Background
About 90% of head and neck cancers (HNCs) constitute head and neck squamous cell carcinomas (HNSCCs). HNSCCs result in a high morbidity and mortality rate with only 50–60% of patients having a 5-year survival rate [
1]. HNSCCs are often associated with carcinogens, such as alcohol and tobacco use, or oncogenic human papillomavirus (HPV) infection [
2,
3], thus, are categorized as HPV
− or HPV
+ HNSCCs. HNSCCs display a high rate of genetic heterogeneity, consisting of hyper-activation of oncogenes (e.g.,
PIK3CA and HRAS) or both loss-of-function mutations and potential gain-of-function mutations in multiple genes (e.g.,
TP53,
CASP8 and
CREBBP/EP300) [
4‐
14]. Phosphoinositide 3-kinase (PI3K) is a frequently dysregulated pathway in HNSCCs with a
PIK3CA gene mutation rate of approximately 16% and gene amplification rate of more than 30% [
4,
15]. However, therapies targeting the PI3K pathway have had limited efficacy in HNSCCs so far [
16]. Another highly mutated gene in HNSCCs is the
TP53 tumor suppressor gene, with over 80% of HPV
− HNSCCs harboring
TP53 mutations, whereas
TP53 mutations occur much less frequently in HPV
+ HNSCCs (~ 3%) [
3,
4]. While clinical trials have tested several therapies targeting p53, they have yet to be proven effective [
17‐
20]. In general,
TP53 mutations in HNSCC are associated with poor prognosis and overall survival with increased rate of recurrence and resistance to therapies [
7‐
9,
20,
21]. Thus, it would be of great interest to better understand how these two genetic alterations influence the aggressive phenotypes of HNSCCs, thereby laying a scientific foundation for developing more effective therapies.
Our prior studies showed that HNSCC patients with
PIK3CA amplification (PIK3CA
Amp) exhibited a higher frequency of harboring
TP53 mutations (TP53
Mutated) compared with patients with WT
PIK3CA [
22]. In addition, we found that HNSCC patients with dual genetic alterations, i.e., PIK3CA
Amp/TP53
Mutated, showed a significantly worse prognosis in their 10-year overall survival than PIK3CA
WT/TP53
WT group [
22]. However, the underlying mechanisms that lead to worse outcomes in PIK3CA
Amp/TP53
Mutated HNSCC patients remain incompletely understood. In this regard, prior studies have generated murine models that mimicked the alterations of
PIK3CA,
p53 or both in HNSCCs [
23‐
25]; however, none of the prior studies showed that genetic alterations in these two genes spontaneously induced HNSCC development. We have established a genetically engineered mouse model, by deleting
p53 and constitutively activating
PIK3CA in mouse keratin 15-expressing (K15
+) stem cells, which leads to the development of multi-lineage tumors including SCCs, termed keratin-15-p53-PIK3CA (KPPA) tumors [
22]. In the current study, we established different KPPA SCC tumor lines and performed in-depth phenotypic characterization of them. We envision that these KPPA cell lines may provide an experimental model system to further elucidate how
TP53 deletion and
PIK3CA hyperactivation cooperate to result in aggressive phenotypes of HNSCCs.
The tumor microenvironment (TME) of HNSCCs is composed of various subsets of tumor-infiltrating cells that can interact with tumor cells or with each other via intricate networks to promote tumor progression or mediate anti-tumor immune responses. We have extensively reviewed how different subsets of immune cells contribute to an immunosuppressive TME of HNSCCs [
26]. In particular, myeloid cells such as myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs) in the TME not only promote tumor progression and angiogenesis, but also suppress anti-tumor immune responses [
26]. MDSCs are CD11b
+ cells and can be phenotypically subdivided into two groups, polymorphonuclear MDSC (PMN-MDSC) and monocytic MDSC (M-MDSC) [
27]. One of the major functions of PMN-MDSCs is to suppress T cells, while M-MDSCs tend to differentiate into TAMs at tumor sites. TAMs are classified into two subpopulations: M1-TAMs, which mediate proinflammatory and anti-tumor responses, and M2-TAMs, which are immunosuppressive and promote tumor growth [
26,
28]. M2-TAMs express a higher level of CD206 and display immunosuppressive properties by expressing arginase-1 (Arg-1), chemoattractant such as IL-10 and TGF-β, and chemokine CCL17 and CCL22 [
29]. Prior studies showed that HNSCC TME largely encompasses M2-TAMs, which may impair effector T cell function [
30]. A higher level of TAMs in the TME correlates with lymph node metastasis and advanced stage of HNSCCs [
26,
28]. While it is conceivable that tumor-derived growth factors or cytokines may be able to modulate the TME, it remains incompletely understood how SCCs harboring genetic alterations in both
TP53 and
PIK3CA drive the expansion of TAMs.
Immune checkpoint inhibitors (ICIs), including monoclonal antibodies against programmed death 1 (PD1) and PD ligand 1 (PD-L1), have been approved for HNSCCs; however, different patients exhibit highly variable responses, and the overall response rate remains low [
31‐
38]. In addition, reproducible and highly reliable markers are still lacking to predict ICI responses in HNSCCs. In the current study, we established two different KPPA tumor lines that mimic human HNSCCs with dual genetic alterations in
TP53 and
PIK3CA, and found they upregulated distinct signaling pathways. Moreover, we showed that these two KPPA tumor lines responded to anti-PD-L1 differentially, although both were initiated by the same oncogenic driver mutations. Our study indicates the limitations of stratifying cancers according to their genetic alterations and suggests that evaluating HNSCC tumor-intrinsic cues along with immune profiles in the TME may help better predict ICI responses.
Materials and methods
Generation of tumor cell lines and in vivo mouse work
The parental TAb2 and TCh3 cell lines were derived from spontaneous tumors that developed in the same female K15.CrePR1(+)p53
f/fPIK3CA
c/c mouse [
22] at different locations (Fig.
S1A). Then, the parental KPPA tumor lines were transplanted into wildtype (WT) C57BL/6 (B6) recipients (Jax Laboratories) (Fig.
S1A). Transplanted tumors were isolated and used for histology analysis and for creating daughter TAb2 and TCh3 cell lines that were employed for all the tumor injection studies (Fig.
S1A). TAb2 and TCh3 cells were cultured in DMEM complete media supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, 1% HEPES buffer at 37 °C CO
2 incubator (5%).
Tumor cells (0.5 × 106 TAb2 or 1 × 106 TCh3) were injected into wild-type (WT) female C57BL/6 (B6) mice (Jackson Laboratories) (6–8 weeks old). Mice were injected subcutaneously at their flank with tumor cells suspended in PBS and 50% Matrigel Basement Membrane Matrix (Corning) to a final volume of 100 μl. When tumor volume reached about 150-200 mm3 approximately 9–12 days post-injection, subsequent treatment was initiated. Tumor-bearing mice were treated with anti-PD-L1 (clone 10F.9G2, BioXCell, Catalog# BE0101) by intraperitoneal (i.p.) injection of 200 μg/mouse/time diluted in PBS for 2 weeks (three times per week). PBS only was used as vehicle control (VC). Tumor length and width were measured with calipers and tumor volume was calculated as length×width2 × (π/6). Relative tumor volume (RTV) was used to assess treatment effects, defined as TVn/TV0, where TVn is the TV at day n and TV0 is the TV when the treatment started. Recipient survival was monitored until mice reached endpoints of severe tumor ulceration, tumor volume reaching 20 mm in diameter or other humane end points, and mice were euthanized in accordance with institutional guidelines. All mice were maintained under specific pathogen-free conditions in the vivarium facility of University of Colorado AMC. Animal work was approved by the Institutional Animal Care and Use Committee of University of Colorado Anschutz Medical Campus (Aurora, CO).
In vitro culture with Bone Marrow (BM) cells and tumor cells
BM cells were collected from WT B6 mice and obtained by using a 25-G needle and syringe as described previously [
39]. BM cells were filtered through 70 μm cell strainer and red blood cells (RBC) were lysed with RBC lysis buffer (Sigma Aldrich, USA). BM cells were then counted (1 × 10
6) and co-cultured with either TAb2 or TCh3 tumor cells (2.5 × 10
4) in 24 well plates. Tumor cells were seeded in DMEM complete media in three 24-well plates 24 h prior to BM collection (Day − 1) for analysis at different time points (Day 2, 3, and 4). The supernatant of tumor culture was removed the next day (Day 0), and BM cells (1 × 10
6) in 1 mL RPMI complete media were added. Media only with BM cells was used as control. Co-cultured cells were collected on different time points, stained for myeloid cell populations, and analyzed by flow cytometry (BD Fortessa). For transwell co-culture, TAb2 or TCh3 tumor cells (2.5 × 10
4) in 200 μl DMEM media were placed in the top insert of a transwell (Corning, CLS3413-48EA), while BM cells (1 × 10
6) in 800 μl RPMI media were seeded on the bottom well. Cells cultured for different time points (Day 2 and 4) were collected and analyzed as described above.
For inhibiting CSF1R or VEGFR, 10 μg/mL of anti-CSF1R mAbs (BioXCell, BE0213) or 500 nM/mL of Axitinib (MedChemExpress, HY-10065) were added into the co-culture of TAb2 tumor and BM cells, respectively. TAb2 cells co-cultured with BM cells alone were used as control. Co-cultured cells were collected on different time points (Day 2, 3, and 4) and analyzed as described above. On Day 2, one plate was used for staining while the other plates were replenished with their corresponding conditioned medias. Subsequently, co-cultured cells were collected on Day 3 and 4 for analysis as described above.
Immune profiling by flow cytometry
Single cell suspensions were prepared from spleens and tumors harvested from WT B6 tumor-bearing mice as previously described [
22]. Single cell suspensions of tumor samples were prepared by finely cutting the tumors with surgical blades into smaller pieces. Then Liberase DL (50 μg/ml) was added to the diced tumor suspensions, and incubated at 37 °C for 30 min. Then, Liberase was neutralized with 2% FBS medium, and tumor suspensions were filtered through 70 μm cell strainers, and centrifuged at 1500 rpm for 5 min at 4 °C to obtain a pellet. Cell pellets were resuspended in culture media and are ready to be stained. Single-cell suspensions were used for immediate staining with flow cytometry antibodies, or for ex vivo stimulation followed by antibody staining as previously described [
22]. Briefly, cells were stained with 1:1000 LIVE/DEAD™ Fixable Aqua Dead Cell Stain (Invitrogen). Cells were then washed twice with 2% FBS in PBS before adding TruStain FcX™ (anti-mouse CD16/32) (BioLegend). Surface staining was then performed by adding Brilliant Stain Buffer Plus (BD Horizon) into each surface antibody flow panel mixture according to manufacturer’s instructions. BD Cytofix/CytoPerm buffer kit (BD Biosciences) was used according to the manufacturer’s instructions before adding intracellular staining antibodies for each panel. Surface and intracellular staining antibodies are listed in Table
1. Data were acquired on BD Fortessa and analyzed with FlowJo™ software V10 (FLOWJO, Oregon, USA).
Table 1
Antibodies used in this study
Flow Antibodies
|
Antibody
|
Fluorophore
|
Company
|
Catalog
|
Clone
|
Concentration
|
CD11c | PerCP/Cy5.5 | BioLegend | 117327 | N418 | 1 μg/mL |
PD-L1 | BV786 | BD Bioscience | 741014 | MIH5 | 1 μg/mL |
MHCII | BV711 | BioLegend | 107643 | M5/114/15/2 | 0.25 μg/mL |
CD19 | Brilliant Violet 605 | BioLegend | 115539 | 6D5 | 1 μg/mL |
Ly6C | BV421 | BioLegend | 128031 | HK1.4 | 1 μg/mL |
Ly6C | FITC | BioLegend | 128006 | HK1.4 | 1 μg/mL |
Ly-6G | APC/Cy7 | BioLegend | 127623 | 1A8 | 1 μg/mL |
CD11b | Alexa Fluor 700 | BioLegend | 101222 | M1/70 | 1 μg/mL |
CD206 | PE/Cy7 | BioLegend | 141719 | C068C2 | 1 μg/mL |
CD86 | BV421 | BioLegend | 105031 | GL-1 | 1 μg/mL |
F4/80 | PE Dazzle | BioLegend | 123145 | BM8 | 1 μg/mL |
TCR beta | BV605 | BioLegend | 109241 | H57–597 | 1 μg/mL |
CD4 | BV421 | BioLegend | 100563 | RM4–5 | 1 μg/mL |
CD8a | Alexa Fluor 700 | BioLegend | 100729 | 53–6.7 | 1 μg/mL |
CD45 | BUV395 | BD Bioscience | 564279 | 30-F11 | 1 μg/mL |
TNFalpha | BV650 | BD Bioscience | 563943 | mp6-xt22 | 1 μg/mL |
IFN gamma | PE | eBioscience | 12–7311-41 | XMG1.2 | 1 μg/mL |
Granzyme B | PE/Cy7 | BioLegend | 372213 | QA16A02 | 1 μg/mL |
Western Blotting and Histology Antibodies
|
Antibody
|
Application
|
Company
|
Catalog
|
Clone
|
Dilution
|
STAT3 | Western | Cell Signaling Technology | 12640S | D3Z2G | 1:1000 |
p-STAT3 | Western | Cell Signaling Technology | 9145S | D3A7 | 1:2000 |
GAPDH | Western | Cell Signaling Technology | 5174S | D16H11 | 1:1000 |
IgG-HRP | Western | Cell Signaling Technology | 7074 | | 1:3000 |
F4/80 | MSI | Cell Signaling Technology | 70076 | D2S9R | 1:500 |
Keratin 5 | MSI | Abcam | ab64081 | SP27 | 1:200 |
Arginase-1 | MSI | Cell Signaling Technology | 93668 | D4E3M | 1:500 |
Western blot, ELISA, and cytokine Array
Cells were harvested and lysed with Lysis Buffer M (Roche) supplemented with complete mini (Roche). Samples were then loaded onto a NuPAGE™ 4 to 12%, Bis-Tris, Mini Protein Gel (ThermoFisher) and transferred to nitrocellulose membrane. Membranes were stained with antibodies against STAT3, p-STAT3 or GAPDH diluted according to manufacturer’s recommendations. Membranes were washed and later stained with anti-rabbit IgG-HRP secondary antibody. Blots were imaged on an Odyssey 9120 Digital Imaging System (Li-Cor). All the antibody information is included in Table
1.
TAb2 or TCh3 tumors cells (1 × 106) were seeded onto 100 mm culture dish and incubated for 48 h. Cell lysate and/or culture supernatant were then collected for ELISA or Proteome Profiler Mouse XL Cytokine Array. For ELISA, cell lysate or supernatant samples were analyzed for cytokine/chemokine levels using the Mouse HGF ELISA Kit (Raybiotech, ELM-HGF), Mouse CXCL17/VCC-1 ELISA Kit (Raybiotech, ELM-CXCL17), Mouse CXCL16 (Sigma-Aldrich, RAB0127), Mouse CXCL12/SDF-1α (Sigma-Aldrich, RAB0125, and Raybiotech, ELM-SDF1a), and Mouse CSF1 (Raybiotech, ELM-MCSF-1) according to the manufacturer’s instructions. Buffer alone served as background and was subtracted from OD reading of 450 nm.
Proteome Profiler Mouse XL Cytokine Array (R&D, ARY028) was performed according to the manufacturer’s instructions. Cytokine array membranes were imaged on an Odyssey 9120 Digital Imaging System (Li-Cor). Data was analyzed by capturing the pixel density (signal) and the signals were then normalized by subtracting the background to calculate the intensity of each cytokine. GraphPad Prism 9.1.2 software (GraphPad Software, Inc.) was used to analyze data.
Histology analysis, immunofluorescence (IF) and multispectral imaging (MSI) staining
Hematoxylin and eosin (H&E) and immunofluorescence (IF) staining of tumor tissues were performed as described previously [
22]. For analyzing the spatial immune profile of mouse tumor tissues by MSI, the Opal™ 4-Color Fluorescent IHC Kit (Akoya Biosciences, NEL810001K) was used according to the manufacturer’s instructions. Slides were stained with primary antibodies against F4/80, Keratin 5, and Arginase-1 (see Table
1).
Survival analysis of TCGA HNSCC patient cohort
Within cBioPortal platform (
https://www.cbioportal.org) and under the category of head and neck cancers, we downloaded data from two cohorts of HNSCCs (TCGA, Firehose Legacy,
n = 530; and TCGA, PanCancer Atlas,
n = 523 samples), including clinical data, DNA mutation data, normalized mRNA expression data (log-transformed mRNA expression z-scores compared to the expression distribution of all samples) and Copy Number Alteration (CNA) data. Data from those two cohorts were merged by utilizing patient IDs (
n = 527) and patients who have both amplification or gain of PIK3CA copy number (PIK3CA
Amp) and truncation (include nonsense mutation, frame shift insertion and frame shift deletion) or missense of TP53 gene (TP53
Mutated) were filtered out for the later survival analysis (
n = 305). However, only 300 patients had analyzable data due to 5 of them missing mRNA expression data. Scores were calculated based on sum of normalized expression for each of the genes, and PIK3CA
Amp/TP53
Mutated HNSCC patients were divided into high-expression group (having a score > the median score) or low-expression group (having a score < = the median score). The association between different group survival was evaluated by cox regression and the
p-value was presented within Kaplan-Meier curves.
Bulk RNA-sequencing, Whole Exome Sequencing (WES), and single cell RNA-sequencing
TAb2 and TCh3 tumor cells were cultured in DMEM complete media and collected for RNA purification. Total RNA was purified with TriPure (Roche) and cleaned up with RNAeasy Kit (Qiagen) according to manufacturer’s instructions. RNA samples were then depleted of ribosomal RNA and subjected to pair-ended RNA sequencing by NovaSEQ 6000 (University of Colorado at Anschutz Genomics and Microarray Core). Raw sequencing data with adapter sequences were filtered using BBduk from BBtools (version 38.86) to remove adapter contamination and obtain clean data for subsequent processing. To obtain transcript quantification from RNA-seq data, alignment tool Salmon was employed. Output files from Salmon were used for visualization and further analysis. Output files were processed using the function DESeqDataSetFromTximport in DESeq2 (V.1.30.1) to create DESeq objects and were normalized to identify differentially expressed genes (DEGs). To remove noise values associated with low count genes, the function lfcShrink in DESeq2 using apeglm estimator was applied to shrink log
2(fold change) [
40]. Volcano plot and heatmap were created with ggplot2 and pheatmap packages in R.
Genomic DNA was purified from TAb2 or TCh3 tumor cells and DNA samples were submitted to Novogene for WES using library preparation kit (Agilent SureSelect Mouse All Exon). The library was checked with Qubit and real-time PCR for quantification and bioanalyzer for size distribution detection. Quantified libraries were pooled and sequenced on Illumina NovaseqS4 PE150. Quality control was performed using FastQC. Reads were aligned to the mouse reference genome (mm10/GRCm38) using Burrows-Wheeler Aligner (BWA) [
41] and the aligned files were further sorted and marked for duplicates by Picard tools. Base quality scores were recalibrated by Genome Analysis Toolkit (GATK, version 4.2.2.0) BaseRecalibrator. Subsequently, two variant-calling pipelines were applied to identify tumor-specific variants. For the first pipeline, GATK (version 4.2.2.0) was applied as follows: Mutect2 function was used to call unique variants by comparing two cell lines. The TAb2 unique variants were called by considering TAb2 as tumor tissue and TCh3 as normal control, while TCh3 unique variants were called vice versa. For the second pipeline, BCFtools
mpileup function [
42] was applied to call variants per site. Mutation callings of the two cell lines were merged and compared. The unique variants for TAb2 compared with TCh3 were defined as: TAb2 (tumor) total count > = 10, TAb2 alternative count > = 4, TAb2 alternative rate > =10% and TCh3 (normal) alternative rate = 0. Similarly, unique variants for TCh3 were defined vice versa. Next, unique variants per cell line were annotated for SNPs and amino acid (protein) changes by tool SnpEff and SnpSift [
43]. For the first pipeline, further filtering was performed with FilterMutectCalls function after annotation and only passed variants were included. Variants with high or moderate putative impact were used for further analysis that might change protein functions or effectiveness.
Single cells suspensions were obtained from tumor-bearing mice treated with PBS or anti-PD-L1 as described previously [
22]. CD45
+ immune cells were isolated using EasySep™ Mouse CD45 Positive Selection Kit (StemCell Technologies, Catalog# 18945) according to manufacturer’s instructions. Purified CD45
+ immune cell samples were submitted to the University of Colorado at Anschutz Genomics and Microarray Core for single cell capture and library preparation. Cells were loaded into a 10 × Genomics Single-cell Chip G for the 3′ captures. Single-cell gene expression libraries were prepared using Chromium Next GEM Single Cell 3′ Reagent Kits (v3.1: Dual Index Libraries) according to the manufacturer’s instructions. Samples were sequenced on the Illumina NovaSeq 6000 platform for an estimated read depth of 100,000 reads per cell. After sequencing, reads were mapped to the reference mm10 genome using the 10 × Genomics CellRanger (V.2.0.2, V.3.0.2 and V.3.1.0) count pipeline. Cells with < 500 genes detected or > 10% mitochondrial RNA content were removed from further analysis. Samples were processed using the functions NormalizeData, FindVariableGenes and ScaleData in Seurat V3.2.3. Integrated variable features were used to cluster and visualize all cells by UMAP with RunUMAP, FindNeighbors, and FindClusters in Seurat V3.2.3. Each cluster was defined by comparing their gene expression to single-cell RNA sequencing databases of known cell types [
44] and PanglaoDB [
45] along with a curated list of commonly known markers. Clusters were then renamed and visualized by their UMAP coordinates. UMAPs, bar graphs and violin plots were created with the R packages. DEGs between anti-PD-L1 treated and control were identified by FindMarkers function in Seurat V4.0.3 and used to find enriched pathways by using Gene Ontology (GO) analysis in clusterProfiler (V.4.0.2).
Ingenuity Pathways Analysis (IPA)
Pathway analysis was performed with the QIAGEN’s Ingenuity® Pathway Analysis (IPA®, QIAGEN Redwood City,
www.qiagen.com/ingenuity) software. Canonical pathways significantly enriched among the DEGs in the dataset were identified using one-sided Fisher’s Exact Test and the Benjamini-Hochberg method was used to adjust canonical pathway to obtain FDR
p-values (significant threshold at 0.05).
Discussion
We uncovered tumor-intrinsic differences that may underlie the differential responses to ICI by establishing and employing two KPPA SCC tumor lines, TAb2 vs. TCh3, both of which harbor TP53 deletion and PIK3CA hyperactivation and originated from the same K15.CrePR1(+)p53f/fPIK3CAc/c mouse. We found that: (1) TCh3 tumors are relatively sensitive to anti-PD-L1, while TAb2 tumors failed to respond completely; (2) Prior to anti-PD-L1 treatment, the TME of TAb2 tumors is highly immunosuppressive evidenced by heavy infiltration of TAMs, especially, M2-TAMs, whereas TCh3 tumors contained more CD8 TILs with better effector functions; (3) TAb2 tumor cells drastically expanded F4/80+ TAMs from BM precursors, which required CSF1 and VEGF; (4) More aggressive phenotypes of TAb2 tumors correlate with upregulation of chemokines/growth factors that may contribute to immunosuppressive TME; and (5) anti-PD-L1 did not affect the TME of TAb2 tumors but significantly increased the number of CD8 TILs in TCh3 tumors. We suggest that tumor-intrinsic differences may contribute to differential ICI responses by orchestrating TME prior to ICI treatment. Although these KPPA tumors harbor same oncogenic driver mutations, they appear to establish differential TME that is highly immunosuppressive or relatively conducive for ICI therapy. These results suggest that evaluating HNSCC tumor-intrinsic cues along with immune profiles in the TME may help better predict ICI responses. Our experimental models may provide a platform for pinpointing tumor-intrinsic differences underlying an immunosuppressive TME in HNSCCs and for testing combined immunotherapies targeting either tumor-specific or TAM-specific players to improve ICI efficacy.
TAMs associate with tumor progression by promoting evasion of immunosurveillance, angiogenesis, metastasis, and therapy resistance or inhibiting effector functions of CD8 TILs [
65‐
67]. A meta-analysis showed that increased density of TAMs, including M2-like subtypes, correlate with poor clinicopathologic markers in HNSCC such as advanced tumor stage and nodal metastasis [
68]. Consistently, we found that TAb2 tumors heavily infiltrated with TAMs exhibited aggressive phenotypes and failed to respond to anti-PD-L1 completely. Furthermore, we found that co-culturing TAb2 tumor cells with BM cells resulted in drastic expansion of F4/80
+ TAMs and CD206
+ M2-TAMs, which is independent of cell-cell contact, suggesting a major role of secretory factors in promoting TAM differentiation. Blocking CSF1/CSF1R and VEGF/VEGFR pathways in the co-culture of TAb2-BM remarkably suppressed TAM production. In line with these findings, TAb2 tumors upregulated CSF1 and VEGF at both mRNA and protein level. While prior studies have reported that CSF1 and VEGF can stimulate differentiation and polarization of TAMs [
69‐
71], their role in HNSCC prognosis and therapy response is less well understood. Recently, CSF1 upregulation was shown to correlate with increased TAM infiltration and poor prognosis in oral SCC [
72]. Taken together, we suggest that CSF1 and VEGF upregulation may serve as predictive markers for worse prognosis and ICI therapy resistance in HNSCCs harboring
TP53 deletion and
PIK3CA amplification.
We verified the expression level of several proteins that are involved in promoting tumor aggressiveness (e.g., invasiveness and angiogenesis) and inducing an immunosuppressive TME. We showed that TAb2 tumor cells expressed higher levels of CSF1, VEGF, HGF and CXCL12. Consistent with our findings, previous studies showed HNSCC patients with a higher level of CXCL12 have poor prognosis [
73,
74]. HGF is a pleiotropic growth factor and cytokine, whose upregulation promotes tumor cell survival, motility, and proliferation [
47,
48], and its receptor, HGFR (a.k.a. c-MET), is a well-known oncogene [
75]. Prior studies also revealed a positive feedback loop of HGF/c-MET/STAT3 signaling that plays an important role in tumorigenesis [
76‐
78]. STAT3 pathway is downstream of receptor tyrosine kinase (RTK) including cytokine and growth factor receptors. Cytokines such as CSF1 can activate RTK (CSF1R) and downstream STAT3 pathway, which can promote tumor migration and invasion in colon cancers [
79]. In this regard, TAb2 tumors upregulated both CSF1 and CSF1R, potentially enforcing a positive signaling cycle. VEGF is another factor involved in a positive feedback loop of STAT3 activation in tumorigenesis and angiogenesis [
53,
80]. IL-6/JAK/STAT3 pathway is predicted to be activated in TAb2 tumors, which is often hyperactivated in various types of cancer that correlates with poor prognosis [
81]. Again, TAb2 tumors upregulated both IL-6 and IL-6Rα transcriptionally, suggesting a positive signal loop. Lastly, we found that TAb2 tumor expressed a higher level of p-STAT3. Hence, our data illustrate a central theme of positive feedback loops reinforcing aggressive phenotypes of TAb2 tumors. In contrast, CXCL17 may be one of the factors that predict less aggressive phenotypes in HNSCCs. Supporting this notion, we found TCh3 tumor expressed a higher level of CXCL17 and HNSCC patients with higher expression of CXCL17 exhibit a better prognosis [
74]. Thus, our studies may provide more information for better predicting HNSCC prognosis and establish an experimental system for testing new therapeutic targets of HNSCC by breaking the vicious positive feedback cycles.
While ICI demonstrated benefits for patients with recurrent or metastatic HNSCC, the response rate remains relatively low (< 20%) [
31‐
38]. Thus, characterizing the tumor-intrinsic signaling pathways and immune landscape that are associated with ICI-response vs. resistance may allow us to develop better strategies to improve ICI efficacy. For instance, TCh3 tumors contained more CD8 TILs prior to anti-PD-L1 treatment, the presence of pre-existing CD8 TILs may be a predictive marker for ICI efficacy [
82]. Furthermore, CD45
+ tumor-infiltrating immune cells in TCh3 tumors clearly undergo more transcriptional changes upon anti-PD-L1 treatment that favor anti-tumor immunity; however, the underlying mechanisms for these observations remain unclear. Compared with TAb2 tumors, TCh3 tumors upregulated completely different chemokines and cytokines, such as CXCL16. Interestingly, CD45
+ immune cells in TCh3 tumors expressed more CXCR6, the only known receptor of CXCL16, upon anti-PD-L1 treatment. CXCL16 may attract CXCR6-expressing naïve CD8 or activated CD8 and CD4 T cells, NK or NKT cells [
83]. On the other hand, CXCL16 was reported to positively correlate with M2-TAM infiltration, increased angiogenesis, and worse prognosis in thyroid cancer [
84]. Hence, the role of different chemokines or cytokines in orchestrating TCh3 TME remains unresolved and needs to be addressed in future studies.
To test the effects of
PIK3CA hyperactivation and
TP53 deletion on tumor-intrinsic cues regulating TME and ICI responses, we employed tumor cell lines derived from in vivo spontaneously generated KPPA SCCs [
22] with these two genetic alterations. Although TAb2 and TCh3 tumors harbor the same oncogenic driver mutations for initial tumorigenesis, they exhibited differential TME and anti-PD-L1 responses, which suggest the limitations of stratifying cancers according to genetic changes and allow us to glance at vast heterogeneity potentially caused by clonal variation in cancers. In this regard, it has been reported that HNSCCs can undergo epigenetic alterations and enhance clonal variation [
85‐
88]. We suggest that the tumor-intrinsic differences in these two cell lines may have arisen during their passage of in vivo transplantation. In line with this idea, our RNA-seq and WES data indicated tumor-specific epigenetic and genetic differences between TAb2 and TCh3 tumors that might contribute to differential responses to anti-PD-L1 by affecting tumor immunogenicity or tumor’s responses to IFN-γ stimulation. Here, we have barely begun to reveal the outcome of such intrinsic differences, it clearly requires substantial work to better understand the fundamental mechanisms defining tumor heterogeneity. Nevertheless, our studies of both tumor cells and tumor-infiltrating immune cells may help to identify new predictive markers associated with anti-PD-L1 responses and our experimental models may facilitate the testing of combinatorial immunotherapy for HNSCCs.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.