High-Baseline Tumor Burden–Associated Macrophages Promote Immunosuppressive Microenvironment and Restrain Immune Checkpoint Inhibitor Ecacy Through the IGFBP2–STAT3–PD-L1 Pathway

Background: In recent decades, immune checkpoint inhibitors (ICIs) have become a conspicuous promising treatment. However, there are still some problems such as limited effective rate and undened suitable patients. With increasing attention to the inuence of baseline tumor burden on the therapeutic ecacy of ICIs, most researchers are currently of the view that patients with high-baseline tumor burden (HTB) have shorter overall survival and poor immunotherapy ecacy; nevertheless, recent studies have been limited to clinical phenomena without deep mechanistic exploration. Methods: RNA-seq and Single-cell RNA sequencing were used to reveal the difference of microenvironment between low-baseline tumor burden (LTB) and HTB tumor. Mice model construction, ow cytometry and cytokine antibody array were used to identify the phenomenon and mechanism of HTB effects on tumor immune microenvironment (TIME) and ICIs effcacy. Results: HTB caused a signicantly higher inltration with M2-type macrophages expressing a high level of programmed death-ligand 1 (PD-L1) to block the inltration by CD8 + T cells and impaired the therapeutic effect of ICIs. HTB-derived IGFBP2 induced the macrophages’ polarization and PD-L1 expression via signal transducer and activator of transcription 3 (STAT3) signaling pathway. Moreover, these HTB–activated macrophages exerted a pro-tumor effect by inhibiting proliferation and cytotoxic function of CD8 + T cells in a PD-L1-dependent fashion. Disappointingly, total tumor burden did not affect the microenvironment or the immunotherapy ecacy of any single tumor. Moreover, macrophage inhibitor in HTB and palliative surgery were not able to increase the therapeutic ecacy in HTB. Only in the early stage of tumor with LTB, the macrophage inhibitor PLX3397 signicantly improved the ecacy of ICIs on HTB. Conclusions: Our study has demonstrated that HTB induced the production of PD-L1 + M2-macrophages through the IGFBP2–STAT3 signaling pathway to generate the suppressive TIME, thereby reducing


Background
In recent decades, immune checkpoint inhibitors (ICIs) have brought a new chapter in cancer treatment.
Above all, therapies based on ICIs have achieved remarkable clinical outcomes and have revolutionized the treatment of various tumors, particularly of those in advanced stage. Compared with traditional cancer therapies, ICIs mainly restore the ability of effector T cells to resist tumor cells by blocking immunosuppressive receptors such as PD1/PDL1 and CTLA4; thus, ICIs have shown a wide range of antitumor biological activities and have achieved long-term remissions [1]. Despite the breakthrough success of ICIs, their effective rate is only 10%-30%; therefore, improving the therapeutic e cacy of ICIs is an urgent problem to be solved in clinical practice.
Effects of the location of tumor metastasis and the baseline tumor burden on the therapeutic e cacy of ICIs have also gradually received research attention. The concept of baseline tumor burden, simply de ned as the size of the tumor or the total number of cancer cells before treatment, can easily be assessed by imaging. Historically, baseline tumor burden has been considered an impediment to the e cacy of ICIs [2,3]. Patients with high-baseline tumor burden (HTB) have shorter overall survival and progression-free survival among non-small cell lung cancer (NSCLC) patients treated with atezolizumab [3]. The KEYNOTE-001 study has pointed out that HTB is associated with a lower objective response rate and decreased overall survival of patients with melanoma treated with pembrolizumab [2].
However, current research on this aspect has been limited to several clinical studies, and there is a lack of in-depth mechanistic exploration.
Tumor immune microenvironment (TIME) is another important factor that affects the e cacy of ICIs.
According to the in ltration by immune cells, especially T cells, the TIME can be divided into the following three subtypes: in amed, immune excluded, and immune desert [4]. The latter two types of tumors respond poorly to ICIs. Previous studies on the correlation between baseline tumor burden and the degree of T lymphocyte in ltration in the TIME are scarce and contradictory. The numbers of CD4 + and CD8 + T cells in peripheral blood of gastric cancer patients with low-baseline tumor burden (LTB) were higher than those of patients with HTB [5]. In diffuse large B lymphoma, there is a lower in ltration degree by Ki-67 + T cells in the HTB microenvironment [6]. In contrast, in ductal breast cancer, cytotoxic T lymphocytes (CTLs) in ltration was signi cantly increased in tumors with HTB [7]. Noteworthily, there have been no research reports that baseline tumor burden can affect tumor response to ICI therapy by in uencing the TIME.
Herein, we found that HTB was able to impair the e cacy of anti-PD-1, and the number of CD8 + T cells was reduced in the HTB microenvironment. Single-cell sequencing revealed that activated macrophages with an immunosuppressive phenotype were highly enriched within HTB melanoma. The speci c mechanism was likely as follows: HTB microenvironment prolonged the lifespan of macrophages, and released large amounts of insulin-like growth factor binding protein 2 (IGFBP2) to induce the macrophages' polarization and PD-L1 expression by activating STAT3. These PD-L1 + M2 macrophages inhibited CD8 + T cells in a PD-L1-dependent fashion, thereby reducing the response of HTB cancer to ICI therapy. To our surprise, palliative surgery was not able to improve the e cacy of ICIs. However, PLX3397-based inhibition of macrophages in the early stage of tumor effectively improved the sensitivity of HTB tumors to ICIs after tumor progression.

Patients and specimens
This study was approved by the Nanfang Hospital Ethics Review Board. A total of 51 para n-embedded samples from patients with colon cancer were collected. According to the medium tumour size, patients were divided into 26 LTB (average TB of 13.00 cm³) and 25 HTB (average TB of 88.11 cm³). Metastatic tumor needle biopsy samples of 2 patients with colon cancer liver metastasis were collected. The patients were all histologically diagnosed with colon cancer at Nanfang Hospital, Southern Medical University (Guangzhou, China). We also downloaded 1 colon cancer datasets with clinical information from TCGA (COAD) and 1 melanoma datasets with clinical information from TCGA (SKCM) (https://portal.gdc.cancer.gov/). Cell culture MC38 colon cancer cell line was grown in 1640 RPMI (Gibco) supplemented with 10% FBS (Hyclone), 2 mM glutamine and 20 μg/ml gentamicin. B16 melanoma cell line was grown in high-glucose DMEM (Hyclone) supplemented with 10% FBS (Hyclone), 2 mM glutamine and 20 μg/ml gentamicin. RAW 264.7 cell line was grown in high-glucose DMEM (Hyclone) supplemented with 10% FBS (Hyclone), 2mM glutamine and 20μg/ml gentamicin. All cells were grown at 37 ℃ in a humidi ed atmosphere with 5% CO 2 . All cell lines were authenticated by the STR (Short Tandem Repeat) pro ling.  Immunohistochemistry (IHC) staining IHC was performed on 4 mm sections of formalin-xed para n-embedded (FFPE) tissues. Para n sections were baked at 65 ℃ for 4 hours. Sections were depara nized, dehydrated through a graded ethanol series. Then sections were repaired in 0.01 M and pH 6.0 citrate buffer. Endogenous peroxidase was blocked by incubation in 0.3% H 2 O 2 for 10 min. After incubating with 5% BSA for 1 hour at room temperature, sections were incubated with appropriate primary antibodies overnight at 4 ℃. Then sections were incubated with relevant secondary antibody for 1 hour at 37 ℃. Sections were incubate with a DAB substrate kit. CD8 protein IOD were calculated by image pro plus 6.0.

Preparation of LTCS and HTCS and supernatant-conditioned macrophages
LTB-tumor tissue culture supernatants (LTCS) or HTB-tumor tissue culture supernatants (HTCS) were prepared by plating tumor tissues in 1 mL DMEM for 24 hours.The supernatant was then centrifuged and harvested. To generate supernatant-conditioned macrophages, RAW 264.7 were rst harvested and cultured with 50% LTCS or HTCS for 24 hours, and then washed with DMEM for three times. RAW 264.7 cultured with DMEM were used as controls.

Flow cytometry
Cells were resuspended to achieve a cell concentration of 1×10 7 / mL. 100 μL cells were incubated with appropriate primary antibodies at 4 ℃ for 30 min. Cells were repeatedly washed twice using PBS, then cells were detected by ow cytometer. Raw data were analysed through the FlowJo_V10 software.
RNA-seq analysis of MC38 subcutaneous tumour cells Total RNA in the tissue samples was extracted using Trizol. The mRNA was isolated and puri ed from the total RNA by using the Oligo (dT) magnetic beads. First-strand Illumina-barcoded libraries were generated using the NEB RNA Ultra Directional kit according to the manufacturer's instructions. Sequencing was performed by a Illumina Hiseq sequencer. Data were aligned to mouse reference genome mm10 using Bowtie2. Normalized counts and differential expression analysis was performed using the DESeq2 R package. Gene enrichment analysis was performed using the R language clusterPro ler analysis package.
Single-cell RNA sequencing Fresh B16 melanoma subcutaneous tumor cells were washed and resuspended in 1x PBS (calcium and magnesium free) containing 0.04% BSA. 10,000 living cells were loaded on a Chromium Single Cell Controller to generate Gel Bead-In-EMulsions containing all cDNAs. Sample demultiplexing, barcode processing, alignment, ltering, UMI counting, and aggregation of sequencing runs were performed using the Cell Ranger analysis pipeline (v1.2). Seurat R package was used to merge the scRNA-seq and cell clustering and calculate the principal components analysis (PCA). Cells in which fewer than 200 genes were detected and in which mitochondrially encoded transcripts constituted more than 10% of the total library were excluded from downstream analysis. Genes detected in fewer than 3 cells across the data set were also excluded. Signi cant PCs were chosen through the jackstraw method. The t-SNE dimensionality reduction was generated using the RunTSNE function and the UMAP was performed with the RunUMAP function. Unsupervised clustering using a shared nearest neighbour modularity optimization based algorithm (resolution parameter 0.8) identi ed 16 distinct clusters.

Mouse cytokine antibody array
Cytokines from LTB and HTB mouse melanoma were detected by AAM-CYT-1000-2 mouse cytokine antibody array. The original data obtained from the scans were background removed and normalized between chips using Raybiotech software. Differential proteins were screened using fold change (expression difference multiple) under the following selection conditions: fold Change ≤ 0.83 or fold Change ≥ 1.2; mean signal value per group > 150 . ELISA Subcutaneous tumors were grinded by the homogenate machine and centrifuged for 20 min at 2000 rpm. The supernatant was collected and then tested by the ELISA kit to determine the OD at 450 nm wavelength.

Statistical analyses
Graphpad Prism software (version 8.0.1) and SPSS software (version 26.0) were used for all statistical analyses. Quantitative data are shown as mean ± s.d. Statistical signi cance was determined using paired two-tailed Student's t-tests. Values of p < 0.05 were considered signi cant. NS p > 0.05, * p < 0.05, ** p < 0.01 and *** p < 0.001.

Baseline tumor burden is associated with the transformation of TIME cellular components
To explore the impact of baseline tumor burden on the TIME, we analyzed the TCGA colon cancer and melanoma data sets. De ning the T1N0M0 staging as the LTB group and the T4N0M0 staging as the HTB group, we found that the population of CD8 + T cells in the HTB group was less abundant than that in the LTB group, and the population of M2 macrophages showed the opposite trend in colon cancer ( Figure  1A). Similar results were observed in melanoma ( Figure 1B). Therefore, we speculated that baseline tumor burden is associated with the transformation of TIME cellular components.
To further verify the above results, we collected postoperative tissues from 51 colon cancer patients and metastatic tumor needle biopsy samples from 2 patients with colon cancer liver metastasis for immunohistochemistry. According to the medium tumor size, 51 colon cancer patients were divided into 26 with LTB and 25 with HTB, with an average tumor volume of 13.00 cm³ in the LTB group and 88.11 cm³ in the HTB group. Furthermore, immunohistochemical staining showed that the number of CD8 + T cells in HTB colon cancer patients was signi cantly lower than LTB ( Figure 1C). Similar observations were made in metastatic tumor needle biopsy samples from patients with colon cancer liver metastasis.
In this patient cohort, the population of CD8 + T cells in a larger metastatic tumor was less than smaller metastatic tumor in the same patient ( Figure 1D). These results suggested that HTB may construct an inhibitory TIME by reducing the proportion of in ltrating CD8 + T cells. However, whether it affects the e cacy of ICIs through the transformation of TIME and the speci c mechanism remain to be explored.
High-baseline tumor burden restricts the number of CD8 + T cells in mouse TIME and attenuates antitumor e cacy of ICIs To de nite the role of baseline tumor burden on the e cacy of ICIs, we used 2×10 6 or 0.5×10 6 B16 melanoma cells or MC38 colon cancer cells to construct HTB or LTB mice models (Figure 2A and supplementary Figures 1A-B). Mice were intraperitoneally injected with physiological saline or anti-PD-1 antibody four times (Figure 2A). With anti-PD-1 treatment, the tumor growth of colon cancer in the LTB group was restrained. In contrast, the tumor growth of colon cancer in the HTB group was not ameliorated ( Figures 2B-D). Similar results were observed in mice with melanoma ( Figures 2E-G). To investigate the speci c mechanism by which HTB weakens the e cacy of ICIs, we rst evaluated the gene expression of HTB and LTB colon cancers by using RNA-Seq ( Figures 2H-I). The gene-ontology (GO) enrichment analysis showed that the highly expressed genes in LTB group were those related to immune-related pathways, while those related to tumor proliferation and metastasis-related pathways dominated in the HTB group ( Figure 2J). In accordance with the ndings in the patient cohort, a decreased percentage of CD8 + T cells was observed in HTB colon cancer TIME and melanoma TIME ( Figures 2K-L). Taken together, these ndings suggest that HTB is associated with poor immune checkpoint inhibitor e cacy and decreased percentage of CD8 + T cells.
Single-cell sequencing reveals the TIME differences between LTB and HTB melanoma Although HTB manifested an immunosuppressive TIME and a low response to immunotherapy, the underlying mechanisms remain poorly understood. Advances in single-cell RNA sequencing technologies have made it possible to comprehensively characterize the fundamental properties of tumor-in ltrating immune cells. We thereby resorted to single-cell sequencing to characterize immune features in LTB and HTB B16 melanoma showing distinct TIME and immunotherapy outcomes. After quality ltering, we obtained the single-cell transcriptome data for high-quality immune cells that were divided into 16 clusters ( Figure 3A). The dot plots compare the proportion of cells expressing cluster-speci c markers and their scaled relative expression levels ( Figure 3B). Similarly, we found that the proportion of M2-type macrophages in the HTB group was higher ( Figures 3C-D). Various studies have reported that tumorassociated macrophages (TAMs) play an important role in the CD8 + T cells function and immune response [8]. Such observations imply that M2-type macrophages may be responsible for the reduction of T cells and the poor e cacy of ICIs in HTB.

Macrophages mediate the formation of inhibitory TIME induced by HTB
We demonstrated that CD8 + T cells in ltration and the e cacy of immune checkpoint inhibitors differed between HTB and LTB constructed with different initial tumor cell numbers ( Figure 2). To ensure that the effects were not speci c to different initial tumor cell numbers and to exclude the physical factors of tumor size itself after rapid cell growth. Our next objective was to explore the effects of prolonged tumor growth time and increased tumor load induced by multiple lesions on the TIME and therapeutic e cacy of ICIs. We constructed the following subcutaneous tumor models: low tumor burden group (LTB, subcutaneous tumor formed on the left side with 0.5×10 6 cells for two weeks); high tumor burden group (HTB, subcutaneous tumor formed on the left side with 2×10 6 cells for two weeks); single-side subcutaneous tumor 3 weeks group (SS3w, subcutaneous tumor formed on the left side with 0.5×10 6 cells for three weeks); two-side subcutaneous tumor 2 weeks group (TS1, subcutaneous tumor formed on the left side with 0.5×10 6 cells for two weeks; TS2, subcutaneous tumor formed on the right side with 0.5×10 6 cells for two weeks); two-side subcutaneous tumor 2 weeks and 3 weeks groups (TS2w, subcutaneous tumor formed on the left side with 0.5×10 6 cells for two weeks; TS3w, subcutaneous tumor formed on the right side with 0.5×10 6 cells for three weeks) ( Figure 4A and supplementary Figure 2A). Figure 2K, in the MC38 colon cancer subcutaneous tumor mice model, the number of CD8 + T cells in ltrating TIME in the HTB group was signi cantly lower than that in the LTB group ( Figure   4B, Top). We also found that the number of CD8 + T cells in TIME in the SS3w group was lower than that in the LTB group ( Figure 4B, Top). These data indicated that the quantity of CD8 + T cells in the TIME decreased with the tumor growth, showing that HTB blocked the in ltration of CD8 + T cells independent of the physical factor of tumor size. Surprisingly, the number of CD8 + T cells in TIME of the TS1 and TS2 groups was similar to that in the LTB group, and was not reduced due to the augmentation in the total tumor burden ( Figure 4B, Middle). Furthermore, the number of CD8 + T cells in the TIME of the TS2w group was similar to that in the LTB group and TS1/2 group, but higher than that in the TS3w group ( Figure 4B, Bottom). The above ndings illustrate that the counts of CD8 + T cells in TIME are only related to single tumor burden, and are independent of the initial tumor cell number and the total tumor burden. Next, we found the same situation in the B16 melanoma subcutaneous tumor mice model ( Figure 4C). In addition, we intended to examine whether reducing the tumor burden could reform the immunosuppressive microenvironment of HTB tumors. To that end, we constructed a debulking surgery model using MC38 colon cancer cells: subcutaneous tumor formed on the left side with 2×10 6 cells for two weeks as HTB cancer and subcutaneous tumor formed on the right side with 0.5×10 6 cells for two weeks as LTB cancer. We reduced the tumor on the left side to the same tumor burden as that on the right side on day 14, and then detected the quantity of CD8 + T cells in TIME on days 16, 18, and 21 ( Figure 4D). Disappointingly, we did not nd that reducing the tumor size of HTB mice was able to reform the immunosuppressive microenvironment of HTB tumors ( Figure 4E-G).

Consistent with
These ndings together imply that HTB promotes the formation of immunosuppressive microenvironment by reducing the number of CD8 + T cells as the tumor grows in the TIME. To understand the reason for this phenomenon, we conducted single-cell sequencing. Our results showed that there were signi cant differences in M2-type macrophages between the HTB and LTB groups; speci cally, the proportion of M2-type macrophages in the HTB group was signi cantly higher than that in the LTB group.
Therefore, we hypothesized that M2-type macrophages in the HTB group affected the in ltration by CD8 + T cells and the anti-tumor e cacy of PD-1 blockade. To test this hypothesis, we used 2×10 6 melanoma cells or MC38 colon cancer cells to construct therapeutic mice models. Mice with different tumor burden were intragastrically administered with or without the macrophage inhibitor PLX3397 ( Figure 4H). Using PLX3397 on day 7 or day 13 did not affect the tumor burden on day 19 (supplementary Figure 2B). Compared with the group that only received PD-1 inhibitor, the group using PLX3397 at HTB (on day 13) and PD-1 inhibitor (on day 19) neither showed a signi cant delay in the tumor growth ( Figure 4I), nor promoted the in ltration of colon cancer TIME by CD8 + T cells ( Figure 4K). Interestingly, the group receiving PLX3397 at an early stage of the tumor (day 9) and PD-1 inhibitor on day 19 displayed a powerful delay in tumor growth ( Figure 4I) and augmented proportion of CD8 + T cells in the TIME ( Figure 4J). Similar results were observed in melanoma mice models ( Figure 4K, L). These ndings indicate that abundant M2-type macrophages in HTB result in reduced number of CD8 + T cells and poor immunotherapy effect. Moreover, the timing of the inhibition of macrophages was incredibly important for remodeling immunosuppressive microenvironment and improving the e cacy of PD-1 inhibitors for HTB tumors. Using PLX3397 at HTB could not restore the number of CTLs in the TIME because the number of CTLs in the HTB microenvironment had already been reduced.
HTB-TIME promotes the formation of PD-L1 + M2 macrophages The above results indicated that HTB environments incur the aggregation of M2-type macrophages.
Thus, we next examined the possible mechanisms. We speculated that HTB colon cancer and HTB melanoma microenvironment sustained the survival of macrophages. To verify this speculation, we tested the survival of macrophages after exposure to colon cancer tissue culture supernatants (TCS) or melanoma TCS. The results showed that macrophages exposed to HTB tissue culture supernatants (HTCS) exhibited a delayed onset of apoptosis when compared with those exposed to LTB tissue culture supernatants (LTCS) and control culture medium (NC) (Figures 5A-B). Furthermore, compared with the other two groups, macrophages exposed to HTCS showed a higher proportion of M2 type (Figures 5C-D).
Simultaneously, we also speculated that HTB environments contribute to the activated immunosuppressive phenotype of macrophages. As expected, compared with LTCS or NC-conditioned macrophages, the co-expression of CD206 and PD-L1 on HTCS-conditioned macrophages was upregulated signi cantly ( Figures 5E-H). These results indicate that HTB colon cancer and melanoma microenvironments promote the survival of macrophages and enhance the expression of PD-L1 on M2 macrophages.

HTB-associated PD-L1 + M2 macrophages inhibit T cells through the PD-1/PD-L1 axis
M2 macrophages can exert inhibitory effects on T cells in various ways, including secretion of inhibitory factors, metabolic competition, or expression of inhibitory effector molecules [9][10][11]. In view of the high PD-L1 expression on HTB-associated macrophages, we next investigated the impact of PD-L1 on HTBassociated macrophages in T cell suppression by adding neutralizing antibodies against PD-L1 into a macrophage/peripheral T cell co-culture system. First, puri ed peripheral CD8 + T cells were co-cultured with tumor-in ltrating macrophages or TCS-conditioned macrophages. HTB tumor-in ltrating macrophages with high expression CD206 and PD-L1 signi cantly suppressed the proliferation of CD8 + T cells (Figures 5A-B). Then, we added neutralizing antibodies against PD-L1 to the macrophages/T cells co-culture system. As expected, PD-L1 neutralizing antibody weakened the suppression of T cells' survival by HTB tumor-in ltrating macrophages (Figures 5A-B). HTCS-conditioned macrophages showed the same effects (Figures 5C-D). Blocking PD-L1 was also able to restore the IFN-γ production by T cells, which had been inhibited by HTCS-conditioned macrophages (Figures 5C-D). These ndings show that tumor-in ltrating macrophages that are activated by HTB-TIME suppress T cell proliferation and cytotoxicity through the PD-1/PD-L1 axis.

HTB induces PD-L1 + M2 macrophages by the IGFBP2-STAT3 axis
Tumor microenvironment contains a variety of soluble factors, including cytokines, in ammatory factors, and chemokines, all of which play an important role in cell-cell interaction, reshaping the tumor microenvironment and affecting the therapeutic e cacy [12]. To detect which cytokines activate and induce PD-L1 expression on macrophages, we rst screened cytokines in mouse melanoma environments by a mouse cytokines' antibody array. Compared with LTB, pro-MMP-9, bFGF, MMP-2, IGFBP-2, TIMP-2, MCSF, PF4, and other cytokines were highly expressed in the HTB microenvironment; among them, the highly expressed cytokine was IGFBP2 ( Figure 7A). In addition, the content of IGFBP2 in melanoma HTCS was signi cantly higher than that in LTCS as detected by ELISA ( Figure 7B Macrophages activated by IGFBP2 for 24 hours were co-cultured with CD8 + T cells. In the meantime, we added neutralizing antibodies or phosphate buffer saline (PBS) to the IGFBP2-induced macrophages/T cells co-culture system. The results showed that IGFBP2-induced macrophages effectively suppressed CD8 + T cells proliferation, and blockade of PD-L1 e ciently revoked T cell suppression mediated by such macrophages ( Figure 7E). Based on this result, we may conclude that HTB mainly promotes the formation of immunosuppressive TIME through PD-L1 + M2 macrophages induced by IGFBP2. However, previous studies on the effect of IGFBP2 on macrophages are limited. Only one study reported that IGFBP2 could activate STAT3 to construct an immunosuppressive tumor microenvironment in ltrated with M2-type macrophages in pancreatic duct cancer [13]. Interestingly, we found that abolishing the phosphorylation of STAT3 with an inhibitor (FLLL32 or HJC0152) effectively suppressed the formation of PD-L1 + M2 macrophages induced by IGFBP2 or HTCS (Figures 7F-G and supplementary Figure 3C). These results suggest that HTB-TIME induces macrophages with immunosuppressive phenotype mainly through the IGFBP2-STAT3 pathway.

Discussion
Although ICIs therapy can tremendously improve the survival of patients with various cancers, the proportion of patients who bene t from ICIs is only 10%-30%, and some treated patients have serious immune-related adverse events. Multiple clinical studies have shown that baseline tumor burden is signi cantly associated with poor prognosis and treatment outcomes after immunotherapy [14][15][16].
Patients with HTB tend to have poorer immunotherapy response and survival outcomes. However, research on the tumor baseline burden and immunotherapy e cacy has been limited to clinical studies, and the speci c mechanism remains unclear. In this study, we showed that immunotherapies of HTB colorectal cancer and melanoma were constrained, and that the number of in ltrating CD8 + T cells in the microenvironment was reduced, which indirectly suggests that HTB promotes the formation of immune desert TIME to diminish the e cacy of ICIs.
A question often arises as to whether palliative surgery can increase the e cacy of subsequent immunotherapy in patients with advanced cancer, and whether patients with early-stage tumors have better immunotherapy e cacy than those with advance-stage tumors [17]. Through the comprehensive animal models, we solved the above questions and obtained other major ndings as follows: 1) The number of CD8 + T cells and M2-type macrophages in the TIME of tumor-bearing mice showed timedependent changes. To rule out that the tumor logarithmic growth of HTB physically prevented CD8 + T cell in ltration, we designed a tumor-bearing mouse model with the same initial number of tumor cells at different growth times. We found that it was not the tumor size that blocked the CD8 + T in ltration; rather, with tumor growth, the abundance of M2-type macrophages gradually increased, while CD8 + T cells showed the opposite trend in a time-dependent manner. 2) Total tumor burden did not affect the microenvironment or the immunotherapy e cacy of a single tumor, which indirectly indicated that metastatic tumors might not affect the e cacy of immunotherapy in the primary tumor. 3) Reducing the tumor size of HTB mice by surgery did not effectively augment the number of CTLs in the TIME and boost immunotherapy e cacy, which further indicates that the effect on TIME transformation and the therapeutic effect of HTB are independent of physical factors, which may be due to the fact that the desert microenvironment caused by M2-type macrophages has not been ameliorated. Some researchers believe that reducing the tumor size by palliative operation, chemotherapy, and radiotherapy should be explored as a way to improve the e cacy of ICIs in HTB cancer patients. Disappointingly, in our study, we demonstrated that the e cacy of immunotherapy after palliative surgery in colon cancer or melanoma patients with advanced tumors was not improved. However, previous studies have suggested that chemoradiation may induce T-cell activation by producing new mutations [18,19], which is a potential entry point for improving the immune bene t of HTB in the future.
This study showed that the effect of HTB on the e cacy of immunotherapy was due to the immunosuppressive microenvironment constructed by HTB-related macrophages. Macrophages in the TIME are mainly divided into M1 and M2 types. M2 macrophages can exert inhibitory effects on T cells by secreting inhibitory factors, performing metabolic competition, expressing inhibitory effector molecules, and inducing stromal brosis to expel T cells [9][10][11]20]. Previous studies have demonstrated that endoplasmic reticulum stress, oxidative stress, fatty acid oxidation, lactate, and matrix stiffness can drive macrophage M2-polarization [21][22][23][24]. Our research innovatively pointed out that HTB may also cause macrophages to switch to M2-type and reduce the apoptosis of macrophages in the TIME. In addition, HTB-associated macrophages inhibited CD8 + T cells through the PD-1/PD-L1 axis; in the macrophage/peripheral T cell co-culture system, anti-PD-L1 antibody effectively decreased the inhibitory effect of HTB-activated macrophages on CTLs. Although HTB-associated macrophages inhibit T cells through the PD-1/PD-L1 axis, immunotherapy is still ineffective in HTB, which may be because CTLs have already been reduced in the HTB microenvironment. Next, we explored the e cacy of a combination regimen of macrophage inhibitors and immunotherapy. Notably, macrophage colony-stimulating factor-1 (CSF-1) inhibitor itself did not affect the tumor growth, but the timing of using CSF-1 inhibitor was crucial; namely, only using CSF-1 inhibitor when the tumor burden is low was able to signi cantly strengthen the e cacy of anti-PD-1 in HTB mice. This conclusion further de nes the immune desert TIME in HTB; at that time, inhibiting M2 macrophages with a CSF-1 inhibitor cannot restore the number of CTLs in the tumor microenvironment. The above results suggest that the combination of CSF-1 inhibitor and ICIs cannot bring signi cant bene t to patients with advanced-stage tumors; nevertheless, earlier application of CSF-1 inhibitor to ameliorate the tumor microenvironment could improve the e cacy of second-line or even post-line immunotherapy treatments for patients. In summary, HTB weakened the e cacy of immunotherapy not due to the physical factor of tumor size, but due to the desert immune microenvironment constructed by HTB-associated macrophages. Therefore, debulking surgery and macrophage inhibitors used in HTB could not change the outcome of poor immunotherapy e cacy in HTB.
Next, we investigated the speci c mechanism of HTB-associated macrophages production. By isolating and obtaining the culture supernatant of tumor tissue, we found that HTB regulates TAM through the secretory pathway. Tumor cells can secrete a variety of soluble factors, including cytokines, in ammatory factors, and chemokines; in that way, they can reshape the tumor microenvironment and affect tumor progression and therapeutic e cacy [12]. In this study, we found that compared with LTB, HTB secreted large quantities of cytokines, including pro-MMP-9, βFGF, MMP-2, IGFBP-2, TIMP-2, MCSF, and PF4.
Among them, only IGFBP2 was able to promote the polarization of macrophages and induce the PD-L1 expression, which were also rst demonstrated in HTB tumors. IGFBP2 was identi ed as a regulator of the IGF system; it plays a signi cant role in several crucial oncogenic processes, such as cancer cells migration, invasion, proliferation, vascular formation, and epithelial-mesenchymal transition, by regulating various signal transductions [25][26][27][28]. Previous studies about the effect of IGFBP2 on macrophages are limited. We con rmed that IGFBP2 not only promoted macrophage polarization but also regulated PD-L1 transcription through the STAT3 signaling pathway, which is consistent with the results of Zhang and Yang [13,29]. STAT3, a member of the signal and activator of transcription (STAT) protein family, is involved in a variety of biological processes, including cell growth, differentiation, and angiogenesis [30,31]. Anomalously activated STAT3 in tumor cells is associated with poor prognosis [32]. Accumulating evidence has revealed that STAT3 participates in regulating the formation of immunosuppressive microenvironment by inhibiting the release of proin ammatory cytokines, damaging antigen presentation, and inhibiting the aggregation and cell killing function of effector cells [33][34][35]. In addition, as a transcription factor, STAT3 can also directly promote the transcription of PD-L1 and PD-L2 [36,37]. Thus, the combination therapy targeting STAT3 is considered a promising direction for improving immunotherapy e cacy. Various STAT3 inhibitors in combination with immune checkpoint blockers are currently being tested in clinical trials; among them, STAT3 inhibitor BBI608 was approved by FDA for gastric and pancreatic cancer based on the promising results of phase I/II clinical trials. The phase II trials of BBI608, in combination with Pembrolizumab (NCT03647839) or nivolumab (NCT02851004), in metastatic colorectal cancer are ongoing. Combined with the conclusions of this study, STAT3 inhibitors might be used as sensitizers for HTB immunotherapy, and their application in rst-line may improve the clinical e cacy of second-line or even post-line immunotherapy. However, how IGFBP2 activates STAT3 and the speci c mechanism of the different levels of IGFBP2 secreted by tumor cells with HTB and LTB have not been clari ed. The large amount of IGFBP2 in HTB may be related to hypoxia or nutrient de ciency in HTB TME [38], which needs to be further investigated.

Conclusion
In summary, we de ned the effects of baseline tumor burden on the TIME and immunotherapy e cacy.
HTB induced the production of PD-L1 + M2 macrophages via the IGFBP2-STAT3-PD-L1 pathway to promote the generation of immunosuppressive TIME (Figure 8)    Single-cell sequencing reveals the TIME differences between LTB and HTB melanoma.  HTB induced the inhibitory TIME through macrophages. HTB-asscociated PD-L1 + M2 macrophages inhibited T cells through PD-1/PD-L1 axis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.