METHODS OF PREDICTING AND PREVENTING CANCER IN PATIENTS HAVING PREMALIGNANT LESIONS

Abstract

As advanced cancer has poor prognosis, its detection and treatment at the earliest stages is critical to increase cancer survival rate. Therefore, elucidating the determinants of the intra-lesion immune reaction during cancer's developments is critical for moving into precision medicine and immunotherapy-based cancer prevention. Adaptive immune response within tumors was shown to be the strongest at the earliest stage of carcinoma. Thus, the inventors hypothesized that the immune microenvironment and adaptive immunity were first established at early stage of lung carcinogenesis. Here they identified changes in the tumor molecular profile and its microenvironment during the successive steps of lung squamous carcinogenesis, using gene expression profiling and multispectral imaging. A unique and invaluable dataset of (9) morphological stages of development was analyzed, including (122) well-annotated biopsies from (77) patients. In particular, the inventors show that immune activation and immune escape occur before tumor invasion, and that immunosuppressive cytokines and checkpoint receptors immune escape mechanisms are concomitant with anti-tumor immunity in high-grade dysplasia. Thus, the present invention relates to methods of predicting and preventing cancer in subjects having premalignant lesions.

Claims

1. A method for determining whether a subject having a premalignant lesion is at risk of having a cancer comprising determining the level of at least one immune marker in a biological sample obtained from the subject and wherein the expression level of the immune marker correlates with the risk of having cancer.

2. The method of claim 1 wherein the cancer results from polygenic or multifactorial phenotypes.

3. The method of claim 1 wherein the cancer is a lung cancer.

4. The method of claim 1 wherein the sample is a body fluid sample or a tissue sample.

5. The method of claim 1 wherein the premalignant lesion is a low grade dysplasia, the at least one immune marker is a CD58 and/or a SERPIN and the expression level of the at least one immune marker correlates with the risk of having cancer.

6. The method of claim 5 wherein the premalignant lesion is a low grade bronchial dysplasia, the at least one immune marker is a CD58 and/or a SERPIN and the expression level of the at least one immune marker correlates with the risk of having lung cancer.

7. The method of claim 1 wherein the at least one immune marker comprises CD4 naive T cells and wherein the expression level of the at least one immune marker correlates with the risk of having cancer.

8. The method of claim 7, wherein the low grade dysplasia is a low grade bronchial dysplasia, the at least one immune marker comprises CD4 naive T cells and the expression level of the at least one immune marker correlates with the risk of having lung cancer.

9. The method of claim 1, wherein the at least one immune marker is selected from the group consisting of TNFRSF18 (GITR), IL18, TNFRSF14 (HVEM), TNFSF4, and TNFRSF17 (BCMA).

10. The method of claim 9 wherein the at least one immune marker is selected from the group consisting of TNFRSF18 (GITR), IL18, TNFRSF14 (HVEM), TNFSF4, and TNFRSF17 (BCMA).

11. The method of claim 1, wherein the at least one immune marker is selected from the group consisting of co-inhibitory molecules, co-stimulatory molecules, immunosuppressive interleukins and immunostimulatory interleukins.

12. The method of claim 11, wherein the premalignant lesion is a high grade bronchial dysplasia and the expression level of the at least one immune marker correlates with the risk of having lung cancer.

13. The method of claim 11 wherein the immune marker is: a co-stimulatory molecule selected from the group consisting of CD137, GITR, ICOS, TNFRSF25 and CD86; or a co-inhibitory molecule selected from the group consisting of PDL1, PD1, IDO1, CTLA4, and TIGIT; or an immunostimulatory interleukin selected from the group consisting of IL-18 and IFNG; or an immunosuppressive interleukin selected from the group consisting of IL6, IL10, and TGFβ.

14. (canceled)

15. (canceled)

16. (canceled)

17. The method of claim 1 wherein the step of determining includes detecting the presence of or measuring the amount of messenger RNA (mRNA) transcribed from genomic DNA encoding proteins which are specifically produced by cells from the immune system, or includes detecting the presence of or measuring the amount of proteins expressed by a cell or released in a soluble form.

18. The method of claim 1 wherein the expression level of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 immune markers is determined.

19. The method of claim 18, wherein a score which is a composite of the expression level of the immune markers is determined and compared to the predetermined reference value, and wherein a difference between said score and said predetermined reference value is indicative of whether or not the subject is at risk of having cancer.

20. The method of claim 18 which comprises a) quantifying the expression level of a plurality of immune markers in the sample; b) implementing an algorithm on data comprising the quantified plurality of immune markers so as to obtain an algorithm output; and c) determining the probability that the subject will develop a cancer from the algorithm output of step b).

21. A method for the prophylactic treatment of cancer in a subject having at least one premalignant lesion comprising administering to the subject a therapeutically effective amount of at least one chemopreventive agent.

22. The method of claim 21 wherein the subject has been considered as being at risk of having cancer by the method of claim 1.

23. The method of claim 21 wherein the chemopreventive agent is an immune checkpoint inhibitor.

24. The method of claim 23 wherein the immune checkpoint inhibitor is selected from the group consisting of PD-1 antagonists, PD-L1 antagonists, PD-L2 antagonists, CTLA-4 antagonists, VISTA antagonists, TIM-3 antagonists, LAG-3 antagonists, GITR antagonists, IDO antagonists, KIR2D antagonists, A2AR antagonists, B7-H3 antagonists, B7-H4 antagonists, and BTLA antagonists.

25. The method of claim 21 wherein the chemopreventive agent is an inhibitor of an immunosuppressive cytokine.

26. The method of claim 25 wherein the immunosuppressive cytokine is IL6, IL10, or TGFβ.

27. The method of claim 21 wherein the chemopreventive agent is a vaccine against an immune checkpoint inhibitor or a suppressive cytokine or suppressive protein.

28. The method of claim 27 wherein the vaccine against the immune checkpoint inhibitor includes proteins or peptides of PD-1, PD-L1, PD-L2 CTLA-4, VISTA, TIM-3, LAG-3, GITR, IDO, KIR2D, A2AR, B7-H3, B7-H4, and BTLA.

29. The method of claim 27 wherein the vaccine is against IL6, IL10 or TGFβ.

30. The method of claim 21 wherein the chemopreventive agent is administered by systemic route to the subject on or by local route in the premalignant lesion.

Description

FIGURES

[0116] FIG. 1. Genes encoding form CD58 and SERPIN members had a biphasic increase in low-grade dysplasia.

[0117] FIG. 2. Continuous shift of immune status for CD4 T cells. Significant differences per stage were highlighted with black asterisks at FDR<0.1, or grey otherwise (Mann-Whitney test, *** p<0.001, ** 0.001≥p<0.01, * 0.01≥p<0.05, −0.05≥p<0.1, BH adjustment).

[0118] FIG. 3 Immune evasion before tumor invasion in early squamous lung carcinogenesis. A) Average expression of co-inhibitory molecules and immunosuppressive interleukins increases in high-grade and is further amplified in tumor samples (non-parametric, rank-based test, Dunn's pairwise multiple comparison test, *** p<0.001, ** 0.001≥p<0.01, * 0.01≥p<0.05, −0.05≥p<0.1). B) Positive fold changes in high-grade and SCC compared to their corresponding normal tissue expression were observed for co-stimulatory (e.g. CD137), co-inhibitory (e.g. TIGIT, PDL1), and suppressive interleukins (IL6, IL10). The p-values and coefficient estimates were derived from the linear mixed-effects model of the four molecular steps, adjusted for smoking history and previous cancer status as fixed effects, and patient as random effect.

EXAMPLE

[0119] Methods:

[0120] 1) Study Population

[0121] Bronchial biopsies were collected between 2003 and 2007 at the Jules Bordet Institute, Brussels, Belgium, during fluorescence bronchoscopy in current or former smokers with a smoking exposure of ≥30 pack-years. Former smokers were defined as individuals who had quit smoking for more than 6 months. The study was approved by the ethics committee of the Jules Bordet Institute and the patients gave informed consent. Based on the fact that high-grade lesions were rare and based on Dobbin et al.'s report.sup.31, we included at least twelve biopsies from each histological stage. The histopathological classification was performed by one pathologist (AH) on three independent blinded occasions. Any discordant diagnoses between successive evaluations were re-evaluated by the local team of pathologists using a multi-head microscope to obtain a consensus. Biopsies were classified using the 2004 histological WHO/IASLC classification of pre-invasive and invasive squamous lesions of the bronchus.sup.32. In addition, normal bronchial biopsies from 16 never-smokers were collected and pooled (same amount of RNA for each) for use as reference RNA.

[0122] A total of 122 biopsies from 77 individuals, 35 former and 42 current smokers, were studied. The median age was 62 years (range 42-78). The male/female ratio was 62/15. The 122 biopsies were distributed according to histology and fluorescence status as follows: 13 biopsies with normal histology and normofluorescent (8/5 biopsies from former/current smokers), 14 with normal histology and hypofluorescent (8/6), 15 hyperplasia (7/8), 15 metaplasia (5/10), 13 mild dysplasia (8/5), 13 moderate dysplasia (7/6), 12 severe dysplasia (2/10), 13 carcinoma in situ (CIS) (5/8) and 14 SCC (5/9). Among the 108 biopsies that were not SCC, 6 biopsies were taken in 4 patients having concurrent lung cancer. Among the 122 samples, matched FFPE blocks were found for 110 of them.

[0123] 2) Sample Collection and RNA Extraction

[0124] During bronchoscopy, two biopsies were taken with clean forceps in the same area: one for routine histopathology and the second, immediately dropped in Tripure Isolation Reagent on ice, homogenized and frozen at −80° C. (Roche Diagnostics, Indianapolis, Ind., USA), for molecular studies. RNA extraction protocols have been previously described.sup.27. Isolated RNAs were assessed for quantity and purity on the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Rockland, Del., USA) and for quality on the Agilent 2100 bioanalyser with RNA 6000 NanoAssay (Agilent Technologies, Palo Alto, Calif., USA). RNA was successfully extracted from 122 fresh frozen biopsies. The median yield of total RNA extracted from the biopsies was 1275 ng (range 244-11000 ng).

[0125] 3) Acquisition and Analysis of Gene Expression Profiles

[0126] After amplification and labelling, cRNAs were hybridized on two Colours Whole Human Genome 4×44K arrays according to the recommendation of the provider (Agilent Technologies) (details in Text Si). Additional normalization steps were performed with Genespring GX, version 7.3.1, software (Agilent Technologies): 1) per spot (divide by control channel), 2) per chip (normalize to the median expression value across chip) and 3) per gene (normalize to median expression value across patients).

[0127] Several steps of data quality control were performed. Random principal component analysis showed that there were no outliers among the samples. The gene expression measurements in the cohort followed Gaussian distribution.

[0128] 4) Identification of Linear Gene Expression Changes and Molecular Phenotypes

[0129] Monotonic gene expression alterations associated with developmental stages were identified using a linear model with mixed-effects. Each gene was modeled as a function of the developmental stage (factor variable), adjusting for smoking status, gender, and history of cancer as fixed effects. Because patient-level observations are not independent, we considered the parameter patient as a random effect. ANOVA tests compared the association of a gene and developmental stage to a null model. The false discovery rate (FDR) was calculated for each ANOVA.sup.33 p-value using the method of Benjamini and Hochberg.sup.34,35. Genes significantly associated with developmental stages was determined by an ANOVA FDR<0.001. Semi-supervised hierarchical clustering of these genes was then used to compare the nine different developmental stages.

[0130] 5) Definition and Functional Characterization of Gene Modules

[0131] To identify trajectories of gene expression during development, we applied a WGCNA.sup.36 on the genes significantly associated with developmental stages. WGCNA network construction and module detection was done using and signed network type, soft-thresholding power of 12, and a dendrogram cut height of 0.3 for merging modules. A minimum cluster size of 50 genes was used to define a module. A p-value ratio threshold of 0 was considered for reassigning genes across modules. The cluster eigengene (the first principal component of a cluster) value was used to evaluate the association of each module with the 9 stages of cancer. Thereby, we determined gene clusters (modules) of highly correlated genes with similar expression patterns across the nine developmental stages.

[0132] To functionally describe the gene modules, we used the cancer hallmark definitions from the mSigDB database.sup.37 and applied the over-representation hypergeometric test using the R package clusterProfiler.sup.38. In addition, we also used single-sample gene set enrichment analysis (ssGSEA).sup.39 on the full gene expression assay to determine whether the cancer hallmark genesets were enriched among the up-regulated or down-regulated genes within a sample (regardless of gene module). Probes were mapped to unique Entrez IDs. The genes were ranked by their z-score transformed expression values per sample. A minimum overlap of 5 genes with a given geneset was required. The enrichment score represents the degrees to which the genes from a given cancer hallmark geneset were up- or down-regulated within a sample.

[0133] 6) Immune Cell-Type Signatures

[0134] To explore a large number of different immune cell subtypes and, even more, to examine their activation status, we compiled a large number of carefully annotated microarray gene expression profiles from almost 2000 publicly available experiments normalized with the frozen Robust Multi-array Averaging (fRMA) method. Building on our previous methods.sup.3,40,41, we identified genes with specific expression for immune cell types, considering their naive, resting, and activated status (manuscript in preparation). Pan-cell type signatures were defined as genes expressed at similar levels in multiple cell types. For examples, the Myeloid-derived category comprised all subtypes of dendritic cells, eosinophils, monocytes, macrophages, neutrophils, and mast cells, while Macrophages-DC was a gene signature comprising common genes expressed in all studied subtypes of both macrophages and dendritic cells.

[0135] 7) Immune Characterization from Gene-Expression Profiles

[0136] The defined immune signatures were used to explore a large variety immune cell types from the gene expression data at different histological stages of SCC development. First, we performed a hypergeometric test between the immune signatures and the gene modules, to pinpoint potential evolutionary trajectories of specific immune cell types.

[0137] We next applied the algorithm for absolute quantification implemented in CIBERSORT.sup.42 and deconvolved immune cell types expression from a mixed gene-expression signal according to the predefined LM22 signature.

[0138] Last, we performed single-sample GeneSet Enrichment Analysis (ssGSEA) using the in-house defined immune gene signatures. Thereby, for each immune cell type, we obtained an enrichment score per sample indicating the extent of up-regulation or down-regulation of the associated genes. The probe IDs were mapped to unique Entrez IDs. A minimum overlap of 5 genes was required.

[0139] 8) Multiplex Immunohistochemistry and Multispectral Image Analysis

[0140] Matched formalin-fixed paraffin-embedded (FFPE) blocks of the 122 fresh frozen samples were available for 110 samples. Two four-μm tick slides were cut from the FFPE blocks, deparaffinized in clarene, rehydrated through an ethanol gradient and fixed in NBF (10% neutral buffered formalin). Slides were then stained according to the Opal 7-plex technology of PerkinElmer allowing the simultaneous visualization of 6 markers on the same slide. Therefore, at each of the 6 cycles of staining, antigen retrieval was performed via microwave treatment (MWT) in antigen retrieval solution pH6 or pH9 (AR6 or AR9) depending on the target, protein blocking was performed using Protein Block-Serum-free (Dako) for 15 min, and primary Abs were then incubated for 30 min at RT. Next, incubation with HRP Labelled Polymer mouse or rabbit (Dako EnVision+ System-HRP Labelled Polymer) was performed at room temperature for 15 min followed by TSA opal fluorophores (Opal 520, Opal 540, Opal 570, Opal 620, Opal 650 or Opal 690) incubation for 10 min. MWT was performed at each cycle of staining to remove the Ab TSA complex with AR solution (pH 9 or 6). At last, all slides were counterstained with DAPI for 5 min and enclosed in ProLong Diamond Antifade Mountant (Thermofisher). The slides were scanned using the PerkinElmer Vectra 3 System and the multispectral images obtained were unmixed using spectral libraries previously built from images stained for each fluorophore (monoplex), using the inForm Advanced Image Analysis software (inForm 2.3.0 PerkinElmer). A selection of representative multispectral images belonging to different samples was used to train the inForm software for tissue segmentation, cell segmentation, and phenotyping, and finally, the settings applied to the training images were saved within an algorithm allowing batch analysis of all the tissue slides. We designed two different 7-plex panels defined as phenotype and functional panels, which were used on 2 sequential slides in order to characterize the immune microenvironment of pre-cancer lesions of the lung, including (in)activated cells, (in)activated immune pathways and immune response type. The phenotype panel included CD3, CD8, FoxP3, CD68, Neutrophil elastase (NE), DAPI, and Cytokeratin (CK) and the functional panel included: CD3, PD-L1, PD1, Ki67, CD137, DAPI, and CK.

[0141] 9) Spatial Statistics

[0142] We performed first- and second-order spatial analysis of multispectral imaging data, which enables a high-definition characterization of the microenvironment architecture. First, we reconstructed whole slides, rather than separate analysis of each image that introduces edge effects and leads to loss of information. We calculated immune cell densities as the number of positive cells per unit of tissue surface area (mm2). Based on the tissue categorization performed with the inForm software, the stroma and the epithelium compartments were annotated on the images, enabling densities and spatial distribution to be calculated individually for the stromal and epithelial tissue category. To compare the spatial localization of different immune cell types, we calculated the distances to the nearest neighbors and their distribution implementing edge corrections, G(r). The function G(r) is the cumulative distribution of the distance from a typical random cell X to its nearest cell Y, where the argument r is the radius of the area in which G(r) is evaluated. Deviations from the empirical and the theoretical G(r) function indicate clustered and dispersed patterns.

[0143] 7) Statistics

[0144] The R statistical software (v 3.3.3) was used for statistical analyses and graphical visualization. The null hypotheses were rejected at p-values lower than 0.05, unless indicated otherwise. When comparing tumor- to normal-tissue gene expression, linear mixed-effects model was used to adjust for the confounding factors smoking history, previous caner, between-patient variability, gender, and age. The Benjamini-Hochberg method.sup.34,35 was applied for multiple testing correction. Post-hoc multiple testing correction was applied for pairwise comparison using Dunn's test.

[0145] Results:

[0146] Despite the developments in targeted therapies and immunotherapy, advanced lung cancer remains incurable.sup.4. There are estimates that US lung cancer deaths could be reduced to more than 70,000 per year by early diagnosis and treatment.sup.1. Recently, the Nelson volume CT screening trials showed a reduction of lung cancer mortality by 26% in men and 39-61% in women.sup.5. Beyond and prior to its early detection, cancer prevention may significantly reduce cancer burden.sup.6. It is critical to understand the mechanisms underlying lung carcinogenesis, to decipher the role of the microenvironment in early lesion, in order to move into precision medicine including immunotherapy for cancer prevention.sup.2. In smokers, a range of successive developmental stages precedes invasive lung squamous cell carcinoma.sup.7 (SCC), making this cancer a convenient model to mechanistically study how cancer develops. However, the rarity of pre-invasive lesion collections explains the limited knowledge of their molecular and immune profiles.sup.8. Using gene expression profiling and multispectral imaging, we sought to identify the changes in the tumor and its microenvironment during the successive steps of lung squamous carcinogenesis.

[0147] We examined a rare dataset of nine morphological stages of lung squamous carcinogenesis, consisting of 122 carefully annotated biopsies from 77 patients (data not shown). Using gene expression profiling, we first identified 7739 genes associated with the nine histological stages of development (linear mixed-effects model, FDR<0.001). Four distinct and successive molecular steps of progression were revealed by semi-supervised hierarchical clustering of the selected genes (data not shown). The first step included normal non-fluorescent and fluorescent biopsies as well as hyperplasia (normal bronchial tissue); the second comprised of metaplasia, mild dysplasia and moderate dysplasia (low-grade); the third combined both severe dysplasia and in situ carcinoma (CIS) (high-grade), while the fourth segregated invasive (SCC) from premalignant lesions (data not shown).

[0148] Carcinogenesis has been described as the process of acquiring advantageous biological capabilities, cancer hallmarks, by the abnormal cells.sup.9. We first isolated modules of genes with specific expression patterns and then searched for significant associations with cancer hallmarks (hypergeometric test, data not shown). Seven evolutionary trajectories of gene expression were discerned by seven gene modules derived from weighted gene co-expression network analysis (WGCNA, data not shown). The two largest modules exhibited linear evolution from normal tissue to cancer, Ascending (n=1848), associated with proliferation and Descending (n=939), linked to genes that are down-regulated in DNA repair (i.e. UV response), suggesting a continuous activation of DNA damage response. A module of 150 genes displayed late expression increase starting from high-grade lesions (High-grade ascending). Interestingly, this module was highly enriched with genes involved in immune response. A set of genes remained unmodified until cancer onset (SCC ascending, n=51). This increase of expression specific to SCC was over-represented by genes involved in epithelial-mesenchymal transition (EMT). The CXCL12-CXCR4 axis known to promote the EMT process, revealed a very low expression of CXCL12, and a significant increase of CXCR4 expression in SCC (data not shown). Two additional modules had biphasic gene expression evolutions, both reaching a peak of expression in low-grade (Biphasic 1, n=164 and Biphasic 2, n=64) (FIG. 1). Here, we found that metabolism regulation had a biphasic trajectory. Specifically, genes involved in fatty acid metabolism, oxidative phosphorylation and citric acid cycle had a transitory increase of expression in low-grade (Biphasic 1).

[0149] To analyze the evolutionary trajectory of immune response, we first compiled genes representing specific immune, stromal, and cancer cell types and matched them to each gene module. We confirmed the highest percentage of immune-related genes in the module High-grade ascending along with a significant under-representation in the linearly decreasing module (both p<0.001, Fisher's exact test, data not shown). Cancer-germline antigens were found in the Ascending module at a significantly higher number than expected (FDR<0.05), along with an over-representation trend of genes involved in neutrophil activation (FDR<0.1, data not shown). Both observations suggest immune sensing at the earliest steps of transformation. Markedly, increased gene expression representing activated T cells was detected in high-grade lesions before tumor invasion, with the same pattern as total neutrophils, M1 macrophages, and overall, the myeloid signature.

[0150] We then estimated the absolute abundance of different immune cell types using a method for deconvolving cell composition of complex tissues from gene expression (data not shown). We confirmed an increase of myeloid-derived cells, neutrophils and macrophage subtypes in high-grade dysplasia (data not shown). Additionally, we observed co-regulation of immune cells from both the innate and adaptive immunity based on correlation of the immune cell abundances (data not shown). Activated T cells (CD4 memory), macrophages (M0), memory B cells, follicular T-helper cells, and dendritic cells followed the same abundance pattern. Interestingly, lesions within the same patient had different immune composition at different developmental stages (data not shown). We also detected a significant shift in the immune status, from resting or naive to activated or memory (data not shown). Resting mast cells were more abundant in the early compared to the late developmental stages, while the activated mast cells followed the opposite pattern (data not shown). A drop of naive B cell abundance was accompanied by an increase of memory B cells. An influx of naive CD4 cells was observed already at the stage of mild dysplasia (stage 4), followed by a sudden decline of naive CD4 abundance and a concurrent increase of activated CD4 memory T cells in the successive stages (FIG. 2).

[0151] To further elucidate the immune transition at each molecular step of transformation, we performed functional analysis of the differentially regulated genes in transformed compared to normal tissues. Accounting for smoking history, previous cancer status, and intra-patient variability as confounding factors, we identified Gene Ontology (GO) immune processes enriched among the differentially regulated genes in low-grade, high-grade, and SCC (linear mixed-effects model, FDR<0.05, data not shown). Few immune functions were specifically modulated for low-grade, not only among up-(n=5) but also among down-regulated genes (n=13, e.g. response to TGFβ). Unlike low-grade, a large number of immune functions were enriched only among the up-regulated genes in high-grade (n=148) and SCC (n=240). Strikingly, negative regulation of the immune system was implicated in all developmental stages, in addition to antigen processing and presentation of peptide antigen (data not shown). Nevertheless, in low-grade, the genes associated with negative regulation were significantly down-regulated, while, in high-grade and SCC, they were up-regulated. Therefore, one of the early immune reactions is immune unleashing by down-regulation of the genes that negatively regulate the immune system such as HVEM (TNFRSF14), CD200, CD59, TGFB3, and HLA-G. Reversely, in high-grade and SCC, there was an up-regulation of genes involved in immunosuppression.

[0152] Closer examination of immunomodulatory gene expression revealed that the average expression of co-inhibitory molecules and suppressive interleukins was significantly higher in severe dysplasia (stage 6) and in the succeeding stages (FIG. 3A). Particularly, the expression of PDL1, PD1, IDO1, CTLA4, and TIGIT marked an increase at the transition point to severe dysplasia (stage 6). Similar evolution patterns were observed for the suppressive interleukins, including IL6, IL10, and TGFβ, with elevated expression at the transition from moderate dysplasia (stage 5). Overall, many immunomodulatory molecules had a positive fold change in high-grade dysplasia compared to normal tissue (FIG. 31B). Not only suppressive molecules (IDO1, PDL1, TIGIT, CTLA4, ICOS, IL10, and IL6) but also stimulatory molecules (CD137, GITR, ICOS, CD80, CD86, CD70, CD137L, TNFRSF25) had increased expression in high-grade (linear mixed-effects model, Benjamini and Hochberg (BH)) and to a greater extent at the invasive stage. Collectively, immune escape occurred before tumor invasion as co-inhibitors and suppressive interleukins increased significantly from high-grade stages onwards.

[0153] For high-definition characterization of the microenvironment architecture, we used two 7-plex staining panels in FFPE blocks from the same bronchial epithelial lesions, a phenotype panel to determine the nature of the immune cells and a functional panel including PD1, PD-L1, Ki67, and CD137, in addition to CD3, Cytokeratin (CK) and DAPI (n=110 and 106, respectively, data not shown). First, we calculated immune cell densities as the number of positive cells per unit of tissue surface area (mm.sup.2), individually for the stromal and epithelial tissue category (data not shown). Overall, we found a relatively large variation in the immune cell densities. However, we observed significant differences among the four developmental stages in the stromal compartment and the same sustained trends in the epithelial compartment (data not shown). CD4 T cells (i.e. CD3.sup.+CD8.sup.−) and CD8.sup.+ lymphocytes both had a transitory increase in high-grade pre-invasive lesions (p<0.01). Consistent with the immune gene expression evolution, myeloid, neutrophil, and macrophage densities increased in high-grade's stroma (p<0.05, FDR<0.1) and epithelium (p<0.1 before BH correction). In accordance with gene expression, PD-L1 (PD-L1.sup.+CK.sup.−) densities significantly increased in high-grade lesions and even more in SCC (p<0.05) (data not shown), similarly to CD137, which did not reach statistical significance. Cells with the CD137, PD-L1, and CD3.sup.+FoxP3.sup.+ phenotype were rarely found in the epithelium at early developmental stages (i.e. stage 0-5, normal and low-grade).

[0154] We next performed second-order spatial statistics and measured distances between each pair of cell phenotypes. We calculated a cross-type cumulative distribution of the nearest neighbor distances, G(r) (data not shown). We expected a potential interaction when two cells were within a distance of 25 μm. By comparison of the observed empirical function G.sub.X,Y(r) to the theoretical curve G.sup.theo.sub.X,Y(r) that shows random sample distribution, we detected segregation among epithelial cells (CK) and CD3, consistently in both panels (p<0.001, FDR<0.1, data not shown). In particular, we observed a lower number of epithelial cells than expected near CD3 cells in high-grade (data not shown). This pattern was observed for all CK.sup.+ cells in the functional panel, total epithelial cells (all CK.sup.+), CK.sup.+PD-L1*, and CK.sup.+Ki67.sup.+ (p<0.01, FDR<0.1, data not shown). Therefore, in high-grade, we discerned reconfiguration of the tumor microenvironment compared to the preceding stages of development, manifested by segregation of epithelial cells from CD3 cells.

[0155] This report shows that both immune activation and immune suppression occur at pre-invasive stages, which reinforces the use of immunotherapy at the earliest steps of treatment and underlines its potential role in chemopreventive approaches. The prognostic impact of immune infiltrates has been demonstrated in various cancer types.sup.10-12 from early stage.sup.13, including lung cancer.sup.14 from stage I.sup.15. Tumor intrinsic factors modestly contributed to the risk of carcinogenesis.sup.16, as compared to extrinsic carcinogen.sup.16 or dysregulation of the immune microenvironment.sup.17. We previously showed that the tumor microenvironment was a critical determinant of dissemination to distant metastasis.sup.18 and of metastatic tumor development, where tumor evolution could be traced back to immune escaping clones.sup.17. These findings could also apply to the pre-malignant transformation and the initiation of carcinoma. Furthermore, major clinical benefit of checkpoint immunotherapy was obtained in various settings of cancer treatment. In non-small cell lung cancer (NSCLC), checkpoint inhibitors are now a standard of care as first-line.sup.19,20 and second-line treatment options.sup.21,22,23 for advanced disease and as maintenance after curative chemo-radiation of locally advanced stages.sup.24. However, up to now, the best opportunity to cure lung cancer patients is early intervention. The positive results of immune checkpoint blockade therapy in adjuvant setting for melanoma.sup.25 and in neoadjuvant setting for lung cancer.sup.26 fortify the importance of using immunotherapy in the early steps of treatment strategies.

[0156] Our study delineated the molecular pathways involved in the four steps of lung squamous cell carcinogenesis (data not shown), whereby the earliest molecular changes affected proliferation and metabolism. A transient influx of naive T cell was observed in low-grade, a pattern previously described for miRNA expression in a subset of the same preneoplastic lesions.sup.21. Collectively, the immune transition unfolds as follows 1) immune sensing and immune unleashing are induced at the earliest step of transformation; 2) continual cell proliferation fosters accumulation of somatic mutations mounting an anti-tumor immune response and, correspondingly, 3) triggering inherent immune suppression mechanisms already in high-grade pre-cancer. Historically, studies have shown that the risk of cancer progression is much higher in high-grade (32-87%) compared to low-grade lesions (2-9%).sup.28-30. Altogether, our results urge to assess the role of immunotherapy and chemoprevention in high-risk individuals for lung cancer.

REFERENCES

[0157] Throughout this application, various references describe the state of the art to which this invention pertains. The disclosures of these references are hereby incorporated by reference into the present disclosure. [0158] 1. Goldberg, S. W., Mulshine, J. L., Hagstrom, D. & Pyenson, B. S. An actuarial approach to comparing early stage and late stage lung cancer mortality and survival. Popul. Health Manag. 13, 33-46 (2010). [0159] 2. Kensler, T. W. et al. Transforming Cancer Prevention through Precision Medicine and Immune-oncology. Cancer Prev. Res. Phila. Pa. 9, 2-10 (2016). [0160] 3. Bindea, G. et al. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity 39, 782-795 (2013). [0161] 4. Herbst, R. S., Morgensztern, D. & Boshoff, C. The biology and management of non-small cell lung cancer. Nature 553, 446-454 (2018). [0162] 5. De Koning, H., Van Der Aalst, C., Ten Haaf, K. & Oudkerk, M. PLO2.05 Effects of Volume CT Lung Cancer Screening: Mortality Results of the NELSON Randomised-Controlled Population Based Trial. J. Thorac. Oncol. 13, S185 (2018). [0163] 6. Umar, A., Dunn, B. K. & Greenwald, P. Future directions in cancer prevention. Nat. Rev. Cancer 12, 835-848 (2012). [0164] 7. Slaughter, D. P., Southwick, H. W. & Smejkal, W. Field cancerization in oral stratified squamous epithelium; clinical implications of multicentric origin. Cancer 6, 963-968 (1953). [0165] 8. Kerr, K. M. Pulmonary preinvasive neoplasia. J. Clin. Pathol. 54, 257-271 (2001). [0166] 9. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646-674 (2011). [0167] 10. Galon, J. Type, Density, and Location of Immune Cells Within Human Colorectal Tumors Predict Clinical Outcome. Science 313, 1960-1964 (2006). [0168] 11. Galon, J., Angell, H. K., Bedognetti, D. & Marincola, F. M. The Continuum of Cancer Immunosurveillance: Prognostic, Predictive, and Mechanistic Signatures. Immunity 39, 11-26 (2013). [0169] 12. Pages, F. et al. International validation of the consensus Immunoscore for the classification of colon cancer: a prognostic and accuracy study. The Lancet 391, 2128-2139 (2018). [0170] 13. Pages, F. et al. In Situ Cytotoxic and Memory T Cells Predict Outcome in Patients With Early-Stage Colorectal Cancer. J. Clin. Oncol. 27, 5944-5951 (2009). [0171] 14. Fridman, W. H., Pages, F., Sautes-Fridman, C. & Galon, J. The immune contexture in human tumours: impact on clinical outcome. Nat. Rev. Cancer 12, 298-306 (2012). [0172] 15. Gentles, A. J. et al. The prognostic landscape of genes and infiltrating immune cells across human cancers. Nat. Med. 21, 938-945 (2015). [0173] 16. Wu, S., Powers, S., Zhu, W. & Hannun, Y. A. Substantial contribution of extrinsic risk factors to cancer development. Nature 529, 43-47 (2016). [0174] 17. Angelova, M. et al. Evolution of Metastases in Space and Time under Immune Selection. Cell 175, 751-765.e16 (2018). [0175] 18. Mlecnik, B. et al. The tumor microenvironment and Immunoscore are critical determinants of dissemination to distant metastasis. Sci. Transl. Med. 8, 327ra26 (2016). [0176] 19. Gandhi, L. et al. Pembrolizumab plus Chemotherapy in Metastatic Non-Small-Cell Lung Cancer. N. Engl. J. Med. 378, 2078-2092 (2018). [0177] 20. Reck, M. et al. Pembrolizumab versus Chemotherapy for PD-L1-Positive Non-Small-Cell Lung Cancer. N. Engl. J. Med. 375, 1823-1833 (2016). [0178] 21. Brahmer, J. et al. Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer. N. Engl. J. Med. 373, 123-135 (2015). [0179] 22. Borghaei, H. et al. Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell Lung Cancer. N. Engl. J. Med. 373, 1627-1639 (2015). [0180] 23. Herbst, R. S. et al. Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial. Lancet Lond. Engl. 387, 1540-1550 (2016). [0181] 24. Antonia, S. J. et al. Overall Survival with Durvalumab after Chemoradiotherapy in Stage III NSCLC. N. Engl. J. Med. (2018). doi:10.1056/NEJMoa1809697 [0182] 25. Eggermont, A. M. M. et al. Adjuvant Pembrolizumab versus Placebo in Resected Stage III Melanoma. N. Engl. J. Med. 378, 1789-1801 (2018). [0183] 26. Forde, P. M. et al. Neoadjuvant PD-1 Blockade in Resectable Lung Cancer. N. Engl. J. Med. 378, 1976-1986 (2018). [0184] 27. Mascaux, C. et al. Evolution of microRNA expression during human bronchial squamous carcinogenesis. Eur. Respir. J. 33, 352-359 (2009). [0185] 28. Bota, S. et al. Follow-up of Bronchial Precancerous Lesions and Carcinoma in Situ Using Fluorescence Endoscopy. Am. J. Respir. Crit. Care Med. 164, 1688-1693 (2001). [0186] 29. Breuer, R. H. et al. The Natural Course of Preneoplastic Lesions in Bronchial Epithelium. 8 [0187] 30. Venmans, B. J. W., van Boxem, T. J. M., Smit, E. F., Postmus, P. E. & Sutedja, T. G. Outcome of Bronchial Carcinoma In Situ. Chest 117, 1572-1576 (2000). [0188] 31. Dobbin, K., Shih, J. H. & Simon, R. Questions and Answers on Design of Dual-Label Microarrays for Identifying Differentially Expressed Genes. JNCI J. Natl. Cancer Inst. 95, 1362-1369 (2003). [0189] 32. Tumours of the Lung, Pleura, Thymus and Heart. in Pathology & Genetics (eds. Travis W. D., Brambilla E., Muller-Hermelink H. K. & Harris C.C.) (IARC Press: Lyon, 2004). [0190] 33. Kerr, M. K., Martin, M. & Churchill, G. A. Analysis of variance for gene expression microarray data. J. Comput. Biol. J. Comput. Mol. Cell Biol. 7, 819-837 (2000). [0191] 34. Qian, H.-R. & Huang, S. Comparison of false discovery rate methods in identifying genes with differential expression. Genomics 86, 495-503 (2005). [0192] 35. Ge, Y., Sealfon, S. C. & Speed, T. P. Some step-down procedures controlling the false discovery rate under dependence. Stat. Sin. 18, 881-904 (2008). [0193] 36. Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008). [0194] 37. Liberzon, A. et al. The Molecular Signatures Database Hallmark Gene Set Collection. Cell Syst. 1, 417-425 (2015). [0195] 38. Yu, G., Wang, L.-G., Han, Y. & He, Q.-Y. clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. OMICS J. Integr. Biol. 16, 284-287 (2012). [0196] 39. Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. 102, 15545-15550 (2005). [0197] 40. Angelova, M. et al. Characterization of the immunophenotypes and antigenomes of colorectal cancers reveals distinct tumor escape mechanisms and novel targets for immunotherapy. Genome Biol. 16, 64 (2015). [0198] 41. Newman, A. M. et al. Robust enumeration of cell subsets from tissue expression profiles. Nat. Methods 12, 453-457 (2015).