Search

The field of cancer diagnostics has been revolutionized by liquid biopsies, which offer a bridge between laboratory research and clinical settings. These tests are less invasive than traditional biopsies and more convenient than routine imaging methods. Liquid biopsies allow studying of tumor-derived markers in bodily fluids, enabling the development of more precise cancer diagnostic tests for screening, disease monitoring, and therapy personalization. This study presents a multiclass approach based on deep learning to analyze and classify diseases based on blood platelet RNA. Its primary objective is to enhance cancer-type diagnosis in clinical settings by leveraging the power of deep learning combined with high-throughput sequencing of liquid biopsy. Ultimately, the study demonstrates the potential of this approach to accurately identify the patient’s type of cancer. Methods: The developed method classifies patients using heatmap images, generated based on gene expression arranged according to the Kyoto Encyclopedia of Genes and Genomes pathways. The images represent samples of patients with ovarian cancer, endometrial cancer, glioblastoma, non-small cell lung cancer, and sarcoma, as well as cancer patients with brain metastasis. Results: Our deep learning-based models reached 66.51% balanced accuracy when distinguishing between those 6 sites of cancer origin and 90.5% balanced accuracy on a location-specific dataset where cancer types from close locations were grouped. The developed models were further investigated with an explainable artificial intelligence-based approach (XAI) - SHAP. They returned a set of 60 genes with the highest impact on the models’ decision-making process. Conclusions: Our results show that deep-learning methods are a promising opportunity for cancer detection and could support clinicians’ decision-making process in finding the solution for the black-box problem. Clinical and Translational Impact Statement— Utilizing TEPs-based liquid biopsies and deep learning, our study offers a novel approach to early cancer detection, highlighting cancer origin. The integration of Explainable AI reinforces trust in predictive outcomes. Category: Early/Pre-Clinical Research.

We have recently demonstrated that fibroblast growth factor receptor 2 (FGFR2)‐mediated signalling alters progesterone receptor (PR) activity and response of oestrogen receptor α (ER)‐positive (ER+) breast cancer (BCa) cell lines to anti‐ER agents. Little is known about whether the crosstalk between ER and PR, shown to be modulated by the hormonal background, might also be affected by FGFR2. Here, PR‐dependent behaviour of ER+ BCa cells was studied in the presence of oestrogen (E2) and progesterone (P4) and/or FGF7. In vitro analyses showed that FGF7/FGFR2 signalling: (a) abolished the effect of P4 on E2‐promoted 3D cell growth and response to tamoxifen; (b) regulated ER and PR expression and activity; (c) increased formation of ER–PR complexes; and (d) reversed P4‐triggered deregulation of ER‐dependent genes. Analysis of clinical data demonstrated that the prognostic value of FGFR2 varied between patients with different menopausal status; that is, high expression of FGFR2 was significantly associated with longer progression‐free survival (PFS) in postmenopausal patients, whereas there was no significant association in premenopausal patients. FGFR2 was found to positively correlate with the expression of JunB proto‐oncogene, AP‐1 transcription factor subunit (JUNB), an ER‐dependent gene, only in premenopausal patients. Molecular analyses revealed that the presence of JunB was a prerequisite for FGFR2‐mediated abrogation of P4‐induced inhibition of cell growth. Our results demonstrate for the first time that the FGF7/FGFR2–JunB axis abolishes the modulatory effects of PR on ER‐associated biological functions in premenopausal ER+ BCa. This may provide foundations for a better selection of patients for FGFR‐targeting therapeutic strategies.

Abstract Immune system plays a dual role in cancer by either targeting or supporting neoplastic cells at various stages of disease, including metastasis. Yet, the exact immune-related transcriptome profiles of primary tumours (PT) and lymph node metastases (LNM) and their evolution during luminal breast cancer (BCa) dissemination remain undiscovered. In order to identify the immune-related transcriptome changes that accompany lymphatic spread, we analysed PT-LNM pairs of luminal BCa using NanoString technology. Decrease in complement C3—one of the top-downregulated genes, in LNM was validated at the protein level using immunohistochemistry. Thirty-three of 360 analysed genes were downregulated (9%), whereas only 3 (0.8%) upregulated in LNM when compared to the corresponding PT. In LNM, reduced expression was observed in genes related to innate immunity, particularly to the complement system (C1QB, C1S, C1R, C4B, CFB, C3, SERPING1 and C3AR1). In validation cohort, complement C3 protein was less frequently expressed in LNM than in PT and it was associated with worse prognosis. To conclude, local expression of the complement system components declines during lymphatic spread of non-metastatic luminal BCa, whilst further reduction of tumoral complement C3 in LNM is indicative for poor survival. This points to context-dependent role of complement C3 in BCa dissemination.

Liquid biopsy demonstrates excellent potential in patient management by providing a minimally invasive and cost‐effective approach to detecting and monitoring cancer, even at its early stages. Due to the complexity of liquid biopsy data, machine‐learning techniques are increasingly gaining attention in sample analysis, especially for multidimensional data such as RNA expression profiles. Yet, there is no agreement in the community on which methods are the most effective or how to process the data. To circumvent this, we performed a large‐scale study using various machine‐learning techniques. First, we took a closer look at existing datasets and filtered out some patients to assert data collection quality. The final data collection included platelet RNA samples acquired from 1397 cancer patients (17 types of cancer) and 354 asymptomatic, presumed healthy, donors. Then, we assessed an array of different machine‐learning models and techniques (e.g., feature selection of RNA transcripts) in pan‐cancer detection and multiclass classification. Our results show that simple logistic regression performs the best, reaching a 68% cancer detection rate at a 99% specificity level, and multiclass classification accuracy of 79.38% when distinguishing between five cancer types. In summary, by revisiting classical machine‐learning models, we have exceeded the previously used method by 5% and 9.65% in cancer detection and multiclass classification, respectively. To ease further research, we open‐source our code and data processing pipelines (https://gitlab.com/jopekmaksym/improving‐platelet‐rna‐based‐diagnostics), which we hope will serve the community as a strong baseline.

Abstract Micronuclei (MN) can form through many mechanisms, including the breakage of aberrant cytokinetic chromatin bridges. The frequent observation of MN in tumors suggests that they might not merely be passive elements but could instead play active roles in tumor progression. Here, we propose a mechanism through which the presence of micronuclei could induce specific phenotypic and functional changes in cells and increase the invasive potential of cancer cells. Through the integration of diverse in vitro imaging and molecular techniques supported by clinical samples from patients with prostate cancer (PCa) defined as high-risk by the D’Amico classification, we demonstrate that the resolution of chromosome bridges can result in the accumulation of Emerin and the formation of Emerin-rich MN. These structures are negative for Lamin A/C and positive for the Lamin-B receptor and Sec61β. MN can act as a protein sinks and result in the pauperization of Emerin from the nuclear envelope. The Emerin mislocalization phenotype is associated with a molecular signature that is correlated with a poor prognosis in PCa patients and is enriched in metastatic samples. Emerin mislocalization corresponds with increases in the migratory and invasive potential of tumor cells, especially in a collagen-rich microenvironment. Our study demonstrates that the mislocalization of Emerin to MN results in increased cell invasiveness, thereby worsening patient prognosis.

Abstract Circulating tumor cells (CTCs) are tumor cells that separate from the solid tumor and enter the bloodstream, which can cause metastasis. Detection and enumeration of CTCs show promising potential as a predictor for prognosis in cancer patients. Furthermore, single-cells sequencing is a technique that provides genetic information from individual cells and allows to classify them precisely and reliably. Sequencing data typically comprises thousands of gene expression reads per cell, which artificial intelligence algorithms can accurately analyze. This work presents machine-learning-based classifiers that differentiate CTCs from peripheral blood mononuclear cells (PBMCs) based on single cell RNA sequencing data. We developed four tree-based models and we trained and tested them on a dataset consisting of Smart-Seq2 sequenced data from primary tumor sections of breast cancer patients and PBMCs and on a public dataset with manually annotated CTC expression profiles from 34 metastatic breast patients, including triple-negative breast cancer. Our best models achieved about 95% balanced accuracy on the CTC test set on per cell basis, correctly detecting 133 out of 138 CTCs and CTC-PBMC clusters. Considering the non-invasive character of the liquid biopsy examination and our accurate results, we can conclude that our work has potential application value.

Abstract Background Cancer-associated fibroblasts (CAFs) have been shown to support tumor development in a variety of cancers. Different markers were applied to classify CAFs in order to elucidate their impact on tumor progression. However, the exact mechanism by which CAFs enhance cancer development and metastasis is yet unknown. Methods Alpha-smooth muscle actin (α-SMA) was examined immunohistochemically in intratumoral CAFs of nonmetastatic breast cancers and correlated with clinicopathological data. Four CAF cell lines were isolated from patients with luminal breast cancer (lumBC) and classified according to the presence of α-SMA protein. Conditioned medium (CM) from CAF cultures was used to assess the influence of CAFs on lumBC cell lines: MCF7 and T47D cells using Matrigel 3D culture assay. To identify potential factors accounting for promotion of tumor growth by α-SMAhigh CAFs, nCounter PanCancer Immune Profiling Panel (NanoString) was used. Results In luminal breast cancer, presence of intratumoral CAFs expressing high level of α-SMA (13% of lumBC group) correlated with poor prognosis (p = 0.019). In in vitro conditions, conditioned medium obtained from primary cultures of α-SMA-positive CAFs isolated from luminal tumors was observed to enhance growth of lumBC cell line colonies in 3D Matrigel, in contrast to CM derived from α-SMA-negative CAFs. Multigene expression analysis indicated that osteopontin (OPN) was overexpressed in α-SMA-positive CAFs in both clinical samples and in vitro models. OPN expression was associated with higher percentage of Ki67-positive cells in clinical material (p = 0.012), while OPN blocking in α-SMA-positive CAF-derived CM attenuated growth of lumBC cell line colonies in 3D Matrigel. Conclusions Our findings demonstrate that α-SMA-positive CAFs might enhance tumor growth via secretion of OPN.

Abstract Background Immunotherapy is emerging as a powerful treatment approach for several types of cancers. Modulating the immune system to specifically target cancer cells while sparing healthy cells, is a very promising approach for safer therapies and increased survival of cancer patients. Tumour-associated antigens are favorable targets for cancer immunotherapy, as they are exclusively expressed by the cancer cells, minimizing the risk of an autoimmune reaction. The ability to initiate the activation of the immune system can be achieved by virus-like particles (VLPs) which are safe and potent delivery tools. VLP‐based vaccines have evolved dramatically over the last few decades and showed great potential in preventing infectious diseases. Immunogenic potency of engineered VLPs as a platform for the development of effective therapeutic cancer vaccines has been studied extensively. This study involves recombinant VLPs presenting multiple copies of tumour-specific mucin 1 (MUC1) epitope as a potentially powerful tool for future immunotherapy. Results In this report VLPs based on the structural protein of Norovirus (NoV VP1) were genetically modified to present multiple copies of tumour-specific MUC1 epitope on their surface. Chimeric MUC1 particles were produced in the eukaryotic Leishmania tarentolae expression system and used in combination with squalene oil-in-water emulsion MF59 adjuvant to immunize BALB/c mice. Sera from vaccinated mice demonstrated high titers of IgG and IgM antibodies which were specifically recognizing MUC1 antigen. Conclusions The obtained results show that immunization with recombinant chimeric NoV VP1- MUC1 VLPs result in high titers of MUC1 specific IgG antibodies and show great therapeutic potential as a platform to present tumour-associated antigens. Graphical Abstract

Liquid biopsies offer a minimally invasive sample collection, outperforming traditional biopsies employed for cancer evaluation. The widely used material is blood, which is the source of tumor‐educated platelets. Here, we developed the imPlatelet classifier, which converts RNA‐sequenced platelet data into images in which each pixel corresponds to the expression level of a certain gene. Biological knowledge from the Kyoto Encyclopedia of Genes and Genomes was also implemented to improve accuracy. Images obtained from samples can then be compared against standard images for specific cancers to determine a diagnosis. We tested imPlatelet on a cohort of 401 non‐small cell lung cancer patients, 62 sarcoma patients, and 28 ovarian cancer patients. imPlatelet provided excellent discrimination between lung cancer cases and healthy controls, with accuracy equal to 1 in the independent dataset. When discriminating between noncancer cases and sarcoma or ovarian cancer patients, accuracy equaled 0.91 or 0.95, respectively, in the independent datasets. According to our knowledge, this is the first study implementing an image‐based deep‐learning approach combined with biological knowledge to classify human samples. The performance of imPlatelet considerably exceeds previously published methods and our own alternative attempts of sample discrimination. We show that the deep‐learning image‐based classifier accurately identifies cancer, even when a limited number of samples are available.

Abstract Blood platelet RNA-sequencing is increasingly used among the scientific community. Aberrant platelet transcriptome is common in cancer or cardiovascular disease, but reference data on platelet RNA content in healthy individuals are scarce and merit complex investigation. We sought to explore the dynamics of platelet transcriptome. Datasets from 204 healthy donors were used for the analysis of splice variants, particularly with regard to age, sex, blood storage time, unit of collection or library size. Genes B2M, PPBP, TMSB4X, ACTB, FTL, CLU, PF4, F13A1, GNAS, SPARC, PTMA, TAGLN2, OAZ1 and OST4 demonstrated the highest expression in the analysed cohort, remaining substantial transcription consistency. CSF3R gene was found upregulated in males (fold change 2.10, FDR q

The most recent classification of the lung cancer expanded the diagnostic criteria of its histological subtypes and included its immunophenotypic profile. We performed the study to compare the reliability of selected markers in high-grade non-small cell lung carcinoma (NSCLC) in the oligobiopsies with the matched postoperative samples. We evaluated expression of p40, p63, TTF1, cytokeratin 5/6, cytokeratin 7, napsin A, desmoglein 3, desmocollin 3 and mucin secretion as detected by mucicarmine staining. The study cohort included 123 cases of poorly-differentiated NSCLC. The tissue oligobiopsy material was available in 38 cases. Tissue microarrays (TMAs) from all postoperative cases were constructed. Comparing the immunophenotype between postsurgical samples and oligobiopsies we found an almost perfect agreement for most of performed IHC reactions. The highest concordance of results was found for desmoglein 3, CK7, and p40, whereas the lowest – for desmocollin 3. Immunoprofile of the oligobiopsies corresponded well to that in the resection specimens. The most useful markers in poorly differentiated ADs are: TTF1 and napsin A, and for non-keratinizing SCCs: p40, p63, CK5/6 and desmoglein 3.

Estrogen (ER) and progesterone (PgR) receptors and HER2 are crucial in the assessment of breast cancer specimens due to their prognostic and predictive significance. Single hormone receptor-positive breast cancers are less common and their clinical course is less favorable than ER(+)/PgR(+) tumors. Their molecular features, especially microRNA (miRNA) profiles, have not been investigated to date. Tumor specimens from 36 chemonaive breast cancer patients with known ER and PgR status (18 ER(+)/PgR(−) and 18 ER(−)/PgR(+) cases) were enrolled to the study. The expression of 829 miRNAs was evaluated with nCounter Human v3 miRNA expression Assay (NanoString). miRNAs differentiating between ER/PgR/HER2 phenotypes were selected based on fold change (FC) calculated for the mean normalized counts of each probe in compared groups. The differences were estimated with Student’s t-test or Two-Way ANOVA (considering also the HER2 status). The results were validated using The Cancer Genome Atlas (TCGA) dataset. Following quality control of raw data, fourcases were excluded due to low sample quality, leaving 14 ER(+)/PgR(−) and 18 ER(−)/PgR(+) cases. After correction for multiple comparisons, we did not find miRNA signature differentiating between ER(−)/PgR(+) and ER(+)/PgR(−) breast cancers. However, a trend for differing expression (p-value ≤ 0.05; FDR > 0.2; ANOVA) in eight miRNAs was observed. The ER(+)/PgR(−) group demonstrated elevated levels of four miRNAs—miR-30a-5p, miR-29c-3p, miR-141-3p and miR-423-5p—while the ER(−)/PgR(+) tumors were enriched in another four miRNAs—miR-514b-5p, miR-424-5p, miR-495-3p, and miR-92a-3p. For one of the miRNAs—miR-29c-3p—the association with the ER(+)/PgR(−) phenotype was confirmed in the TCGA cohort (p-value = 0.024; t-test). HER2 amplification/overexpression in the NanoString cohort was related to significant differences observed in 33 miRNA expression levels (FDR ≤ 0.2; ANOVA). The association with HER2 status was confirmed in the TCGA cohort for four miRNAs (miR-1180-3p, miR-223-3p, miR-30d-5p, and miR-195-5p). The main differences in miRNA expression amongst single hormone receptor-positive tumors were identified according to their HER2 status. However, ER(+)/PgR(−) cases tended to express higher levels of miRNAs associated with ER-positivity (miR-30a-5p, miR-29c-3p, miR-141-3p), whereas ER(−)/PgR(+) cancers showed elevated levels of miRNAs characteristic for double- and triple-negative tumors (miR-92a-3p, miR-424-5p). Further studies are necessary to comprehensively analyze miRNA signatures characteristic of ER(−)/PgR(+) and ER(+)/PgR(−) tumors.

Intratumor heterogeneity implies heterogeneous protein function, facilitating tumor adaptation which results in therapeutic failure. We hypothesized that tumor heterogeneity at protein level may influence the course of the disease. As a single biopsy might not represent the full biologic complexity of the tumor, we have analyzed immunohistochemically four different cores obtained from each primary tumor within the cohort of 364 patients with endometrial cancer (EC). The following proteins were examined: estrogen receptor 1 (ESR1), progesterone receptor, epidermal growth factor receptor, v-erb-b2 erythroblastic leukemia viral oncogene homolog 2, receptor tyrosine-protein kinase erbB-3, v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4, phosphatidylinositol-4,5-bisphosphate 3-kinase, phosphorylated v-akt murine thymoma viral oncogene homolog 1, v-myc avian myelocytomatosis viral oncogene homolog, DNA topoisomerase II alpha 170 kDa (TOP2A), cyclin-dependent kinase inhibitor 2A (CDKN2A), tumor protein p53, RAD21 homolog, S. pombe, and runt-related transcription factor 1. Particularly strong correlation was found between TOP2A and CDKN2A heterogeneity and higher stage of the disease (P = .0002 and P = .0003, respectively). Most correlations with clinicopathologic data were observed for ESR1 heterogeneity that correlated with non-endometrioid carcinomas (P=.02), higher stage (P=.005), grade (P=.01), and the presence of metastases (P = .01). Thirty-nine (11.0%) patients were classified as “globally heterogeneous”. Cumulative tumor heterogeneity strongly correlated with the presence of metastases, higher stage, and higher grade of the disease (all P b .05). It also carried negative prognostic value (P=.0008). We show that the degree of heterogeneity in EC might serve as a clinically valid molecular marker.

Intratumoral heterogeneity of breast cancer remains a major challenge in successful treatment. Failure of cancer therapies can also be accredited to inability to systemically eradicate cancer stem cells (CSCs). Recent evidence points to the role of epithelial-mesenchymal transition (EMT) in expanding the pool of tumor cells with CSCs features. Thus, we assessed expression level as well as heterogeneity of CSCs markers in primary tumors (PT), lymph node metastasis (LNM), and circulating tumor cells (CTCs)–enriched blood fractions in order to correlate them with signs of EMT activation as well as clinicopathological data of breast cancer patients. Level of CSCs markers (ALDH1, CD44, CD133, OCT-4, NANOG) and EMT markers was quantified in PT (N=107), LNM (N=56), and CTCs-enriched blood fractions (N=85). Heterogeneity of CSCs markers expression within each PT and LNM was assessed by calculating Gini Index. Percentage of ALDH1-positive cells was elevated in PT in comparison to LNM (P = .005). However, heterogeneity of the four CSCs markers: ALDH1 (P = .019), CD133 (P = .009), OCT-4 (P = .027), and CD44 (P < .001) was decreased in LNM. Samples classified as mesenchymal (post-EMT) showed elevated expression of CSCs markers (OCT-4 and CD44 in PT; OCT-4 in LNM; ALDH1, OCT-4, NANOG, CD44 in CTCs). Patients with mesenchymal-like CTCs had worse prognosis than patients with epithelial-like or no CTCs (P = .0025). CSCs markers are enriched in PT, LNM, and CTCs with mesenchymal features, but their heterogeneity is decreased in metastatic lymph nodes. Mesenchymal CTCs phenotype correlates with poor prognosis of the patients.

Amplification of the ESR1 gene, coding for estrogen receptor alpha, was shown to predict responsiveness to tamoxifen, however its prognostic impact in breast cancer patients has not been thoroughly investigated. Other factors that could contribute to responsiveness to tamoxifen treatment are polymorphisms in ESR1 gene and genes involved in tamoxifen metabolism. The aim of this study was to assess the prognostic role of ESR1 gene dosage in a consecutive group of breast cancer patients and to correlate this feature with clinico-pathological factors. Additionally, ESR1 PvuII, CYP2C19*2 and UGT2B15*2 polymorphisms were analyzed in the tamoxifen-treated subgroup of patients.Primary tumor samples from 281 stage I-III consecutive breast cancer patients were analyzed for ESR1 gene dosage using real-time PCR with locked nucleic acids hydrolysis probes. In the tamoxifen-treated subgroup of patients, ESR1 PvuII, CYP2C19*2 and UGT2B15*2 polymorphism in leukocytes genomic DNA were analyzed. Results were correlated with clinico-pathological factors and with disease-free survival (DFS) and overall survival (OS).ESR1 amplification (with a cut-off level of 2.0) was found in 12% of the entire group of breast cancer patients, and in 18% of the ER-negative subgroup. This feature was associated with decreased DFS both in the entire group (P=0.007) and in the ER-negative subgroup (P=0.03), but not in the tamoxifen-treated patients. Patients with ESR1 PvuII wt/wt genotype and at least one UGT2B15 wt allele had a worse DFS (P=0.03) and showed a trend towards decreased Os (P=0.08) in comparison to patients with ESR1 PvuII wt/vt or vt/vt genotype and UGT2B15 *2/*2 genotype.ESR1 amplification can occur in ER-negative tumors and may carry poor prognosis. In the tamoxifen-treated subgroup, poor prognosis was related to the combined presence of ESR1 PvuII wt/wt and UGT2B15wt/wt or wt/*2 genotype.