Exploring platelet-covered and naked circulating tumor cells - a single-cell transcriptomic perspective

The study included single-cell sequencing data from 29 donors: 12 diagnosed with high-grade serous ovarian cancer (HGSOC), 4 serving as a control group with histologically confirmed benign ovarian lesions (including tumors of the genital cords, Brenner tumor, endometriosis ovari, and borderline ovarian tumors), 3 healthy controls, and 10 diagnosed with breast cancer (BRCA). Importantly, blood was collected at diagnosis, prior to any treatment. Clinical characteristics of ovarian and breast cancer patients enrolled in the study are presented in Tables  1 and 2 , respectively. CTC candidates were detected in patients HGSOC 09, HGSOC 10, HGSOC 12, HGSOC 02 and BRCA 07. Table 1 Information about ovarian cancer patients (HGSOC – high-grade serous ovarian cancer) Patient ID Group Grade Stage Age HGSOC 01 high-grade serous ovarian carcinoma G3 IIIC 74 HGSOC 02 high-grade serous ovarian carcinoma G3 IIIC 48 HGSOC 03 high-grade serous ovarian carcinoma G3 IV 63 HGSOC 04 high-grade serous ovarian carcinoma G3 IIA 50 HGSOC 05 high-grade serous ovarian carcinoma G3 IIIC 47 HGSOC 06 high-grade serous ovarian carcinoma G3 IIIC 77 HGSOC 07 high-grade serous ovarian carcinoma G3 IIIA 33 HGSOC 08 high-grade serous ovarian carcinoma G3 IIIB 77 HGSOC 09 high-grade serous ovarian carcinoma G3 IIIC 61 HGSOC 10 high-grade serous ovarian carcinoma G3 IIIC 63 HGSOC 11 high-grade serous ovarian carcinoma G3 IIIC 73 HGSOC 12 high-grade serous ovarian carcinoma G3 IIIB 67 Information about ovarian cancer patients (HGSOC – high-grade serous ovarian cancer) Table 2 Information about breast cancer patients (BRCA – breast cancer) Patient ID Group Grade Stage Age BRCA 01 Luminal B, HER2 -negative G1 IIA 50 BRCA 02 Triple-Negative Breast Cancer G3 IIIB 59 BRCA 03 Luminal B, HER2 -negative G3 IIAc 41 BRCA 04 Luminal B, HER2 -positive G3 IIA 56 BRCA 05 Triple-Negative Breast Cancer G3 IIIA 33 BRCA 06 Luminal B, HER2 -positive G2 IV 78 BRCA 07 Triple-Negative Breast Cancer G3 IIA 49 BRCA 08 Triple-Negative Breast Cancer G3 IV 68 BRCA 09 Triple-Negative Breast Cancer G3 IIA 49 BRCA 10 Triple-Negative Breast Cancer G3 IIA 38 Information about breast cancer patients (BRCA – breast cancer) CTC identification and phenotyping was performed using immunofluorescent staining (pan-K/vimentin/DAPI/aSMA/CD29/CD45/CD31) in PBMCs fractions putatively containing CTCs using Amnis ® ImageStreamX MkII (Cytek Biosciences) as described in Muchlinska et al. [ 34 ] in the cohort of 24 ovarian cancer patients selected in Wenta et al. [ 29 ]. Briefly, the antibody panel included antibodies targeting pan-keratins (pan-K; epithelial marker), vimentin (V; mesenchymal marker), alpha-smooth muscle actin (α-SMA; myofibroblast/mesenchymal marker), CD29 (integrin β1 (platelet glycoprotein [GP] IIa); adhesion marker), CD31 (endothelial marker), and CD45 (hematopoietic marker). In addition, nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) to enable DNA visualization, while side scatter (SSC) was used to capture information on intracellular granularity and structural complexity. Platelets were assessed visually in the brightfield channel based on their morphology. For bulk platelets data we used a publicly available dataset (GEO: GSE183635 ), comprising 2,351 samples from asymptomatic and symptomatic controls, as well as 18 cancer types. From this dataset we selected only samples with ovarian and breast cancer − 236 samples in total. Following the methodology outlined by Pastuszak et al. [ 35 ], we normalized gene expression data using the DESeq2 package (version 1.44.0) [ 36 ] in R (version 4.4.1) through variance-stabilizing transformation (VST) [ 37 ]. The human reference genome hg19 was used for annotation throughout. All samples underwent standardized pre-processing and were normalized collectively. For transcript selection, we applied the criteria described in [ 38 ], retaining 5,346 transcripts with at least 30 reads in a minimum of 10% of samples in at least one group. Whole blood was collected into 10 ml EDTA tubes before any treatment. It was then transferred to 15 ml Falcon tubes pre-coated with DB buffer. An equal volume of 1xPBS was added, and the mixture was layered onto Histopaque 1.077 (SigmaAldrich), a density gradient medium commonly used for PBMC isolation, and centrifuged (400 g, 30 min, room temperature). The resulting buffy coat was placed on ice, centrifuged again (300 g, 5 min, 4 °C), and the supernatant was discarded. The pellet was resuspended in a dedicated medium After adding 0.5 mL of freezing medium 90% FBS + 10% DMSO the cells were counted using a Countess cell counter (Thermo Fisher Scientific), transferred to cryotubes, and stored at -80 °C until single-cell sequencing was performed. Thawed buffy coat was used for single cell sequencing with the target of 10 000 cells/patient. No enrichment of PBMCs for CTC selection was used. Ovarian cancer and benign gynecologic control samples were processed at the Institute of Bioorganic Chemistry, Polish Academy of Sciences, in Poznań. Cryovials were thawed in a 37 °C water bath, and cells were transferred to 50 ml conical tubes using wide-bore pipette tips. After dilution in a warm 10% FBS complete growth medium, cells were centrifuged and resuspended. Cell concentration and viability were assessed with trypan blue staining using the Countess™ 3 FL Automated Cell Counter. Samples with viability over 70% (average 84%) were resuspended in 1X PBS + 0.04% BSA to a final concentration of 700–1,200 cells/µl. Single-cell libraries were prepared using the Chromium Next GEM Single Cell 3′ Reagent Kit (10x Genomics, USA) following the manufacturer’s protocol. Breast cancer and healthy donor samples were processed by Singleron Biotechnologies (Germany). The protocol mirrored that of the Poznań site, with similar cell viability obtained, but the library preparation was performed using the SCOPE-chip ® micro-well microfluidics system according to the manufacturer’s instructions. Regardless of the processing site, all samples were sequenced on the NovaSeqX platform, in separate runs, targeting 20,000 reads per cell. Sequencing data were demultiplexed and converted to FASTQ format using CellRanger mkfastq. Unique Molecular Identifier (UMI) count matrices representing raw gene expression data were produced with the standard Cell Ranger software (version 5.01) from 10x Genomics, employing the GRCh38 human reference genome for mapping. This processing involved filtering out low-quality cells as well as removing apoptotic cells. Specifically, cells with fewer than 1000 reads overall, cells expressing fewer than 100 unique genes, and cells with over 10% mitochondrial RNA were excluded from further analysis. Following quality control, each read count matrix was normalized to the total expression, scaled by a factor of 10,000, and log-transformed. We also identified platelets in single-cell RNA-sequencing data from peripheral blood mononuclear cells of patients with high-grade serous ovarian cancer and breast cancer. After the aforementioned quality control and normalization, cell types were annotated using SingleR with the Human Primary Cell Atlas as the reference dataset. Cells classified as platelets were designated as candidates, representing potential trace contamination in PBMC isolates. After platelets classification additional filtering was performed, excluding cells with positive expression of genes characteristic for erythrocytes (hemoglobin and glycophorin C) as well as leukocytes ( LYZ , CD3D , CD14 , CD19 , CD56 , IGHM , PTPRC/CD45 ) [ 39 ]. Normalized expressions of retrieved platelets were extracted for subsequent validation and analysis. To detect CTC and pcCTC candidates, we applied strict filtering criteria. We removed cells expressing markers characteristic of leukocytes, endothelial cells, fibroblasts, dendritic cells, stromal cells, adipocytes, and pericytes. The list of markers used is shown in Table  3 . Only cells with no expression of these markers were included in the further analysis. Table 3 Markers used to filter out cells for further analysis. The references justifying the selection of these markers may be found in Supplementary Table 1 Cell type Markers Leukocytes MS4A1 , CD19 , CD8A , CD8B , KLRF1 , NCAM1 , CD68 , FCN1 , LYZ , FCGR3B , S100A9 , CD14 , FCGR3A , CD4 , IL7R , SELL , CCR7 , CD27 , IL2RA , FOXP3 , TIGIT , FUT4 , CEACAM8 , ITGAM , ITGB2 , PTPRC , CD3E , CD3D , CD3G Endothelial CDH5 , CD34 , VWF , ID3 , EGFL7 , CD93 Fibroblasts LUM , FAP , COL6A2 , COL1A2 , COL1A1 , COL6A3 Dendritic cells FCER1A , CD1C , AXL , CD83 , CLEC4C , THBD , NRP1 Stromal cells COL15A1 , GDF10 , COL4A1 , BMP4 , BMP5 , BMP6 Adipocytes LRP1 , LOX , AOC3 , TFE3 , STAT6 , CAVIN1 Pericytes MCAM , KCNJ8 , CSPG4 , MYO1B , ANGPT2 Markers used to filter out cells for further analysis. The references justifying the selection of these markers may be found in Supplementary Table 1 Next, we selected marker genes for the detection of CTC candidates representing epithelial and mesenchymal phenotypes, as well as markers associated with ovarian and breast cancer. We also included markers characteristic of platelets, most of which have been previously reported as upregulated in pcCTCs according to the literature [ 31 , 40 , 41 ]. The results are summarized in Table  4 . Table 4 Markers used to detect circulating tumor cells. The references justifying the selection of these markers may be found in Supplementary Table 2 Cell type Markers Epithelial KRT8 , KRT7 , KRT18 , KRT19 , KRT80 , KRT13 , MUC1 , EPCAM , EGFR , EMP2 , CDH1 , CLDN3 , CLDN4 , CLDN7 Mesenchymal VIM , FN1 , CDH2 , TWIST1 , TWIST2 , SNAI1 , SNAI2 , ZEB1 , ZEB2 , AKT1 , AKT2 , AXL , SPARC , ACTA2 , MMP9 , CXCR4 Ovarian cancer-associated CD44 , ALDH1 , SOX2 , ERCC1 , NANOG4 , OCT4 , PIK3A , IL6 , CXCL1 , CCL20 Breast cancer-associated NOTCH1 , NOTCH2 , NOTCH3 , NOTCH4 , ESR1 , HER2 , WNT1 , WNT2 , WNT3 , WNT4 , ERBB2 , WNT10B Platelets VEGFB , MYC , IL33 , TGFB2 , PTGER2 , VEGFA , FGF2 , GP1BA , MMP2 , F2 , PPBP , ITGB3 , PF4 , GP9 , ITGA2B , CD63 , SELP , MMP7 , MMP9 , TGFB1 , CCL5 , CCL2 , PECAM1 , ADGRE5 , SERPINH1 Markers used to detect circulating tumor cells. The references justifying the selection of these markers may be found in Supplementary Table 2 After filtering cells according to the criteria described in Table  3 , we calculated the expression levels of genes listed in Table  4 for both cancer and control groups using the UCell scoring method. Scores were also computed for the control group to determine the maximum score that a non-CTC cell could achieve. Subsequently, we selected cells from cancer patients whose scores exceeded the respective control thresholds. Epithelial CTC candidates were defined as cells exceeding the epithelial control threshold while having both mesenchymal and platelets scores equal to zero. Mesenchymal CTC candidates were defined analogously. Ovarian cancer–associated CTC candidates were detected only among ovarian cancer patients and defined as cells exceeding the ovarian cancer–associated control threshold with a platelets score of zero. Breast cancer–associated CTC candidates were detected analogously among breast cancer patients. Additionally, cells showing a positive platelets score were classified as potential pcCTC candidates. Functional enrichment analyses were performed using gene sets from the MSigDB Hallmark and Reactome databases. From Hallmark we verified pathways such as epithelial–mesenchymal transition and angiogenesis and from Reactome we investigated processes such as extracellular matrix organization and platelet signaling and aggregation. Using both databases in parallel allows complementary insights: Hallmark gene sets highlight overarching biological trends, while Reactome pathways reveal the underlying molecular details. This combined approach offers a more comprehensive understanding of the cellular processes enriched in the analyzed data [ 42 , 43 ]. Single-cell analysis was performed in R (version 4.5.1). Quality control, normalization, and cell filtering were carried out using the Seurat package (version 5.3.1). Platelet detection was done using the SingleR package (version 2.10.0). Scores for CTC candidates were calculated with the UCell package (version 2.12.0). Heatmaps were generated using the ComplexHeatmap package (version 2.24.1). Genes associated with biological processes were identified from the Hallmark and Reactome gene sets using the msigdbr package (version 25.1.1). Statistical comparisons between two groups were performed using the Mann–Whitney test due to the non-normal distribution of the data. A p -value threshold of 0.05 was considered statistically significant.

In order to identify the exact occurrence of CTCs and pcCTCs, we used imaging flow cytometry imFC). In brief, the antibody cocktail included proteins targeting pan-keratins (pan-K; epithelial cell marker), vimentin (V; mesenchymal cell marker), CD31 (PECAM-1; endothelial cell marker), and CD45 (leukocyte antigen; hematopoietic marker). In addition, nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) to enable DNA visualization [ 29 , 34 ]. We examined an unmatched cohort of 24 ovarian cancer patients and we found 454 circulating tumor cells in this group from which about 64% of these cells were covered with platelets. Figure  1 A and B present representative images of CTC and pcCTC of epithelial (K + V-) phenotype from two ovarian cancer patients, both diagnosed with high-grade serous ovarian carcinoma (grade G3). The patient from whom the naked CTC candidate was identified is 61 years old with stage IB disease, whereas the patient from whom the pcCTC candidate was identified is 80 years old with stage IIIB disease. Fig. 1 Imaging flow cytometry images of CTC ( A ) and pcCTC ( B ) candidates. The lower panel shows several small objects surrounding the cell that, based on size analysis, were identified as platelets Imaging flow cytometry images of CTC ( A ) and pcCTC ( B ) candidates. The lower panel shows several small objects surrounding the cell that, based on size analysis, were identified as platelets Next, CTC and pcCTC candidates were identified using single cell sequencing data, based on the developed gene panel and filtering for genes associated with leukocytes, endothelial cells, fibroblasts, dendritic cells, stromal cells, adipocytes, and pericytes. The CTC and pcCTC candidates were found in 4 ovarian cancer patients and 1 breast cancer patient. No CTC candidates were found in benign and healthy donor samples. In this study, we identified 12 CTC candidates: one epithelial pcCTC, three mesenchymal pcCTCs, three ovarian cancer-associated pcCTCs, three ovarian cancer-associated naked CTCs, and two naked mesenchymal CTCs. Figure  2 illustrates the normalized expression of selected genes defining the EMT phenotype in the aforementioned cells, accompanied by metadata including cancer type and platelets score. The platelets score was defined as a UCell score used to detect platelet signals in CTC candidates, normalized to range from 0 to 1 for readability. A score of 0 indicates that the CTC is not covered with platelets, whereas higher scores correspond to progressively stronger platelet signals detected on the CTC. While the epithelial cell exhibited expression of KRT18 , all remaining cells displayed positive expression of at least one mesenchymal marker ( VIM , ZEB1 , ZEB2 , AKT2 , SPARC , or CXCR4 ). Among these mesenchymal cells, OC-associated CTC candidates additionally expressed either CD44 or ERCC1 . Two pcCTC candidates exhibited a significantly higher platelets score than the other pcCTC candidates (two on the left side of the heatmap). Regarding the platelet-associated genes - all pcCTC candidates expressed at least one platelet gene marker ( GP1BA , PPBP , PF4 , GP9 , ITGA2B , CD63 , SELP , CCL5 , TGFB1 or PECAM1 ). Fig. 2 Heatmap of normalized gene expression levels of the identified CTC and pcCTC candidates detected using the proposed gene panel. The platelets score is UCell score used to detect platelet signals in CTCs, normalized to range from 0 to 1 for readability Heatmap of normalized gene expression levels of the identified CTC and pcCTC candidates detected using the proposed gene panel. The platelets score is UCell score used to detect platelet signals in CTCs, normalized to range from 0 to 1 for readability Furthermore, we examined the number of genes with non-zero expression in CTC and pcCTC candidates. The mean number of expressed genes (± standard deviation) was 751.6 ± 130.0 for CTC candidates and 855.7 ± 273.4 for pcCTC candidates. It is notable that pcCTC candidates tended to have higher numbers of genes with non-zero expression, however the differences were not statistically significant ( p  = 0.81), likely due to the small sample size. Next, we performed a comparison between CTC and pcCTC candidates regarding most differentiating genes. We calculated the mean normalized expression for each gene of CTC and pcCTC candidates, then identified the top 30 genes with the greatest absolute differences. The results are presented in Fig.  3 as a heatmap showing these candidates and their expression profiles. Among these genes, the following were the part of the gene panel to evaluate the platelet signal in CTC candidates: PF4 , PPBP , and CCL5 . The pcCTC candidates with the highest platelets score (the three cells on the left side of the heatmap) are clearly separated from the remaining cells. While genes associated with platelets activation, cytoskeleton and interaction with the endothelium, namely: PPBP , PF4 , TUBA4A , TUBB1 , CAVIN2 , RGS18 and GNG11 , are upregulated in pcCTC candidates with highest platelets score, genes such as: PITPNC1 , ATM , AKT3 , SRPK2 , NEMF , ANKRD12 , CCDC91 , XRCC5 , ETS1 and UTRN , among others connected with anoikis, oxidative stress, DNA damage and survival adaptations, are upregulated in naked CTC candidates. There is also B2M (β2-microglobulin) gene, encoding a component of the MHC class I complex, which has high expression in both CTC and pcCTC candidates, but higher for pcCTC candidates. Fig. 3 Heatmap of normalized gene expression levels of the identified pcCTC and naked CTC candidates demonstrating the most differentiating genes between these groups Heatmap of normalized gene expression levels of the identified pcCTC and naked CTC candidates demonstrating the most differentiating genes between these groups Next, we compared the transcriptomes of pcCTC candidates and single platelets from the same single cell sequencing dataset. Before the analysis, single platelets were filtered for red and white blood cell contamination. Analogically to the previous analysis, we calculated the mean normalized expression of each gene for both pcCTC candidates and platelets, and selected the top 30 genes showing the largest absolute differences. Interestingly, many genes exhibited higher normalized expression in single platelets. We hypothesize that this pattern may result from the mixed transcript composition of pcCTC candidates, which contain both platelet and CTC transcripts — with the latter likely being predominant. During normalization, this imbalance could lead to an apparent decrease in platelet-associated gene expression in pcCTC candidates. Notably, only two of the top 30 genes showed higher expression in pcCTC candidates — MALAT1 (Metastasis Associated Lung Adenocarcinoma Transcript 1) and ARHGAP15 (Rho GTPase-activating protein 15) connected with cancer invasiveness — and their expression was also detected in naked CTC candidates. We also performed a similar comparison between the two pcCTC candidates with the highest platelets scores and single platelets. In this case, a greater number of genes showed higher expression in the pcCTC candidates than in single platelets. The genes are as follows: FAM172A , ORAI1 , CAB39 , MYCBP2 , RPS19 , TRAPPC3L , ELOVL7 , CD27-AS1 , TUBB2A , TACC3 , ISCA1 , FUS , MBP , ZFR , FAU , MFAP3L , MEPCE , AKIRIN1 , RPS7 , ATL1 , ARHGAP5 , PADI4 , CSNK1A1 , EIF1B , ARPC2 , RPL4 associated with many processes among others: proliferation, cytoskeletal migration, remodeling, survival and RNA biogenesis. We assessed the similarity between pcCTC and CTC candidates and bulk platelet samples derived from ovarian and breast cancer. We used part of the publicly available bulk platelets dataset (GEO: GSE183635 ), which contains material from different patients then in single cell sequencing dataset. Although direct comparison of gene expression between single-cell and bulk data is limited by sequencing resolution, we investigated whether genes highly expressed in bulk platelets are also upregulated in pcCTC and CTC candidates. Figure  4 shows a heatmap of gene expression in pcCTC and CTC candidates for the top 30 genes with the highest mean expression in bulk platelet samples from ovarian and breast cancer patients. Most of these genes were also upregulated in cells displaying the highest platelet-associated signal compared with other CTCs. Notably, this included genes used to calculate the platelets score, such as PF4 , PPBP , and CCL5 ; cytoskeleton-associated genes ACTB and ACTN1 ; and B2M , a component of the MHC class I complex, which is upregulated in nearly all candidates. Fig. 4 Heatmap of normalized gene expression in pcCTC and CTC candidates for the top 30 genes with the highest mean expression in bulk platelet samples from ovarian and breast cancer patients Heatmap of normalized gene expression in pcCTC and CTC candidates for the top 30 genes with the highest mean expression in bulk platelet samples from ovarian and breast cancer patients | We further analyzed genes associated with selected biological processes from the Hallmark and Reactome databases [ 42 , 43 ], focusing on comparisons among pcCTCs, naked CTCs, and single platelets. Specifically, we selected angiogenesis and epithelial-mesenchymal transition from the Hallmark gene sets, and extracellular matrix organization along with platelet activation, signaling, and aggregation from Reactome. These pathways were chosen based on prior literature suggesting that pcCTCs may play a role in modulating these processes [ 13 , 22 , 40 ]. Figures 5  A and 5 B show heatmaps of EMT-related gene expression for pcCTC and CTC candidates and single platelets, respectively. CD44 expression was upregulated in most CTC candidates and in some pcCTC candidates, while only a few single platelets showed positive expression of this gene. The CD44 gene was used for ovarian cancer CTC candidates. Regarding the single platelets, several genes associated with adhesion, cytoskeletal remodeling and interaction with the extracellular matrix were upregulated, such as: SAT1 , TPM4 , TGFB1 , TIMP1 , TPM1 , EMP3 , GADD45A , WIPF1 , ITGB1 , SPARC , FLNA , MYLK , ITGB5 and MYL9. Notably, these genes were also upregulated in at least one pcCTC candidate. Among these genes SPARC was used as mesenchymal marker and TGFB1 as platelets marker. Fig. 5 Heatmaps of EMT-related normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). SPARC – Secreted Protein Acidic and Rich in Cysteine; osteonectin; TGFB1 (Transforming Growth Factor β1) Heatmaps of EMT-related normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). SPARC – Secreted Protein Acidic and Rich in Cysteine; osteonectin; TGFB1 (Transforming Growth Factor β1) Figure  6 A and B present genes related to platelets activation, signaling and aggregation for pcCTC and CTC candidates and single platelets, respectively. Most of the genes were upregulated for single platelets and for pcCTC candidates with the highest platelets score (3 cells on the left side of the chart). Especially genes such as PPBP , PF4 , GNG11 and TUB4A had the highest expression among pcCTC candidates and platelets. From these genes PPBP and PF4 were used for platelets score. Moreover, gene RAP1B was upregulated in most pcCTC as well as CTC candidates, and it also had high expression among single platelets. Fig. 6 Heatmaps of normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). The genes are related to platelets activation, signaling and aggregation. PPBP – pro-platelet basic protein, CXCL7; PF4 – platelet factor-4, CXCL4; RAP1B – Ras-related protein-1B Heatmaps of normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). The genes are related to platelets activation, signaling and aggregation. PPBP – pro-platelet basic protein, CXCL7; PF4 – platelet factor-4, CXCL4; RAP1B – Ras-related protein-1B Figure  7 A and B present heatmaps of normalized expression of genes involved in the angiogenesis process. The two pcCTC candidates with the highest platelets scores showed particularly high expression of the PF4 and TIMP1 genes, a pattern that is also observed in single platelets. The PF4 gene was also used for calculating the platelets score. Moreover, except for the VCAN gene, increased expression of angiogenesis-related genes was observed exclusively in pcCTC candidates and not in CTC candidates. Fig. 7 Heatmaps of angiogenesis related normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). TIMP1 – tissue inhibitor of metalloproteinases-1; VCAN – versican; chondroitin sulfate ECM proteoglycan Heatmaps of angiogenesis related normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). TIMP1 – tissue inhibitor of metalloproteinases-1; VCAN – versican; chondroitin sulfate ECM proteoglycan Figure  8 A and B present genes related to extracellular matrix organization for pcCTC and CTC candidates and single platelets respectively. Genes such as CD44 and UTRN were downregulated for single platelets, but upregulated for most of the CTC and some pcCTC candidates. Additionally, several genes were highly expressed in both single platelets and the pcCTC candidates with the highest platelets score (three on the left side of the chart). These genes were as follows: ACTG1 , TGFB1 , TIMP1 , CD47 (integrin associated protein (IAP)), SPARC , ADAM10 , ITGA2B and ACTB . Among these genes ITGA2B was used as a platelets marker. Figs. 8 Heatmaps of normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). The genes are related to extracellular matrix organization Heatmaps of normalized gene expression for pcCTC and CTC candidates ( A ) and for single platelets from cancer patients ( B ). The genes are related to extracellular matrix organization

Metastasis is the primary cause of cancer-related deaths [ 1 ]. Circulating tumor cells, which enter the bloodstream after detaching from the tumor, are key in this process, acting as the “seeds” that disseminate the disease to distant organs [ 2 ]. In the bloodstream, circulating tumor cells are exposed to multiple hostile factors, including hemodynamic shear stress [ 3 ] and immune surveillance [ 4 ], which together severely limit their survival. As a result, CTCs are extremely rare, typically occurring at fewer than 10 cells per milliliter of blood, among billions of red blood cells and millions of leukocytes. The abundance of CTCs varies depending on cancer type, disease stage, and tumor burden [ 5 ]. Beyond their scarcity, CTCs exhibit considerable heterogeneity, particularly with regard to their epithelial-mesenchymal transition phenotype [ 6 ], but also in other characteristics, partially recapitulating the heterogeneity of cancer cells in the tumor [ 7 ]. The rarity and heterogeneity of CTCs make these cells challenging to identify and further characterize. In addition to their intrinsic heterogeneity, CTCs dynamically interact with components of the bloodstream—most notably platelets—which can shield them from immune detection and may influence their transcriptional programs, including EMT and stemness-associated pathways. Complementing CTC-based approaches, platelet RNA is emerging as a novel field in cancer research, offering a complementary and potentially more accessible biomaterial [ 8 ]. Platelets are anuclear cell fragments derived from mature megakaryocytes that play crucial roles beyond their well-established functions in hemostasis and thrombosis. These abundant blood components can absorb RNA from tumor cells and other components of the tumor microenvironment, including extracellular vesicles (EVs) [ 9 ]. This process, known as tumor education of platelets, results in the formation of tumor-educated platelets (TEPs), which carry specific RNA profiles reflective of the tumor’s presence and activity [ 10 ]. Interactions between platelets and tumor cells have been extensively studied. Platelets may infiltrate into the tumor stroma and adhere to the cancer cells via cell adhesion receptors [ 11 ]. They can modulate the microenvironment through the release of immunomodulatory [ 12 ], pro- and anti-angiogenic factors [ 13 ], and may induce proliferation of cancer cells [ 14 ]. In the bloodstream, platelets may also adhere to CTCs, similarly to how they adhere to cancer cells within the tumor. The platelet-CTC interaction is thought to be strong and occur quickly after intravasation of the cancer cell [ 15 ] platelet-secreted factors were observed to decrease the anti-cancer functions of NK cells by modulating the HLA-E: CD94-NKG2A checkpoint [ 16 ]; platelets were also observed to modulate the functions of other immune cells [ 17 ]. Cloaking platelets are also known to promote a “normal-like” phenotype of CTCs by transferring their MHC-I molecules presenting physiological antigens, which inhibits the function of NK cells, thus protecting the CTCs from lysis [ 18 ]. CTCs, as they are detached and under blood pressure, are prone to anoikis. The platelet cloak is postulated to shield cancer cells from fluid shear stress and, enhancing CTC survival and thus facilitating metastasis formation [ 19 , 20 ]. Furthermore, platelet-covered CTCs may promote interactions with the endothelium and extracellular matrix (ECM) during extravasation, aiding in the reorganization of ECM structure to facilitate tumor cell adhesion, migration, and metastatic colonization [ 15 ]. Moreover, platelets can enhance CTC aggressiveness by transferring EMT-stimulating factors, particularly transforming growth factor β (TGFβ) and platelet-derived growth factor (PDGF) [ 21 , 22 ]. Recent findings suggest that platelets influence other aspects of cancer cell biology such as metabolic reprogramming [ 21 ], immune evasion [ 23 ], and stemness [ 24 ]. Despite their potential importance, the molecular features of pcCTCs remain poorly understood. Advances in CTC isolation and single-cell RNA sequencing revealed differences in gene expression signatures relating to EMT, stemness, drug resistance and other cancer-related processes, leading to a better understanding of the role of these cells in disease progression [ 25 ]. Importantly, CTCs may exist in three main phenotypic states: epithelial, mesenchymal, or an in-between epithelial-mesenchymal; all these states may be found in samples from the same patient, although CTCs from the same patient were observed to be skewed in the direction of one particular phenotype [ 26 ]. The delineation of the CTC phenotype is crucial, as the the presence of mesenchymal CTCs are related to a worse outcome [ 27 ] and CTCs belonging to each of these phenotypes may potentially interact with platelets differently. The study of all the CTC phenotypes is challenging due to many isolation methods, based on size, rigidity and EpCAM expression, yielding primarily epithelial CTCs [ 28 ]. Wenta et al. have shown that in breast and ovarian cancer patients’ blood samples CTCs of the epithelial, mesenchymal and mixed phenotypes there may be found using flow cytometry with imaging, and that CTCs of different phenotypes have a markedly different size [ 29 ]. Moreover, few studies have focused on detecting platelet‑covered circulating tumor cells. In one study the authors first used hydrodynamic, size‑based sorting to remove free platelets, and then employed a herringbone CTC chip functionalized with platelet markers CD41 (integrin α2b (ITGA2B)) and CD61 (integrin β3 (platelet glycoprotein [GP] IIIa)) to selectively capture pcCTCs from blood samples from patients with lung and breast cancer as well as melanoma [ 30 ]. Another study used the high-definition single-cell assay methodology based on cell imaging data analysis to identify rare cells in the blood. Whole blood samples from prostate cancer patients were fixed and stained for immunofluorescence, targeting cytokeratins to identify epithelial tumor cells, vimentin (VIM) to capture mesenchymal tumor cells, CD45 (protein tyrosine phosphatase receptor type C (PTPRC), leukocyte common antigen (LCA)) to exclude leukocytes and CD31 (platelet endothelial cell adhesion molecule-1 (PECAM-1)) to detect platelets signal. Cells positive for pan-cytokeratin with associated CD31 signals were classified as epithelial pcCTCs and those positive for pan-cytokeratin, VIM and CD31 as mesenchymal pcCTCs. This study also performed survival analysis connecting pcCTC with the worst prognosis [ 31 ]. In [ 32 ] pcCTCs were detected in blood samples from prostate cancer patients using size‑based filtration with the ScreenCell ® device to enrich CTCs. Captured cells on the filters were stained with May‑Grünwald‑Giemsa and analyzed by a skilled pathologist using stringent morphological cytology criteria. CTCs exhibiting at least one platelet adherent to their surface were classified as platelets‑cloaked. In another study, pcCTCs were isolated from peripheral blood of patients with pancreatic cancer using a label-free centrifugal microfluidic device that separates single CTCs and CTC clusters based on cell size and deformability. Captured cells were subjected to molecular characterization by RT-qPCR and targeted RNA profiling of 48 genes that are associated with EMT phenotypes, stemness and the presence of platelets, neutrophils, macrophages, and fibroblasts. The expression of platelet-associated genes ( ITGA2B , SELP , PDGFB ) within isolated CTCs or CTC clusters was used to identify platelet cloaking. Interestingly, most CTCs expressing platelet markers displayed a predominantly mesenchymal rather than epithelial phenotype. The authors further demonstrated that the presence of CTC clusters expressing platelet markers was associated with poor prognosis [ 33 ]. Overall, the reported unfavorable outcomes associated with the detection of pcCTCs (vs. naked CTCs) indicate the biomarker potential of pcCTCs. While few studies have confirmed the presence of pcCTCs in peripheral blood, there is a gap in understanding the true phenotype of these cells and platelets attached to them. Prior to addressing this gap, we performed high-resolution immunofluorescence imaging with Cytek ® Amnis ® ImageStream ® X Mk II imaging flow cytometer to directly confirm the presence of platelet-covered CTCs in peripheral blood of ovarian cancer (OC) patients whose blood was biobanked according to the protocol developed in our lab. Next, in PBMCs fractions of unmatched patients, we conducted single-cell transcriptomic profiling to explore possible platelet-induced transcriptional changes in pcCTCs as opposed to naked CTCs. By applying a validated marker panel, we were able to distinguish candidate pcCTCs from free platelets and non-cloaked candidate CTCs, and further dissect their gene expression profiles, including signatures related to EMT, platelet activation or angiogenesis. In this study we will be referring to detected cells as CTC or pcCTC candidates, since transcriptome-based identification alone is non-definitive and may be affected by CTC heterogeneity and technical confounders, requiring orthogonal validation to confirm malignant origin. This integrative approach not only provides direct visual evidence of pcCTCs in circulation but also offers, for the first time in breast and ovarian cancer, a detailed transcriptional characterization of these potential hybrid cellular entities. The results are additionally verified with bulk RNA sequencing of platelets. Our findings lay the groundwork for future biomarker development and suggest that pcCTCs may serve as a dynamic window into tumor–platelet interactions, with potential relevance for metastasis prediction, therapy monitoring, and liquid biopsy design.

CTCs are exceptionally rare cells in the bloodstream of cancer patients, typically occurring at frequencies of only a few cells per several million leukocytes. Moreover, many currently available enrichment approaches are unable to capture the full phenotypic and biological heterogeneity of CTCs. These factors make the investigation of CTC biology, including platelet–CTC interactions, particularly challenging, especially in the context of epithelial–mesenchymal transition (EMT). The abundance of these cells may be influenced by tumor stage, as tumors at later stages may shed more cells entering the bloodstream. CTCs themselves shift to a mesenchymal phenotype at later stages [ 44 ]. Importantly, post-EMT CTCs were observed to be more likely to be cloaked by platelets [ 45 ], and as such, more pcCTCs may be found as cancer progresses. In this study, we focused on detecting CTC and pcCTC candidates in PBMCs. We identified 12 CTC candidates: one epithelial pcCTC, three mesenchymal pcCTCs, three ovarian cancer–associated pcCTCs, three ovarian cancer–associated naked CTCs, and two naked mesenchymal CTCs. In addition, we characterized the transcriptomic profiles of the detected CTC and pcCTC candidates and highlighted the differences between these cell types and platelets. In our study, in regards to CTC candidates with clearly ascribed EMT phenotype, we observed a single naked epithelial CTC candidate, 2 naked mesenchymal CTC candidates and 2 mesenchymal pcCTC candidates. These low numbers of identified (pc)CTC candidates, stemming from a low number of patients enrolled and the relative homogeneity of the cohorts (e.g. all of the OC patients having stage III or IV disease), make it challenging to refer to the aforementioned observations regarding CTC’s EMT phenotype’s influence on interactions with platelets, although it should be noted that the single epithelial CTC found was not coated. A larger, more heterogeneous patient cohort should be similarly studied to elucidate the relationship between CTC phenotype, disease stage and cloaking status. Eventually we distinguished three groups of genes: (1) genes elevated in the most platelet‑covered circulating tumor cells (pcCTCs), (2) genes elevated in “naked” CTCs as well as in less platelet‑covered, and (3) genes shared by both groups. In the first group, we included platelet‑associated genes such as PPBP and PF4 , which reflect strong coupling of CTCs with platelets (immune shielding, facilitation of emboli and adhesion) [ 19 ]. Moreover, the increased expression of tubulins ( TUBA4A , TUBB1 , TUBB2A ) as well as actin and motor‑regulator proteins (ACTG1 , ACTB , ACTN1 , TPM1 , TPM4 , MYL9 , MYLK , FLNA ) in the most platelet‑covered CTCs suggests a phenotype favoring deformability, transit through the circulatory microenvironment, and adhesion during sequestration by platelets or the vessel wall [ 46 ]. Additional genes, such as integrins ( ITGB1 , ITGB5 , ITGA2B ) and SPARC , may contribute to extracellular matrix remodeling, adhesion, and signal transduction [ 47 ]. Upregulation of CD47 further facilitates evasion of phagocytosis [ 48 ]. Importantly, ORAI1 (calcium release-activated calcium channel protein 1) has been reported to act pro‑tumorigenically in breast cancer and to increase resistance to therapy in ovarian cancer [ 49 , 50 ]. We also observed elevated levels of ribosomal and translation‑associated genes ( RPS19 , RPS7 , RPL4 , FAU , EIF1B , MEPCE ), which often reflect enhanced protein synthesis or translational stress in malignant cells [ 51 ]. Furthermore, TIMP1 influences epithelial–mesenchymal transition, metastasis, and chemoresistance in ovarian cancer [ 52 ]. Together, these genes indicate a phenotype better adapted for metastasis and survival of tumor cells. Among the genes elevated in naked CTC candidates and less platelet‑covered CTC candidates, we detected genes involved in DNA repair ( XRCC5 and ATM ) [ 53 , 54 ] and metastasis promotion ( PITPNC1 and ETS1 ) [ 55 , 56 ]. Among the genes common in both groups, B2M plays a role in immune evasion, as it is essential for the stabilization and surface expression of MHC class I [ 57 ]. In addition, MALAT1 , a well‑characterized lncRNA, is associated with EMT and metastasis [ 58 ]. ARHGAP15 may influence the directionality of cell movement and invasiveness [ 59 ], while RAP1B can regulate CTC adhesion to the endothelium. These represent core mechanisms that facilitate CTC survival and transmigration [ 60 ]. According to our knowledge, this is the first study to identify pcCTC candidates within PBMCs sequenced at single cell resolution. As shown in the literature [ 31 ], the presence of pcCTCs is associated with poorer patient prognosis, and our findings indeed demonstrate that pcCTCs exhibit a broader range of pro-survival and pro-metastatic mechanisms. This indicates that antiplatelet treatments indeed can be useful in cancer therapy [ 61 ], specifically P-selectin inhibitors combined with immune therapy due to a reported immune evasion-enhancing properties of the platelet coat [ 16 – 18 ], as the “uncoating” of CTCs would expose them to immune cells. Moreover, for all of the patients in whom CTC candidates were detected, distant metastasis was not diagnosed, which indicates that our method can work on at earlier stages of disease dissemination. In addition, in both the single-cell dataset and the Amnis ImageStream dataset, most of the detected cells (7 of 12–58% in the single-cell dataset and 289 of 454 − 64% in the Amnis ImageStream dataset) were covered with platelets, underscoring the importance of pcCTC detection. Therefore, our detection method may hold potential clinical value. This study has several limitations. First, the number of samples from breast and ovarian cancer patients is limited, and therefore not fully representative of the overall heterogeneity of these cancers. A similar limitation applies to the detected cells—their number is also restricted and may not capture the full heterogeneity of cancer cell populations. In addition, we focused on genes associated with specific biological processes and selected the top 30 most differentiating genes between pcCTCs and CTCs, as well as between pcCTCs and platelets (both at the single-cell and bulk level.). This targeted approach does not encompass the entire transcriptome, and additional insights could potentially be gained from genes that were not included in our analysis. Moreover, although we made reasonable, literature-based decisions when selecting gene panels for detecting CTCs, pcCTCs, and for filtering out various cell types, we acknowledge that these panels could be further expanded, optimized, and made more specific in future studies. We also acknowledge that some CTC candidates may be lost during sample processing and recovery. Although recovery rates were not assessed as part of the present study, previous spike-in experiments performed at the Laboratory of Translational Oncology using the same PBMC-based CTC enrichment workflow demonstrated recovery rates of 54–72% following the addition of MDA-MB-361 breast cancer cells to healthy donor blood samples [ 62 ]. These findings indicate that, while some cell loss is unavoidable, the workflow is capable of retaining a substantial proportion of rare tumor cells. Accordingly, our analyses demonstrate that CTC candidates, including cloaked pcCTC candidates, can be delineated from PBMC-derived material, although the identified populations may represent only a subset of the full CTC spectrum. Importantly, all samples were processed under standardized conditions and within predefined time intervals, with viability and quality-control assessments performed prior to single-cell library preparation to maximize cell integrity and data quality. Additionally, BC (Singleron Biotechnologies) and OC/benign (10x Genomics) samples were processed using different single-cell sequencing platforms. Although these platforms were not benchmarked directly against one another, the consistent identification of CTC candidate populations across these datasets supports the robustness of the applied detection strategy and suggests that the observed signatures are not restricted to a single technological platform. Besides expanding gene panels, current CTC isolation methods should be optimized to obtain (pc)CTCs at a higher rate, especially taking into consideration the phenotypic diversity of CTCs and the reported increased proclivity of platelets to cloak mesenchymal CTCs, for example by shifting away from the EpCAM-dependent isolation. The degree of coating should also be studied in a hypothetical optimized isolation pipeline. Currently, there is no agreement upon gene expression panel that would delineate low- and highly-coated pcCTCs. The analysis results presented herein point to a higher expression of deformability-, immune evasion-, and adhesion-related genes in pcCTCs with a high expression of platelet genes as calculated using UCell. We hypothesize that this demonstrates the upregulation of these processes in the CTC related to a high degree of coating. The previously-mentioned genes could be taken into account when designing a standard “Cloaking Index” gene panel. However, it should first be validated in a larger cohort, preferably also using imaging techniques; an AI-powered approach, such as Deepcell, would allow to accurately determine that parameter [ 63 ]. Only after the development of a pipeline allowing for an accurate CTC isolation may the pcCTC-related gene signatures be viable clinically. Additionally, CTC inference was based on transcriptomic profiles from PBMC sequencing without dedicated CTC enrichment, reflecting a trade-off between maximizing cell viability/RNA quality and achieving definitive CTC identification. Moreover, we are not able to ensure that detected cells are indeed circulating tumor cells, due to CTC heterogeneity and technical confounders such as dropouts, ambient RNA and contamination as well as aforementioned limitations [ 64 ]. While the identified pcCTC candidates exhibited transcriptomic features consistent with platelet-coated CTCs, the possibility of occasional CTC–platelet doublets cannot be completely excluded. To our knowledge, no validated gene panel or analytical framework is currently available to reliably distinguish between these entities in single-cell transcriptomic datasets.

In this study, we developed a method for detecting putative platelet-cloaked circulating tumor cells within the peripheral blood mononuclear cell fraction, alongside naked circulating tumor cells. Comparative transcriptomic analyses showed substantial similarities between pcCTC candidates and platelet profiles derived from both single-cell and bulk RNA sequencing datasets, supporting the presence of platelet-related features in a subset of detected CTCs. Integration of flow cytometry and single-cell sequencing data further indicated that most CTC candidates were associated with platelet markers, highlighting the biological relevance of pcCTC detection. Gene-level analyses suggested that pcCTC candidates may exhibit enhanced mechanisms supporting survival in the bloodstream, immune evasion, and metastatic potential. Together, these observations imply that platelet cloaking could contribute to the increased adaptability of CTCs and may be linked to poorer clinical outcomes. Future work should aim to validate the proposed detection strategy in larger, independent cohorts of ovarian and breast cancer patients. Extending the analysis to additional cancer types would help establish a more comprehensive view of pcCTC biology. Moreover, evaluating the impact of antiplatelet therapies on pcCTC prevalence, cellular behavior, and disease progression may provide important insight into their potential clinical and therapeutic relevance.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2