BioLegend Cat# 100412; RRID: AB_312697
anti-mouse CD40 BioLegend Cat# 124612; RRID: AB_1134072
anti-mouse CD19 BioLegend Cat# 115549;
RRID: AB_2563066
anti-mouse FOXP3 BioLegend Cat# 126406; RRID: AB_1089113
anti-mouse CD25 BioLegend Cat# 102008;
RRID: AB_312857
anti-human CD8 Zsbio Cat# ZA-0508; RRID: AB_2890107
anti-human CD68 Zsbio Cat# ZM-0060; RRID: AB_2904190
anti-human CD163 Zsbio Cat# ZM-0428; RRID: NA
anti-human CD57 Zsbio Cat# ZM-0058;
RRID: NA
anti-human PD-1 Zsbio Cat# ZM-0381;
RRID: AB_2921363
anti-human PD-L1 CST Cat# 13684;
RRID: AB_2687655
Biological samples
Human primary CRC samples The sixth affiliated hospital #IRB 2021ZSLYEC-173,
of Sun Yat-sen University See Table S1 for details
Human adjacent normal tissues The sixth affiliated hospital #IRB 2021ZSLYEC-173,
from CRC patients of Sun Yat-sen University See Table S1 for details
Human peripheral blood from CRC patients The sixth affiliated hospital #IRB 2021ZSLYEC-173
of Sun Yat-sen University
Paraffin-embedded formalin fixed slides The sixth affiliated hospital #IRB 2021ZSLYEC-173
prepared from colon biopsy of Sun Yat-sen University
and surgical samples
(Continued on next page)

Cancer Cell 41, 1–18.e1–e7, June 12, 2023 e1

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Critical commercial assays
Human XL Cytokine Luminex Luminex Cat# LKTM014
Performance Panel Premixed Kit
Sequencing: Chromium Single 10X Genomics Cat# PN-120237
Cell 30 Library & Gel Bead Kit v2
Sequencing: Chromium Single 10X Genomics Cat# PN-1000075
Cell 30 Library & Gel Bead Kit v3
Deposited data
Processed single cell RNA-seq data in This paper GEO: GSE205506
human dMMR/MSI-H CRC patients
TCGA COAD and READ datasets GDC data portal https://portal.gdc.cancer.gov/
Experimental models: Cell lines
MC38 BMCR/NICR Cat# 1101MOU-PUMC000523
Experimental models: Organisms/strains
C57/BL6N mice Jiangsu GemPharmatech C57BL/6NGrl
Software and algorithms
Cell Ranger-4.1.0 10 X Genomics http://10xgenomics.com/
R-4.1.0 The R Foundation https://www.r-roject.org/
Seurat-4.1.0 Hao et al.80 https://satijalab.org/seurat/
fgsea (v1.18.0, R package) Korotkevich et al.81 https://bioconductor.org/packages/
release/bioc/html/fgsea.html
CellChat (version 1.1.3) Jin et al.82 https://github.com/sqjin/CellChat
Monocle2 (v2.10.1) Qiu et al.83 https://github.com/cole-trapnelllab/
monocle-release

RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Yanhong
Deng ([email protected]).

Materials availability
This study did not generate new unique reagents.

Data and code availability

Processed single cell RNA-seq data of this study can be obtain from Gene Expression Omnibus (GEO) with an accession number of
GSE205506. The raw FASTQ files in this study will be provided for scientific research upon request complying with the law due to
human patient privacy concerns. Code used for all processing and analysis is available upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Patient population and clinical study

Patient cohorts were from a randomized phase 2 PICC study (NCT03926338), with locally advanced, primary invasive carcinoma of
the d-MMR/MSI-H colon and rectum treated with toripalimab with or without celecoxib for six cycles before curative surgical resec-
tion.9 The primary goal of the study was to investigated the efficacy and safety of PD-1 blockade with toripalimab, with or without the
COX-2 inhibitor celecoxib, as neoadjuvant treatment for d-MMR/MSI-H locally advanced CRCs. The study enrolled 34 patients be-
tween May 1, 2019, and April 1, 2021. In both treatment groups, toripalimab 3 mg/kg (Junshi BioSciences, Shanghai, China) was
given intravenously over 30 min on day 1 of each 14-day cycle, for six cycles, before surgical resection. Patients in the toripalimab
plus celecoxib group were also given celecoxib 200 mg (Pfizer, New York, NY, USA) orally twice daily from day 1 to 14 each 14-day
cycle. Surgery was planned for within 4 weeks after the last neoadjuvant toripalimab dose. Pathological complete response was
defined as tumours without any viable tumour cells in the resected primary tumour sample and all sampled regional lymph nodes.
Primary tumor tissue and adjacent normal tissue, serum samples before and/or after ICI treatment, as well as paraffin-embedded
formalin fixed (FFPE) slides prepared from enteroscope biopsy and surgical samples were collected. The samples before ICI treatment
were obtained after the initial diagnosis without immunotherapy, while the serum samples after ICI treatment were collected before

e2 Cancer Cell 41, 1–18.e1–e7, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

surgical resection, and the tissue samples after ICI treatment were from radical surgical resection. Overall, we collected 40 samples
from 19 patients for single cell RNA profiling (See Table S1 for details) and 74 serum samples from 27 patients for serum cytokine/che-
mokine quantification. In addition, we obtained 90 FFPE slides from 23 patients for multiplex immunofluorescence staining to verify the
results of scRNA sequencing. Patients are all from the prospective PICC study and the expanded PICC study, 6 of which have partic-
ipated in the scRNA sequencing in this study. In general, 90 samples were collected, of which 42 were - ICI (27 pCR, 15 non-pCR) and
48 were +ICI (38 pCR, 10 non-pCR). These samples were distributed into 4 panels for multiplex fluorescence immunohistochemical
staining. All patients were provided written informed consent for the collection of tissue and blood samples for research and genomic
profiling, as approved by The Sixth Affiliated Hospital of Sun Yat-sen University Review Board (#IRB 2021ZSLYEC-173).

In vivo mouse models

Male C57/BL6N mice aged 5–6 weeks were purchased from Jiangsu GemPharmatech and quarantined for one week before use.
Animal care and experiments involved in this study were performed in accordance with Accreditation of Laboratory Animal Care
International guidelines. Animal experiment protocols were approved by the guidelines established by the Animal Care Committee
of the Sixth Affiliated Hospital, Sun Yat-sen University. 5 3 105 MC38 cells were suspended in 50 mL of normal saline (NS) and
subcutaneously injected. 500 mg/kg IL1b protein (HY-P7073A, MedChemExpress, USA) was intratumorally injected every day
(five injections total). One day after the final injection, the tumor tissues were collected and analyzed by flow cytometry.

METHOD DETAILS

Sample collection and dissociation for scRNA-seq

All sample collection procedures complied with the regular routine in clinical practice. Tumor and adjacent normal samples were first
stored in liquid nitrogen with the SenotechTM Tissue Storage Buffer (#JZ-SC-5802, Senotech Genomics, Shanghai, China) or pro-
ceed directly to tissue disassociation. The frozen tissue was heated in a 37 C water bath, and then was transferred to a new 15 mL
centrifuge tube together with the frozen storage solution. 5 mL of preheated RPMI-1640 medium (GIBCO) was added to the centri-
fuge tube, which was further centrifuged at room temperature at 3003g for 5 min to remove the supernatant without disturbing the
tissue sample at the bottom of the centrifuge tube. According to the manufacturer’s protocol, the tissue samples were enzymolysis
with Thermo Scientific at 37 C for 20–30 min. The isolated cells were then passed through a 40 mm cell filter (BD) containing 10% FBS
in RPMI-1640 medium (Invitrogen) until a uniform cell suspension was obtained. Subsequently, the suspended cells were passed
through a cell filter and centrifuged at 400 3g for 10 min. Red blood cells were removed according to the manufacturer’s instruction.

Single-cell RNA-seq library preparation and sequencing

Single-cell RNA-seq libraries were prepared using the Chromium Next GEM Single Cell 30 Kit from 10x Genomics, following the man-
ufacturer’s instructions. In brief, single cells were resuspended in PBS containing 0.04% BSA to a final concentration of 500–1,200
cells/mL as determined by Countess 3 Automated Cell Counters (Invitrogen). 10,000 cells were captured in droplets to generate
nanoliter-scale Gel bead in EMulsion (GEMs). GEMs were then reverse transcribed in a C1000 Touch Thermal Cycler (Bio-Rad)
programmed at 53 C for 45 min, 85 C for 5 min, and held at 4 C. After reverse transcription and cell barcoding, emulsions were
broken and cDNA was isolated and purified with Cleanup Mix containing DynaBeads and SPRIselect reagent (Thermo Fisher Scien-
tific), followed by PCR amplification. Amplified cDNA was fragmented, end-repaired, double-sided size-selected, PCR-amplified
with sample indexing primers, and subjected to library construction. Libraries prepared according to the manufacturer’s user guide
were then purified and profiled for quality assessment. Single-cell RNA libraries were sequenced by an Illumina NovaSeq 6000 or
BGISEQ DNBSEQ-T7 sequencer with 150 bp paired-end reads.

Multiplex fluorescence immunohistochemistry

For multiplex fluorescence immunohistochemical staining, formalin-fixed and paraffin-embedded (FFPE) tumor tissue blocks were
serially sectioned into 4–8 mm sections. Multiplex immunofluorescence staining was conducted using the Akoya OPAL Polaris
7-Color Automation IHC kit (NEL871001KT) according to the manufacturer’s protocol. Briefly, FFPE tissue slides were first deparaffi-
nized in a BOND RX system (Leica Biosystems) and then incubated sequentially with primary antibodies CD4 (Abcam, Cat#ab133616,
1:100), FOXP3 (Abcam, Cat#ab20034, 1:100), CTLA4(Abcam, Cat#ab237712, 1:500), TIM3(Abcam, Cat#ab241332, 1:1000), CD68
(Abcam, Cat#ab213363, 1:1000), IL1B(NOVUS, Cat#NB600-633, 1:400), CD14(Abcam, Cat#ab133335, 1:500), VEGFA(Abcam,
Cat#ab1316, 1:600), MKI67(CST, Cat#9449, 1:400), HLA-DR(Abcam, Cat#ab92511, 1:500), CD3 (Dako, Cat#A0452, 1:100), CD8
(Abcam, Cat#ab178089, 1:100), CD40 (Abcam, Cat#ab224639, 1:100), CD20 (Proteintech, Cat#60271-1-Ig, 1:400), CD68 (Zsbio,
Cat#ZM0060, 1:500), CD163 (Zsbio, Cat#ZM0428, 1:100), PD-1 (Zsbio, Cat#ZM0381, 1:50) and PD-L1 (CST, Cat#13684, 1:100),
CD8(Zsbio, Cat#ZA-0508, clone SP16, 1:100) and CD57 (Zsbio, Cat#ZM0058, 1:100). This was followed by incubation with secondary
antibodies (Vector Laboratories) and corresponding reactive Opal fluorophores. Nuclei acids were stained with DAPI (Akoya,
Cat#210514003). Tissue slides that were bound with primary and secondary antibodies but not fluorophores were included as negative
controls to assess autofluorescence. Multiplex stained slides were scanned using a Vectra Polaris Quantitative Pathology Imaging
System (Akoya Biosciences) at 20 nm wavelength intervals from 440 nm to 780 nm with a fixed exposure time and an absolute magni-
fication of 3200. All scans for each slide were then superimposed to obtain a single image. Multilayer images were imported to inForm

Cancer Cell 41, 1–18.e1–e7, June 12, 2023 e3

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

v.2.4.8 (Akoya Biosciences) for quantitative image analysis. The quantities of various cell populations were expressed as the number of
stained cells per square millimeter and further as the percentage of positively stained cells.

Luminex-based cytokine/chemokine detection for serum samples

Magnetic Luminex Assay kits were purchased from R&D Systems (catalog number: LKTM014). The tested cytokines/chemokines
were as follows: CCL2 (MCP-1), CCL3 (MIP-1a), CCL4 (MIP-1b), CCL5 (RANTES), CCL11 (Eotaxin), CCL20 (MIP-3a), CCL19
(MIP-3b), CX3CL1 (Fractalkine), CXCL2 (GRO-b), CXCL10 (IP-10/GRO-a), IL-1b, IL-1ra, IL-2, IL-4, IL-6, IL-7, IL-8, IL-10, IL-12
p70, IL-13, IL-15, IL-33, VEGF, PDGF-AA, PDGF-AB/BB, TGF-a, G-CSF, GM-CSF, PD-L1 (B7-H1), CD40 Ligand, FLT-3Ligand,
IFN-a, IFN-g, TNF-a, TRAIL, Granzyme B, and EGF. Serum samples cryopreserved at
80 C were thawed and analyzed according
to the manufacturers’ instructions. Data was acquired on a calibrated and validated Luminex MAGPIX instrument (R&D Systems,
Abingdon, UK) as per manufacturer’s instructions.

Tissue digestion and flow cytometry analysis for mouse samples

The tumor tissues for flow cytometry were cut into small pieces and digested with 1 mg/mL collagenase type IV (Sigma, USA; Cat#C5138)
and 0.6 ku mL^-1 DNAse (Sigma, USA; Cat#D5025) for 2.5h. Samples were then filtrated to single-cell suspension. Mouse 13 lympho-
cyte separation medium (DAKEWE, China; Cat#DKW33-R0100) was used to enrich tumor-infiltrating lymphocytes. Cells were stained
with CD45-FITC (Biolegend, Cat#103108, 1:100), CD45-PE/Cyanine7(Biolegend, Cat#103114, 1:100), CD8-APC(Biolegend,
Cat#100712, 1:50), PD1-Brilliant Violet 421(Biolegend, Cat#135218, 1:100), CD4-APC(Biolegend, Cat#100412, 1:100), CD25-PE
(Biolegend, Cat#102008, 1:100), CD40-APC(Biolegend, Cat#124612, 1:100), CD19-Brilliant Violet 421(Biolegend, Cat#115549,
1:100). For intracellular staining, cells were proceeded with Foxp3/Transcription Factor Staining Buffer Set (eBioscience, Cat#00-
5523-00) according to the manufacturers instruction and stained with FOXP3-Alexa Fluor 488(Biolegend, Cat#126406, 1:100). Subse-
quently, cells were analyzed by flow cytometry. The results were analyzed by CytExpert.

Serum cytokine/chemokine data analysis

Before Z score standardization of quantitative data (pg/mol), we set the value below the detection line to 0 and the value above the
detection upper limit to the value of high concentration control. Normalized data were analyzed using R (version 4.1.0) software. Sig-
nificant variations between the groups were analyzed using Wilcoxon test. p values of less than 0.05 were considered statistically
significant.

Pre-processing of single cell RNA-seq data

Raw single-cell RNA-seq data were processed using 10X Genomics Cell Ranger toolkit (version 4.1.0), including demultiplexing the
FASTQ reads, aligning them to the human reference genome (GRCh38, v3.0.0, from 10X Genomics), and generating the gene-cell
unique molecular identifier (UMI) matrix (by using ‘‘cellranger count’’ function). Quality control and integration were performed using
Seurat v4 R package.80 We carried out several steps to filter out poor quality data. First, genes covered by less than 3 cells were
filtered out. Then, all cells expressing <500 or >5,000 genes were removed, as well as cells that contained <400 or >25000 unique
molecular identifiers (UMIs) to filter out the most of barcodes associated with empty partitions or doublet cells. On average, we de-
tected 5,011 cells and 19,186 genes per sample after filtering. Cells with high mitochondrial gene expression are removed according
to the broad cell type identified after batch effect correction and unsupervised clustering, as described below. To integrate and
embed single cells from different individuals into a shared low-dimension space, we utilized integrated analysis (RPCA) by the Seurat
v4 function ‘‘IntegrateData’’ to perform batch effect correction and normalization. The new integrated matrix was obtained and was
used only for clustering and cell type classification.

Unsupervised clustering analysis and identification of broad cell type

After generation of the integrated matrix, we used an unsupervised graph-based clustering algorithm to cluster single cells by their
expression implemented in Seurat v4.80 The default parameters of Seurat were used, unless mentioned otherwise. Briefly, the UMI
count matrix was normalized by using ‘NormalizeData’ function with default parameters. Based on the natural-log transformed
normalized gene expression matrix, 2000 highly variable genes were generated using the ‘FindVariableFeatures’ function with the
‘vst’ method. For the clustering of all cell types, 2,000 variable genes were identified and 20 principal component analysis (PCA)
was applied to the dataset to reduce dimensionality after regressing for the number of UMIs (counts). We used the function
‘‘FindClusters’’ on 20 PCs with resolution 1.2 to perform the first-round cluster and annotated each cluster by known markers (T cells:
CD3D, CD3E, TRAC, TRBC1; B cells: CD79A, CD79B, MS4A1, TNFRSF17, MZB1; Myeloid cells: CD14, CD68; Epithelial cells: EP-
CAM, CD24; Fibroblast: COL1A2, COL3A1, MYH11, ACTA2; Endothelial cells: VWF, PECAM1). Then cells with high expression level
of mitochondrial genes in each cell type were then filter out, as descried below. The remaining 155,397 cells were used to perform
re-clustering using Seurat with the same parameters for visualization.

Removing cells with high expression level of mitochondrial genes

Furthermore, cells with unusually high detection rate of mitochondrial gene expression were excluded. Since mitochondrial RNA con-
tent are highly cell type dependent (notably higher in epithelial cells).84 Thresholds were derived individually for cells within each
compartment following an initial clustering. We first using several well-known markers to identify the cell types of the clusters and

e4 Cancer Cell 41, 1–18.e1–e7, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

combined the clusters belong to same cell type after removing extremely low abundance and conflicting clusters that co-express
multiple broad markers (2.41%). Epithelial cells were removed if their proportions of mitochondrial gene expression were larger
than 75% (29.75% of epithelial cells). Then, we fit the expression level of mitochondrial genes by using a median-centered median
absolute deviation (MAD)-variance normal distribution,73,85 and then removed the cells with significantly higher expression levels
than expected (determined by Bonferroni-corrected p < 0.05, 9.20% for lymphoid immune cells including T cells, B cells; 12.84%
for myeloid cells; 8.11% for fibroblasts and 8.50% for endothelial cells). The remaining 155,397 cells were used to perform re-clus-
tering using Seurat. As a result, high qualitied cells were clustered into 6 groups: epithelial cells, fibroblasts, endothelial cells, T cells,
B cells and myeloid cells.

Dimensionality reduction using UMAP

For visualization, the dimensionality of each dataset was further reduced using the Uniform Manifold Approximation and Projection
(UMAP) with Seurat functions ‘‘RunUMAP’’. The PCs used to calculate the embedding were as the same as those used for clustering.

Re-clustering of broad cell types

Using T cell markers (CD3D, CD3E, CD2), we identified a total of 29,033 cells, which were annotated as T cells. Re-cluster analysis
found that these T cells were mixed with NK cells. Clusters of cells with high mitochondrial gene expression and hybrid cells (10%)
were excluded in the subsequent analysis. The remaining cells (n = 25,986) revealed well-defined subpopulations that were anno-
tated based on published signatures, including the major T cell lineages (CD4+ T cells and CD8+ T cells) and three subsets of innate
immune cells among including gd T cells expressing gd receptor genes (TRDC, TRGC1 and TRGC2), NK cells with CD3 negative and
overexpressed NK cell genes (KLRD1, KLRC1, FCGR3A, TYROBP), as well as a small amount of innate lymphocyte expressing KIT,
and KRT86.84 (Figure S2C, Table S2) In addition, proliferating T cells were shown as a mixture of CD4+ and CD8+ T cells. We further
classified them according to the normalized expression of CD4 and CD8A (CD4 Expr. > CD8A Expr., CD4+ Prolif. T cells, CD4 Expr. %
CD8A Expr., CD8+ Trm-mitotic cells).
Using B cell markers (MS4A1, CD79A, CD79B), we identified a total of 17,008 B cells to reveal 7 phenotypes after removing cells
with high mitochondrial or ribosome gene expression (16%). Cells were first classified into plasma cells (MZB1, XBP1, TNFRSF17A,
IGHA1, IGHD1) and CD20+ B cells (MS4A1, CD79A, CD79B), and further annotated according to marker genes (Figure S2D,
Table S2).
Using myeloid cell markers (CD14), we identified a total of 7,282 myeloid cells to reveal 10 clusters, including dendritic cell (DC)
subsets, macrophage subsets and monocyte-like subsets (Figures S2E and S5A) after removing cells with high mitochondrial
gene expression and hybrid cells (14%). DC subsets including plasmacytoid DCs (pDC, LILR4A+ pDC), migratory DCs (CCL19+
mDC) and conventional DCs (CD1C+ cDC), which are characterized by high expression of HLA genes and low expression of
CD14. Clusters were identified as macrophages based on their high expression of CD68 or CD163 and negative expression of mono-
cyte markers (FCN1, S100A8, S100A12). Cells express monocyte signature genes (FCN1, S100A8, S100A12) without the expression
of dendritic and macrophage markers were annotated as mono-like cells (FCN1+ Mono, IL1B+ Mono).
Using endothelial cell markers (VWF, PECAM1), we identified a total of 13,376 endothelial cells to reveal 6 phenotypes after
removing hybrid cells (5.8%). Cells were classified into lymphatic endothelial cells (LEC) by expressing LYVE1, CCL21, PROX1 genes
and high endothelial venule (HEV) by expressing MADCAM1, ACKR1 genes or annotated according to marker genes (Figure S2F).
Using fibroblast cell markers (COL1A2, COL3A1, MYH11, ACTA2), we identified a total of 7,326 fibroblast cells to reveal three phe-
notypes after removing hybrid cells (7.7%). Cells were classified into myofibroblast by expressing MYH11, ACTA2, MCAM, CAV1
genes or annotated according to marker genes (Figure S2G).

Identification of marker genes

To characterize each cluster, we applied the ‘‘FindAllMarkers’’ procedure in Seurat which identified markers using log fold changes
(FC) of mean expression, and used Wilcoxon Rank-Sum test by default (min.pct = 0.25, logfc.threshold = 0.25). Marker gene list were
show in Table S2.

Differential expression and functional annotation

Differential expression genes (DEGs) were identified using combination of ‘‘FindMarkers’’ function implemented in the Seurat pack-
age and pseudobulk method (‘Libra’ package).86 Genes that met the criteria of adjusted P-value <0.05 in Seurat method and
P-value < 0.01 in pseudobulk method were considered as significant DEGs. Gene Ontology and pathway enrichment analyses of
DEGs were performed using the ‘clusterProfiler’ R package.87 Annotation Dbi R package ‘‘org.Hs.eg.db’’ was used to map gene
identifiers. In each case, gene sets were tested for over-representation in cluster markers or DEGs by computing enrichment p values
(‘enricher’ R function, default parameters) from the hypergeometric distribution. Hypergeometric p value was adjusted in each case
for multiple testing using Benjamini-Hochberg correction. The results were visualized as bar plots using ‘clusterProfiler’ and ‘ggplot2’
R packages.

Definition of gene signature score

For computation of signature scores, average expression of the signature genes was counted based on the normalized matrix gener-
ating by Seurat Package. Cytotoxicity, exhaustion, and Treg score were defined by the average expression of published signature

Cancer Cell 41, 1–18.e1–e7, June 12, 2023 e5

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

gene according to the gene lists from Zheng et,al.17 Proliferation score was defined by the average expression of known proliferation-
related genes include AURKA, BUB1, CCNB1, CCND1, CCNE1, DEK, E2F1, FEN1, FOXM1, H2AFZ, HMGB2, MCM2, MCM3,
MCM4, MCM5, MCM6, MKI67, MYBL2, PCNA, PLK1, TOP2A, TYMS and ZWINT.88 Acute inflammatory response and chronic
inflammatory response scores were defined by the average expression of genes annotated in ‘‘GOBP_CHRONIC_INFLAMMATOR-
Y_RESPONSE.v7.5.1 (GO:0002544)’’ and ‘‘GOBP_ACUTE_INFLAMMATORY_RESPONSE.v7.5.1 (GO:0002526)’’ from MSigDB
gene sets.89 Gene list to define CD8+ T cell signatures according to previous literatures were provided in Table S3.

Gene set enrichment analysis

We used a fast per-ranked gene set enrichment analysis (GSEA) named fgsea (v1.18.0, R package)81 to perform functional enrich-
ment analysis for defined gene signatures and previously reported gene signatures of cluster-specific DEGs (CD8 subset-A vs other
CD8+ T cells, IL1B+ Mono vs other myeloid cells) and ICI treatment associated DEGs (+ICI/pCR vs -ICI, +ICI/non-pCR vs -ICI). First,
DEGs were identified using the fast Wilcoxon rank sum test with ‘wilcoxauc’ functions implemented in ‘presto’ R package (https://
github.com/immunogenomics/presto) and sorted according to AUC value, and then we fed each differentially expressed gene list
together with AUC value of each gene and the gene sets to the ‘fgsea’ function (nperm = 1000). Gene sets with adjusted P value < 0.05
were selected as the functionally enriched biological states or processes. Positive normalize enrichment score (NES) means that a
certain functional gene set is enriched in front of the sorting sequence, indicating upregulation in the specific cell population
compared with other cells. While on the contrary, negative NES means certain functional gene set is enriched behind the sorting
sequence, indicating down regulation in the specific cell population compared with other cells.

Correlation analysis
For single-cell based correlation analysis, we first calculated the cellular proportions of each immune cell cluster by their fractions in
corresponding major immune compartments, which involve T/I/NK cell, B cell and myeloid cell compartments for each patient. We
then evaluated the correlations of immune cell clusters by Pearson correlation test in the entire patient cohort (n=40).

Cell-to-cell communication of scRNA-seq data

We used CellChat (Version 1.1.3)82 to infer the cell-cell interactions across immune and stromal cell types using the expression of
known ligand-receptor pairs. All databases curated in CellChatDB.human database including the "Secreted Signaling", "Cell-Cell
contact" and "ECM-Receptor" were included. Cells were annotated according to their cell type and sample groups: "-ICI", "+ICI/
pCR", "+ICI/non-pCR". To identify the changes of the cell-cell communication networks induced by ICI treatment in pCR response
tumors, we followed the tutorial for comparison analysis of multiple datasets in the CellChat github repository (https://htmlpreview.
github.io/?https://github.com/sqjin/CellChat/blob/master/tutorial/Comparison_analysis_of_multiple_datasets.html) and computed
the major signaling changes between -ICI and +ICI/pCR groups by joint manifold learning and quantitative contrasts of multiple
cell-cell communication networks. We compared the outgoing and incoming interaction strength of each pair of cell types to identify
significant changes in sending or receiving signals between groups, and used the interaction strength to qualify the ICI treatment
effect. We then visualized the most significant ligand-receptor pairs that mediate the cell-cell interaction changes across compari-
sons by using the netVisual_bubble function in CellChat.

Cell developmental trajectory

The cell lineage trajectory of CD8+ T and myeloid cells was inferred by using Monocle2.83 We used the ‘‘differentialGeneTest’’ func-
tion to derive top 500 DEGs from each cluster and genes with a q-value < 0.05 were used to order the cells in pseudotime analysis.
After the cell trajectories were constructed, differentially expressed genes along the pseudotime were detected using the ‘‘differen-
tialGeneTest’’ function.

Cell subtype abundance estimated from TCGA bulk gene expression profiles
We integrated sc-RNA data and TCGA CRC bulk expression profiles to verified the correlation between cell subsets and function-
alities in d-MMR/MSI-H CRC. We identified immune and stromal cell subtypes in human d-MMR/MSI-H CRC in this study (Figure 1E),
and their specific gene signatures were derived marker genes of each subtypes identified using ‘‘FindAllMarkers’’ procedure in
Seurat (Marker gene list were show in Table S2) based on following criteria: 1) FDR adjusted P-value <0.05; 2) log2FC > 0.58. The
TCGA-COAD and TCGA-READ data were used to evaluate the abundance of each cluster. The gene expression from these bulk
RNA-seq datasets (‘HTSeq – FPKM’) were downloaded from TCGA data portal (https://portal.gdc.cancer.gov/). MSI classification
results are derived from previously published study.90 Subsequently, we estimated the relative abundance of each cell subtype
by the average expression of Z score normalized log-transformed expression of the cell type specific genes defined above (Table S4).

Correlative cell-cell interactions inferred by combined scRNA-seq and TCGA MSI-H CRC datasets
To inferred the correlative cell-cell interactions between immune cell subtypes by combined scRNA-seq and TCGA datasets, we
applied methods from Zhang et al.38 Briefly, cell subtype abundances were first estimated from TCGA MSI-H CRC bulk expression
profiles as described above. Then we computed the Pearson correlation coefficient between the expression of each gene and the
relative abundance of each cell subtype based on the bulk RNA-seq profiles and obtained a correlation matrix between genes
and abundances of cell subtypes. Since the expression levels of certain signature genes in a given cell subtype were highly correlated

e6 Cancer Cell 41, 1–18.e1–e7, June 12, 2023

 Please cite this article in press as: Li et al., Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient
colorectal cancer, Cancer Cell (2023), https://doi.org/10.1016/j.ccell.2023.04.011

ll
Article

with the abundance of this specific cell subtype, we defined these genes as ‘‘self-expressed genes’’ based on our sc-RNA dataset
according to the following criteria: 1) FDR adjusted p-value <0.05; 2) log2FC > 0.58; 3) cell frequency of expression >35%. We then
filtered these genes and transformed their correlation values to 0 to obtain an adjusted correlation matrix. Finally, for a specific cell
subtype, to identify which cell subtypes could contribute to the highly correlated non-self-expressed genes, we performed a gene set
enrichment analysis based on our sc-RNA dataset. In detail, we calculated the mean expression of each gene across each cell
subtype and then performed Z score transformation. Next, the top 13 highly correlated non-self-expressed genes were selected
from the adjusted correlation matrix. We then calculated the mean value of Z score transformed expression of these genes in
each cell subtype as the enrichment score. Finally, the correlated cell subtype(s) was identified as the Z score transformed enrich-
ment score >2.576. The correlative networks were built separately with corresponding correlated cell subtypes (Figure S6F).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed as described in Figure legends. Pearson correlation was used to estimate correlations among
immune cell subsets. Statistical significance was determined by Kruskal-Wallis test, Wilcoxon test, Student T test and hypergeomet-
rical test with the Benjamini-Hochberg method for multiple comparisons correction.

Cancer Cell 41, 1–18.e1–e7, June 12, 2023 e7