Fat1 deletion promotes hybrid EMT state, tumour stemness and metastasis.

FAT1, which encodes a protocadherin, is one of the most frequently mutated genes in human cancers1-5. However, the role and the molecular mechanisms by which FAT1 mutations control tumour initiation and progression are poorly understood. Here, using mouse models of skin squamous cell carcinoma and lung tumours, we found that deletion of Fat1 accelerates tumour initiation and malignant progression and promotes a hybrid epithelial-to-mesenchymal transition (EMT) phenotype. We also found this hybrid EMT state in FAT1-mutated human squamous cell carcinomas. Skin squamous cell carcinomas in which Fat1 was deleted presented increased tumour stemness and spontaneous metastasis. We performed transcriptional and chromatin profiling combined with proteomic analyses and mechanistic studies, which revealed that loss of function of FAT1 activates a CAMK2-CD44-SRC axis that promotes YAP1 nuclear translocation and ZEB1 expression that stimulates the mesenchymal state. This loss of function also inactivates EZH2, promoting SOX2 expression, which sustains the epithelial state. Our comprehensive analysis identified drug resistance and vulnerabilities in FAT1-deficient tumours, which have important implications for cancer therapy. Our studies reveal that, in mouse and human squamous cell carcinoma, loss of function of FAT1 promotes tumour initiation, progression, invasiveness, stemness and metastasis through the induction of a hybrid EMT state.

Regulatory T cells constrain the TCR repertoire of antigen-stimulated conventional CD4 T cells.

To analyze the potential role of Tregs in controlling the TCR repertoire breadth to a non-self-antigen, a TCRß transgenic mouse model (EF4.1) expressing a limited, yet polyclonal naïve T-cell repertoire was used. The response of EF4.1 mice to an I-Ab-associated epitope of the F-MuLV envelope protein is dominated by clones expressing a Vα2 gene segment, thus allowing a comprehensive analysis of the TCRα repertoire in a relatively large cohort of mice. Control and Treg-depleted EF4.1 mice were immunized, and the extent of the Vα2-bearing, antigen-specific TCR repertoire was characterized by high-throughput sequencing and spectratyping analysis. In addition to increased clonal expansion and acquisition of effector functions, Treg depletion led to the expression of a more diverse TCR repertoire comprising several private clonotypes rarely observed in control mice or in the pre-immune repertoire. Injection of anti-CD86 antibodies in vivo led to a strong reduction in TCR diversity, suggesting that Tregs may influence TCR repertoire diversity by modulating costimulatory molecule availability. Collectively, these studies illustrate an additional mechanism whereby Tregs control the immune response to non-self-antigens.

The Prognostic Value of the Combination of Low VEGFR-1 and High VEGFR-2 Expression in Endothelial Cells of Colorectal Cancer.

Research on tumor angiogenesis has mainly focused on the vascular endothelial growth factor (VEGF) family and on methods to block its actions. However, reports on VEGF receptor (VEGFR) expression in tumor-associated endothelial cells (ECs) are limited. Thus, we evaluated VEGF, VEGFR-1 and VEGFR-2 expression in ECs of colorectal cancer (CRC) using immunohistochemistry. VEGF, VEGFR-1 and -2 expression in ECs was quantitatively evaluated by digital image analysis in a retrospective series of 204 tumor tissue samples and related to clinical variables. The data show that the VEGF, VEGFR-1 and VEGFR-2 expression in ECs is heterogeneous. Multivariate analysis including a set of clinicopathological variables reveals that high EC VEGFR-1 expression is an independent prognostic factor for overall survival (OS). The combination of low VEGFR-1 and high VEGFR-2 expression in ECs outperforms models integrating VEGFR-1 and VEGFR-2 as separate markers. Indeed, this VEGFR-1_VEGFR-2 combination is an independent negative prognostic factor for OS (p = 0.012) and metastasis-free survival (p = 0.007). In conclusion, this work illustrates the importance of studying the distribution of VEGF members in ECs of CRC. Interestingly, our preliminary data suggest that high VEGFR-1 and low VEGFR-2 expression in ECs appear to be involved in the progression of CRC, suggesting that targeting EC VEGFR-1 could offer novel opportunities for CRC treatment. However, a prospective validation study is needed.

Initial Condition Assessment for Reaction-Diffusion Glioma Growth Models: A Translational MRI-Histology (In)Validation Study.

Reaction-diffusion models have been proposed for decades to capture the growth of gliomas. Nevertheless, these models require an initial condition: the tumor cell density distribution over the whole brain at diagnosis time. Several works have proposed to relate this distribution to abnormalities visible on magnetic resonance imaging (MRI). In this work, we verify these hypotheses by stereotactic histological analysis of a non-operated brain with glioblastoma using a 3D-printed slicer. Cell density maps are computed from histological slides using a deep learning approach. The density maps are then registered to a postmortem MR image and related to an MR-derived geodesic distance map to the tumor core. The relation between the edema outlines visible on T2-FLAIR MRI and the distance to the core is also investigated. Our results suggest that (i) the previously proposed exponential decrease of the tumor cell density with the distance to the core is reasonable but (ii) the edema outlines would not correspond to a cell density iso-contour and (iii) the suggested tumor cell density at these outlines is likely overestimated. These findings highlight the limitations of conventional MRI to derive glioma cell density maps and the need for other initialization methods for reaction-diffusion models to be used in clinical practice.

The potential of tumour microenvironment markers to stratify the risk of recurrence in prostate cancer patients.

The tumour micro-environment (TME) plays a crucial role in the onset and progression of prostate cancer (PCa). Here we studied the potential of a selected panel of TME-markers to predict clinical recurrence (CLR) in PCa. Patient cohorts were matched for the presence or absence of CLR 5 years post-prostatectomy. Tissue micro-arrays (TMA) were composed with both prostate non-tumour (PNT) and PCa tissue and subsequently processed for immunohistochemistry (IHC). The IHC panel included markers for cancer activated fibroblasts (CAFs), blood vessels and steroid hormone receptors ((SHR): androgen receptor (AR), progesterone receptor (PR) and estrogen receptor (ER)). Stained slides were digitalised, selectively annotated and analysed for percentage of marker expression with standardized and validated image analysis algorithms. A univariable analysis identified several TME markers with significant impact on CR: expression of CD31 (vascular marker) in PNT stroma, expression of alpha smooth muscle actin (αSMA) in PCa stroma, and PR expression ratio between PCa stroma and PNT stroma. A multivariable model, which included CD31 expression (vascular marker) in PNT stroma and PR expression ratio between PCa stroma and PNT stroma, could significantly stratify patients for CLR, with the identification of a low risk and high-risk subgroup. If validated and confirmed in an independent prospective series, this subgroup might have clinical potential for PCa patient stratification.

A Novel Approach for Quantifying Cancer Cells Showing Hybrid Epithelial/Mesenchymal States in Large Series of Tissue Samples: Towards a New Prognostic Marker.

In cancer biology, epithelial-to-mesenchymal transition (EMT) is associated with tumorigenesis, stemness, invasion, metastasis, and resistance to therapy. Evidence of co-expression of epithelial and mesenchymal markers suggests that EMT should be a stepwise process with distinct intermediate states rather than a binary switch. In the present study, we propose a morphological approach that enables the detection and quantification of cancer cells with hybrid E/M states, i.e., which combine partially epithelial (E) and partially mesenchymal (M) states. This approach is based on a sequential immunohistochemistry technique performed on the same tissue section, the digitization of whole slides, and image processing. The aim is to extract quantitative indicators able to quantify the presence of hybrid E/M states in large series of human cancer samples and to analyze their relationship with cancer aggressiveness. As a proof of concept, we applied our methodology to a series of about a hundred urothelial carcinomas and demonstrated that the presence of cancer cells with hybrid E/M phenotypes at the time of diagnosis is strongly associated with a poor prognostic value, independently of standard clinicopathological features. Although validation on a larger case series and other cancer types is required, our data support the hybrid E/M score as a promising prognostic biomarker for carcinoma patients.

Strategies to Reduce the Expert Supervision Required for Deep Learning-Based Segmentation of Histopathological Images.

The emergence of computational pathology comes with a demand to extract more and more information from each tissue sample. Such information extraction often requires the segmentation of numerous histological objects (e.g., cell nuclei, glands, etc.) in histological slide images, a task for which deep learning algorithms have demonstrated their effectiveness. However, these algorithms require many training examples to be efficient and robust. For this purpose, pathologists must manually segment hundreds or even thousands of objects in histological images, i.e., a long, tedious and potentially biased task. The present paper aims to review strategies that could help provide the very large number of annotated images needed to automate the segmentation of histological images using deep learning. This review identifies and describes four different approaches: the use of immunohistochemical markers as labels, realistic data augmentation, Generative Adversarial Networks (GAN), and transfer learning. In addition, we describe alternative learning strategies that can use imperfect annotations. Adding real data with high-quality annotations to the training set is a safe way to improve the performance of a well configured deep neural network. However, the present review provides new perspectives through the use of artificially generated data and/or imperfect annotations, in addition to transfer learning opportunities.

Segmentation of glandular epithelium in colorectal tumours to automatically compartmentalise IHC biomarker quantification: A deep learning approach.

In this paper, we propose a method for automatically annotating slide images from colorectal tissue samples. Our objective is to segment glandular epithelium in histological images from tissue slides submitted to different staining techniques, including usual haematoxylin-eosin (H&E) as well as immunohistochemistry (IHC). The proposed method makes use of Deep Learning and is based on a new convolutional network architecture. Our method achieves better performances than the state of the art on the H&E images of the GlaS challenge contest, whereas it uses only the haematoxylin colour channel extracted by colour deconvolution from the RGB images in order to extend its applicability to IHC. The network only needs to be fine-tuned on a small number of additional examples to be accurate on a new IHC dataset. Our approach also includes a new method of data augmentation to achieve good generalisation when working with different experimental conditions and different IHC markers. We show that our methodology enables to automate the compartmentalisation of the IHC biomarker analysis, results concurring highly with manual annotations.

Comparing the expression profiles of steroid hormone receptors and stromal cell markers in prostate cancer at different Gleason scores.

The recent developments in anti-angiogenic and immunomodulatory drugs show that the tumour micro-environment (TME) becomes increasingly important in cancer research. Here we investigated the correlation between the Gleason score (GS) and the TME by comparing tissue expression profiles of steroid hormone receptors, cancer activated fibroblast (CAF) markers and vessel densities between different GS groups. Therefore, matched patient cohorts were composed for different GS (6-7-8). Tissue micro-arrays with 6 samples/patient were processed for immunohistochemistry. Stained slides were digitised, stroma and epithelium were selectively annotated, and all selected areas were quantitatively analysed for marker expression. The most striking findings were decreased stromal expression levels of several steroid hormone receptors, increased CAF-phenotypes and increased vessel densities in high GS prostate cancer compared to low GS prostate cancer and paired prostate non-tumour tissue. The present data reveal a complex correlation between prostate cancer differentiation and TME components and suggest that different GS can be associated with different possible actionable targets in the TME. The use of standardised digital image analysis tools generated robust and reproducible quantitative data, which is novel and more informative compared to the classic semi-quantitative and observer-dependent visual scoring of immunohistochemistry.

Image processing in digital pathology: an opportunity to solve inter-batch variability of immunohistochemical staining.

Immunohistochemistry (IHC) is a widely used technique in pathology to evidence protein expression in tissue samples. However, this staining technique is known for presenting inter-batch variations. Whole slide imaging in digital pathology offers a possibility to overcome this problem by means of image normalisation techniques. In the present paper we propose a methodology to objectively evaluate the need of image normalisation and to identify the best way to perform it. This methodology uses tissue microarray (TMA) materials and statistical analyses to evidence the possible variations occurring at colour and intensity levels as well as to evaluate the efficiency of image normalisation methods in correcting them. We applied our methodology to test different methods of image normalisation based on blind colour deconvolution that we adapted for IHC staining. These tests were carried out for different IHC experiments on different tissue types and targeting different proteins with different subcellular localisations. Our methodology enabled us to establish and to validate inter-batch normalization transforms which correct the non-relevant IHC staining variations. The normalised image series were then processed to extract coherent quantitative features characterising the IHC staining patterns.

High-throughput analysis of tissue-based biomarkers in digital pathology.

By simultaneously processing a large number of tissue samples, the tissue microarray (TMA) technology allows standardized screening of protein expression using immunohistochemistry thereby providing a very efficient way for tissue-based biomarker analysis. Nowadays, whole slide imaging is becoming standard in digital pathology and enables image sharing, archiving and also processing. In this paper, we present methods for processing TMA images in order to correctly identify the numerous tissue samples and to register images from consecutive TMA sections.

Registration of whole immunohistochemical slide images: an efficient way to characterize biomarker colocalization.

BACKGROUND AND OBJECTIVE: Extracting accurate information from complex biological processes involved in diseases, such as cancers, requires the simultaneous targeting of multiple proteins and locating their respective expression in tissue samples. This information can be collected by imaging and registering adjacent sections from the same tissue sample and stained by immunohistochemistry (IHC). Registration accuracy should be on the scale of a few cells to enable protein colocalization to be assessed. METHODS: We propose a simple and efficient method based on the open-source elastix framework to register virtual slides of adjacent sections from the same tissue sample. We characterize registration accuracies for different types of tissue and IHC staining. RESULTS: Our results indicate that this technique is suitable for the evaluation of the colocalization of biomarkers on the scale of a few cells. We also show that using this technique in conjunction with a sequential IHC labeling and erasing technique offers improved registration accuracies. DISCUSSION: Brightfield IHC enables to address the problem of large series of tissue samples, which are usually required in clinical research. However, this approach, which is simple at the tissue processing level, requires challenging image analysis processes, such as accurate registration, to view and extract the protein colocalization information. CONCLUSIONS: The method proposed in this work enables accurate registration (on the scale of a few cells) of virtual slides of adjacent tissue sections on which the expression of different proteins is evidenced by standard IHC. Furthermore, combining our method with a sequential labeling and erasing technique enables cell-scale colocalization.