Computational ranking identifies Plexin-B2 in circulating tumor cell clustering with monocytes in breast cancer metastasis

Schuster, Emma; Dashzeveg, Nurmaa K.; Tong, Fangjia; Jia, Yuzhi; El-Shennawy, Lamiaa; Zhang, Tong; Hoffman, Andrew D.; Kitata, Reta Birhanu; Kibria, Golam; Zhang, Youbin; Squires, Joshua R.; Zheng, Chunlei; Ramos, Erika; Taftaf, Rokana; Scholten, David; Almubarak, Hannah F.; Adorno-Cruz, Valery; Sullivan, David P.; Reduzzi, Carolina; Minor, Allegra C.; Purev-Ochir, William; Spahija, Sabina; Xu, Rong; Siziopikou, Kalliopi P.; Platanias, Leonidas C.; Shah, Ami; Muller, William A.; Gradishar, William J.; Cristofanilli, Massimo; Tsai, Chia-Feng; Shi, Tujin; Liu, Huiping

doi:10.1038/s41467-025-62862-z

Download PDF

Article
Open access
Published: 16 August 2025

Computational ranking identifies Plexin-B2 in circulating tumor cell clustering with monocytes in breast cancer metastasis

Nature Communications volume 16, Article number: 7649 (2025) Cite this article

7398 Accesses
4 Citations
18 Altmetric
Metrics details

Subjects

Abstract

Multicellular circulating tumor cell (CTC) clusters can be up to 50 times more efficient than single CTCs in mediating viable metastasis. Here, combining computational ranking and functional determination, we identify the transmembrane protein Plexin-B2 (PLXNB2) as one of the top molecular targets associated with unfavorable distant metastasis-free survival, showing enriched expression in CTC clusters versus single CTCs from patients with advanced breast cancer (mostly female). Loss of PLXNB2 (Plxnb2) reduces the formation of homotypic tumor cell clusters and heterotypic tumor-myeloid cell clusters, reducing spontaneous metastases in female mice bearing human (mouse) breast cancer. Interactions of PLXNB2 with its ligands SEMA4C on tumor cells and SEMA4A on myeloid cells (monocytes) promote homotypic and heterotypic CTC cluster formation, respectively, thereby driving lung metastasis. Global proteomic analysis reveals downstream effectors of the PLXNB2 pathway associated with tumor cell clustering. Thus, PLXNB2 is a therapeutic target for preventing new metastasis in breast cancer.

Comprehensive analysis of the PLXNA3 gene on prognosis and immune characteristics in breast cancer

Article Open access 04 August 2025

Identification of epithelial and mesenchymal circulating tumor cells in clonal lineage of an aggressive prostate cancer case

Article Open access 21 June 2022

Detection of circulating tumor cells by means of machine learning using Smart-Seq2 sequencing

Article Open access 14 May 2024

Introduction

The rapid development of cutting-edge multi-omic analyses and computational modeling have facilitated the integration and transformation of bioinformatic data into phenotype-related discoveries^1,2,3,4,5. Reciprocally, comprehensive experimental determination feeds back computational analysis-based data mining and prediction. While genomic and transcriptomic studies have been relatively extensive, our study seeks to synergize the power of phenotype-driving proteomic analysis and functional exploration for cancer discoveries. One of the most devastating features of solid tumors is distant spreading or stage IV cancer metastasis, predicting unfavorable overall survival for all breast cancers⁶, especially estrogen receptor (ER) negative breast cancers, such as triple-negative breast cancer (TNBC) which lacks expression of ER, progesterone receptor, and epidermal growth factor receptor 2 (HER2)^7,8,9 with the lowest 5-year survival rate (~10%) after metastasis to the lungs, brain, and liver¹⁰, followed by HER2-positive breast cancer (~40%)⁶. As such, our goal is to develop a computational ranking of protein candidates and identify new therapeutic targets that drive breast cancer metastasis.

Cancer is disseminated by circulating tumor cells (CTCs) that shed off the primary tumor and are capable of seeding and regenerating tumors in distant organs, including lung, liver, brain, and bone, due to their inherent and acquired properties, such as plasticity, proliferation, and intercellular interactions^11,12,13,14. The dogma of single CTC-mediated cancer dissemination has been challenged by the detection of rare CTC clusters in the blood of patients with advanced breast cancer^{11,15,16,17,18,19}. The existence of CTC clusters predicts unfavorable outcomes^{11,15,16,17,18,19}, as they are 20–50 times more likely to seed metastases than single CTCs^{11,20,21,22,23,24}. To control and prevent metastatic disease, it is imperative to discover the diverse mechanisms of CTC clusters in cancer. Two cellular mechanisms have been proposed for CTC cluster formation¹⁸; one is collective dissemination or cohesive shedding^11,25, and another is tumor cell aggregation^16,26, which would lead to both homotypic and heterotypic tumor clusters, thereby promoting metastasis with unfavorable overall survival (OS)¹⁸. It remains an open question whether primary tumor proteomic profiles can be used to identify phenotypic drivers for early intervention, and guide targeting approaches to prevent homotypic and heterotypic tumor cluster formation and block metastasis.

In this work, we hypothesize that aggregation or cohesion phenotype-related adhesion proteins can regulate CTC cluster formation. Taking advantage of systems biology and mass spectrometry (MS)-based global proteomic data, we have developed a ranking method to assess all 608 cell adhesion molecule candidates (derived from the Molecular Signature Database^27,28,29 of the Gene Ontology Biological Processes^30,31) in breast cancer. As a proof-of-concept, our study reveals a single-pass transmembrane protein, Plexin-B2 (PLXNB2), as one of the top candidates. Plexin-B2 is up-regulated in primary tumors and enhances the formation of both homotypic tumor cell clusters and heterotypic CTC-myeloid cell clusters in metastasis of breast cancer, especially TNBC.

Results

Proteomic ranking of cell adhesion molecules in breast cancer

To seek an unbiased in silico screen of candidate proteins in primary tumors that may regulate cancer progression and metastasis, we developed a mathematical ranking score, Rscore, to assess the significance ranking of adhesion network proteins in contributing to intercellular interactions. In our initial proof-of-concept analyses, the Rscore integrates individual rankings (r_i) of each adhesion molecule in various datasets with an adjustable weight of constant factors (c_i), including relative protein abundance (normalized intensity), tumor-specific expression versus normal adjacent tissues (p value, fold change, and absolute change), and clinical associations with OS and distant metastasis-free survival (DMFS) (p value and hazard ratio) (Fig. 1a, and Supplementary Fig. S1a). We have assessed the Rscore rankings of the proteins, especially adhesion proteins overlappingly detected in lab-obtained and public datasets via MS proteome quantifications, including human breast tumors³² (N = 122), TNBC tissue voxels versus normal adjacent regions (N = 6), breast cancer cell lines (N = 2), and patient-derived CTC specimens (N = 19), in association with clinical outcomes, especially OS and DMFS in multiple datasets of breast cancer.

**Fig. 1: PLXNB2 expression in breast tumors is associated with poor prognosis and enriched in CTC clusters.**

Based on the MS proteomic analyses of human TNBC tissues from laser capture microdissection (tumor regions versus normal adjacent), we obtained a list of tumor-specific proteins, including 627 differentially expressed proteins in tumor tissues (Supplementary Data 1). After expanding the analysis to the tandem MS proteomic datasets of treatment-naive human breast tumors³² (N = 122), 398 adhesion proteins were detected and ranked based on the relative protein abundance, calculated as spectral counts per protein and normalized by the ratio of a given protein’s length (# of amino acids) versus the median protein length (Supplementary Fig. S1a, Supplementary Data 1). To further narrow down to the intrinsic adhesion proteins in cancer cells, we combined the MS proteomic profiles of human TNBC cells (MDA-MB-231³³ and Hs578T³⁴) and identified 124 overlapped adhesion proteins with ranking in normalized peptide-spectrum match (PSM) counts (Supplementary Fig. S1a, and Supplementary Data 1-tabs 3–4). We also detected 258 adhesion proteins from patient CTCs and CTC-derived PDX CTC-205 and CTC-92 (N = 19 specimens) (Supplementary Fig. S1a, and Supplementary Data 1-tabs 5)

The Venn diagram of the above datasets produced a list of 30 overlapped proteins as top candidates for integrated assessment via Rscore of relative abundance and tumor specificity in the above MS datasets as well as clinical association with OS at protein and mRNA expression levels via KMPlotter³⁵ and Protein Atlas³⁶. We identified Plexin-B2 (encoded by PLXNB2, PB2) as the top-ranked oncoprotein in a negative association with patient OS across three clinical datasets (Fig. 1a, and Supplementary Fig. S1b-d, Supplementary Data 1-tabs 6-7). Of note, the list of the top 30 includes a previously characterized surface protein regulator of homotypic CTC cluster formation, CD44¹⁶, within the most abundant proteins in cancer cells (Supplementary Fig. S1c), validating the potential relevance of listed adhesion molecules.

Our previous studies demonstrated that cancer cell-derived extracellular vesicles (EVs) and their surface proteins CD44 and CD81 contribute to the modulation of the microenvironment and EV-recipient cells³⁷. To further assess the potential relevance of candidate proteins in cancer-specific secretome for cancer progression^38,39,40,41, we performed MS proteomic analyses of small EVs secreted by breast cancer cells (MDA-MB-231, SKBR3, BT4T4, and MCF7) and normal immortalized epithelial cells (HEK293, MCF10A and MCF12A). While most of the EV proteins were glycoproteins and membrane-bound proteins, PLXNB2 was also among the top enriched proteins in cancer EVs versus normal cell EVs (Supplementary Fig. S1e, and Supplementary Data 1-tabs 8–9). However, subsequent immunoblotting analysis revealed a truncated form of PLXNB2 (also known as β subunit⁴²) in breast cancer cell-derived EVs which did not contain the full-length protein (Supplementary Fig. S1e), therefore the PLXNB2 functions in EVs were not further investigated and we focused on characterizing its cellular functions instead in this study.

PLXNB2 expression is associated with unfavorable survival and enriched in CTC clusters

In KMPlotter³⁵ analysis of human breast cancer datasets³⁶, we found that high PLXNB2 expression is negatively associated with both OS and DMFS in all breast cancers and ER-negative breast cancers or grade 3 breast cancers (Fig. 1b–d, and Supplementary Fig. S1f and S1h). The other candidates following PLXNB2 include PPIA, EZR, and TLN1 (Supplementary Fig. S1g). We also analyzed a tissue microarray of advanced breast cancers collected at Northwestern University which showed a high expression of PLXNB2 across different breast cancer subtypes in association with distant metastasis (Supplementary Fig. S2a–b, and Supplementary Data 2).

We then measured the PLXNB2 expression in human blood CTCs and white blood cells (WBCs) from patients with stage III-IV breast cancer using multiple complementary methods^16,26: FDA-approved CellSearch® with blood collected into fixative-containing CellSave tubes (N = 15 patients) (Fig. 1e, and Supplementary Fig. S2b–i), live cell flow cytometry^18,26 with blood drawn into EDTA tubes (N = 17 patients) (Supplementary Fig. S3a–b), immunoblotting (Supplementary Fig. S2d), and MS proteomic analysis (Supplementary Data 1-tabs 5). WBCs were recognized with the leukocyte marker CD45 expression.

Based on the CellSearch analyses, PLXNB2^high breast cancer correlated with detectable blood CTCs, stained as cytokeratin (CK)⁺DAPI⁺CD45⁻ cells after enrichment via anti-EpCAM magnetic beads (Supplementary Fig. S2b–c). Compared to single CTCs, PLXNB2 expression was significantly higher in CTC clusters, both homotypic CTC-CTCs and heterotypic CTC-WBC clusters (Fig. 1f–g). Flow cytometry analysis confirmed that a larger proportion of homotypic CTC clusters were PLXNB2-positive than single CTCs (Supplementary Fig. S3b–e). In orthotopic TNBC patient-derived xenografts (PDXs) labeled with luciferase 2-eGFP (L2G)¹³, we observed a dynamic increase of surface PLXNB2 expression in CTCs and lung metastases in comparison to the primary tumor cells (Supplementary Fig. S3f). Furthermore, the detection of PLXNB2 on cell surface was resistant to trypsin digestion (Supplementary Fig. S3g). Breast tumor cells presented PLXNB2 in full-length ( ~ 200 kD) and truncated ( ~ 75 kD) proteins, both of which were lost upon CRISPR/Cas9 mediated gene knockout (KO) (Supplementary Fig. S2d).

Loss of PLXNB2 reduces metastasis and CTC cluster formation in human breast tumors in vivo

PLXNB2 is a single-pass transmembrane plexin family member, and its primary function is to direct neural cell growth and migration in brain development; it also plays a role in the function of the vascular and endocrine systems, wound healing, monocyte function, and neuro progenitor cells^{43,44,45,46,47,48,49,50}. However, PLXNB2’s molecular mechanism was not previously studied in the context of CTC clusters and metastatic breast cancer. The existence of CTC clusters in patient blood has previously been shown to be associated with metastasis and reduced OS in breast cancer^11,22. Since our work revealed that PLXNB2 is highly expressed in patient CTCs, we continued to determine if PLXNB2 promotes CTC cluster formation and spontaneous metastasis in vivo.

Taking advantage of the CRISPR/Cas9 technologies and the target protein detection via flow cytometry, we transduced the MDA-MB-231 cells with lentiviral PLXNB2 gRNAs and then sorted multiple pools of PLXNB2 knockout cells (KO1 pool 1 and KO2 pool 2) without clonal selection based on the negative expression of the surface receptor PLXNB2 within 72 hours. These PLXNB2 KO cells did not significantly alter cellular growth (confluence), viability, cell migration, and invasion in vitro as well as 10,000 tumor cell-mediated growth curves in vivo (Supplementary Fig. S4a–h) whereas the tumorigenesis in serial dilutions of 10 to 1000 cell implants was compromised (Supplementary Fig. s4i–k). Notably, the tumor growth phenotype of stable PLXNB2 KO tumor cells (via Cas9/gRNAs) might have resulted from selective pressure in KO cell maintenance, differing from the variable effects of siRNA-mediated transient KD of PLXNB2 in our following studies (shown in Supplementary Fig. S5) as well as the reported inhibitory effects of shRNA-mediated PLXNB2 knockdown (KD) on tumor cell proliferation⁵¹. We speculate that the CRISPR Cas9-generated KO and selective growth pressure resulted in a compensation of pathways for the restored proliferation of these cells.

After orthotopically implanting 10,000 tumor cells into the fourth mammary fat pads of NSG mice to ensure tumor growth of both L2G-labeled WT (pool control clonalities) and PLXNB2 KO MDA-MB-231 cells (pooled KO cells), we assessed spontaneous metastasis of these tumors to the lungs (Fig. 2a). Mice were monitored for 8–10 weeks until the experimental endpoint for collections of tumors, lungs, and blood. Compared to control tumors, the PLXNB2 KO tumors did not show a significant difference in Ki67-indicated proliferation (P = 0.14) but a trend of lighter tumor weight (P = 0.06) (Fig. 2b, c, and Supplementary Fig. S4c–e). These data suggest that stromal factors and host interactions with KO tumor cells might contribute to the borderline changes in tumor burden in vivo. The mouse lungs, however, showed a significant 23-fold reduction of the metastatic burden from PLXNB2 KO tumors compared to those of control tumors (Fig. 2d–e), which were normalized by tumor weight. L2G⁺ CTCs were analyzed after blood collection via heart puncture from each mouse and by H&E staining analyses of vascular CTC in situ within the lung sections. PB2 KO reduced the CTC cluster formation, as detected from the blood and within lung tissue sections; and PB2 KO inhibited spontaneous metastatic colonization (metastatic lesions and cell numbers) (Fig. 2f–h, and Supplementary Fig. S4f). These data demonstrated that PLXNB2 depletion inhibits CTC cluster formation and blocks spontaneous metastasis of human breast cancer in vivo.

**Fig. 2: PLXNB2 depletion abolishes spontaneous metastasis and CTC cluster formation in TNBC in vivo.**

To determine if PLXNB2-mediated CTC clustering was associated with increased co-colonization in metastasis, we orthotopically implanted red L2T- and green L2G-labeled PLXNB2⁺ WT control tumors (ConT and ConG) and PB2⁻ KO tumors (KOT and KOG) into separate left and right 4^th mammary glands with 4 groups of combinations: (1) ConT-ConG, (2) KOT-KOG, (3) ConT-KOG, and (4) KOT-ConG (Fig. 2i). After 6 weeks of orthotopic tumor growth, only the mice bearing the ConT-ConG tumors in dual colors showed dual-color lung colonies whereas the counts of both single-color and dual-color metastatic colonies dramatically decreased in any of the three groups with one or two KO tumor implants (Fig. 2j-m). The CTC clusters were dramatically in higher frequencies in control tumor-bearing mice than the KO tumor-bearing mice (Fig. 2m). These data are in consistency with previous demonstrations that lung co-colonization (polyclonality) are contributed by CTC clusters of breast cancer¹⁶ and pancreatic cancer⁵², albeit a possibility of sequential seeding of polyclonal tumor cells.

To explore the effects of PB2 KO on animal survival, we found that the control tumor-bearing mice survived up to 8 weeks whereas the KO tumor-bearing mice went up to 10 weeks with relatively comparable tumors before they exceeded the allowable tumor sizes for euthanasia (Fig. 2n). Nevertheless, the blood CTC clusters and lung metastases from the control tumors remained over 30 times higher than those from the KO tumors in mice (Fig. 2o–p). Notably, the majority of the CTCs from both control and KO tumor groups were arrested in G0/G1 phases without significant proliferation differences (Fig. 2o), in consistency with comparable Ki67-indicated proliferation of lung metastases of these tumors (Supplementary Fig. S4e). These results suggest that PLXNB2 depletion-reduced metastases are independent of its proliferation effects on the CTCs, or the cells disseminated into the lungs.

PLXNB2 depletion compromises tumor cell clustering and mammosphere formation

After identifying PLXNB2’s role in promoting CTC clusters and metastasis, we continued to determine its function in tumor cell clustering in specific breast cancer subtypes such as TNBC and HER2⁺ in which metastasis is common. To do this, we sorted L2G-labeled PLXNB2^high and PLXNB2^low/− tumor cells from orthotopic TNBC PDX models (TN3) which develop spontaneous lung micro-metastases¹³ (Fig. 3a). Next, using the IncuCyte Live Cell Imager® as previously described¹⁶, the clustering of primary PDX tumor cells on collagen-coated plates was monitored over time. After 6–8 h of clustering, PLXNB2^high cells formed significantly more clusters ( > 2-3 cells) than PLXNB2^low cells (Fig. 3b, c). We then transiently knocked down PLXNB2 in multiple TNBC cell lines (human MDA-MB-231 and HS578T, and mouse 4T1) as well as HER2⁺ cell line SKBR3, using both SmartPool siRNA (siPB2) and individual siRNAs (siPB2−09, −10, −11). The reduction or loss of the full-length and truncated PLXNB2 was validated by immunoblotting and/or flow cytometry (Supplementary Fig. S5a, S5c–d, S5j). MDA-MB-231 cells transfected with siPB2 started to show slower growth (confluence) than the control cells at 24-48 h after seeding, whereas siPB2−10 did not have a significant effect on the growth (Supplementary Fig. S5a). However, over a short period of 4-6 h without compromised cell viability and with minimal influence by cell proliferation/growth, both siPB2 and siPB2−10 comparably reduced the size of tumor clusters of all five tested models, including human MDA-MB-231, MDA-MB-468, SKBR3, and HS578T cells, and mouse 4T1 cells (Fig. 3d–g, and Supplementary Fig. S5a-n), suggesting that PLXNB2 is required for tumor cell clustering in breast cancers.

**Fig. 3: PLXNB2 promotes tumor cell clustering and metastatic dissemination.**

Consistent with the effects of siPB2 and siPB2−10 (Supplementary Fig. S5), both PB2-KO1 and -KO2 cell pools depleted tumor cell clustering (Supplementary Fig. S6a–b). However, PB2 KO cells did not significantly alter the FAK phosphorylation, cell growth (confluence), and cell motility (migration and invasion) in vitro (Supplementary Fig. S6c–e), possibly due to selective pressure. These results demonstrate that PLXNB2 is necessary for tumor cell clustering but acts independently of the phenotypes on proliferation or cell motility.

PLXNB2 is a large single-pass transmembrane protein made up of multiple domains⁵³ (Fig. 3h): the extracellular region (ECTO) including a semaphorin domain (Sema) responsible for binding to receptor/ligands, three Ig-like fold domains and three plexin-semaphorin-integrin domains; a transmembrane domain; the intracellular Rho-Binding Domain (RBD); the intracellular GTPase activating protein domain, and the intracellular PDZ-binding domain (VTDL). Using the PB2 mutant constructs shared by the Friedel Lab⁵⁴, we overexpressed the full-length PLXNB2 and its various mutant forms (dECTO, mRBD, and dVTDL) back into PB2 KO cells and validated the expression via immunoblotting (Fig. 3i) or flow cytometry (Supplementary Fig. S6f). While full-length PLXNB2 overexpression effectively rescued the clustering of PB2 KO cells, the PB2 mutant depleting the ECTO domain (dECTO) had the lowest cluster-rescuing function, followed by the mutants depleting or modifying the intracellular signaling domains (dVTDL and mRBD) (Fig. 3j, k). This data demonstrates that both extracellular and intracellular signaling properties in full-length PLXNB2 are required to promote tumor cell clustering.

As self-renewal is often required for CTCs to regenerate tumors^13,55,56,57,we next determined if PLXNB2 regulates self renewal-related tumorigenesis in vivo and mammosphere formation of breast cancer cells in vitro¹⁶. We found that of PB2 depletion via KO or KD significantly compromised the orthotopic tumorigenesis at serial dilutions (down to 10 cells) with reduced frequencies of tumor-initiating cells (TICs) and reduced the number of mammospheres formed in TNBC cells (Supplementary Fig. S4i, j, S6g, h). The mammosphere phenotype in PB2 KO cells was restored by overexpression of the full-length PLXNB2 and to a lesser extent by overexpressed dECTO PB2 mutant, suggesting that the extracellular domain of PLXNB2 is necessary for its functions in promoting self-renewal of tumor cells (Supplementary Fig. S6i-j).

Given that tumor cell clusters are powerful precursors to metastatic colonization, we wanted to determine if PLXNB2 levels would impact tumor cell seeding to the lungs. Within 4 days after tail vein injection for experimental colonization, the mice injected with L2G-labeled¹³ PB2 KD cells (MDA-MB-231) either by siPB2 or si-PB2−11 had shown 80-90% less disseminated colonization to the lungs compared to the control group (Fig. 3l-o). These data suggest that PLXNB2 promotes tumor clustering, self-renewal, and experimental lung colonization, contributing to metastasis.

PLXNB2 signals through SEMA4C to promote tumor cell clustering

To elucidate the molecular mechanism by which PLXNB2 facilitates tumor cluster formation, we investigated whether in breast cancer PLXNB2 interacts with any of its canonical ligands on the cell surface, such as single-pass transmembrane proteins, semaphorin (Sema) family members SEMA4A, 4C, 4D, and 4G^{51,54,58,59,60,61,62}. From our ranked list of adhesion proteins, we found expression of SEMA4C, SEMA4D, and SEMA4A in treatment-naïve breast cancer³² (Supplementary Fig. S6k); however, only SEMA4C was detected in breast cancer cells (human/mouse cell lines and/or PDX tumor cells) via immunoblot analyses and mass spectrometry (Fig. 4a, b, and Supplementary Data 1). Based on publicly available databases, we also found that high SEMA4C mRNA expression in breast cancer correlates with unfavorable DMFS and OS (Supplementary Fig. S6l-m). We hypothesized that SEMA4C is a primary ligand of PLXNB2 in driving homotypic tumor cell clusters in metastasis.

**Fig. 4: PLXNB2 interacts with SEMA4C to promote homotypic clustering of breast tumor cells.**

To confirm if Sema molecules interact with PLXNB2 in breast cancer cells, we performed a co-immunoprecipitation for human PLXNB2 using clustered TNBC cells; and found that SEMA4C was specifically enriched in the PLXNB2 protein complex, whereas SEMA4D was nearly undetectable (Fig. 4c), demonstrating SEMA4C-PLXNB2 interactions in the homotypic CTC clusters. In the anti-PLXNB2 pull-down protein complex, we also detected CDC42 (Fig. 4c), which is a Rho GTPase family member known to be regulated by PLXNB2 signaling^63,64.

To determine the functional importance of SEMA4C in tumor clustering, we knocked down SEMA4C in the breast cancer cells (Fig. 4d). SEMA4C KD alone mimicked PB2 KO in reducing the average size and numbers of WT tumor cell clusters, whereas loss of both PLXNB2 and SEMA4C together did not have an additional impact on cluster size compared to PB2 KO alone (Fig. 4e–g). This indicates that these two interacting proteins might belong to the same interaction pathway necessary for homotypic tumor clustering.

To gain a better understanding of the downstream molecules and pathways in PLXNB2-mediated tumor cell clustering, we performed an MS-based global proteomics analysis of MDA-MB-231 breast cancer cells in suspension before and after clustering. Principal component analysis indicates three groups of samples and Heatmap comparisons show 6 distinct groups of protein signatures, up-regulated or down-regulated within 4 h of clustering or by siPB2-mediated KD (Fig. 4h-i, Supplementary Data 3). Additional GO analysis of the altered proteins in siPB2 cells includes biological processes in chromosome condensation, cell cycle, RNA export, DNA methylation, and others (Fig. 4j). Groups C and F, with clustering-specific up-regulation and down-regulation in protein expression, respectively, identified pathways in protein folding and hydroxylation, UV protection, and response to gamma radiation, all of which were over 10-fold enriched in PLXNB2⁺ tumor cell clusters (Supplementary Fig. S7a and Supplementary Table S1).

In search of PLXNB2 downstream target proteins related to TNBC phenotypes, we found that MCM7, an essential component of the DNA helicase, was one of the top candidates significantly decreased in the cells with siPB2-mediated KD (Supplementary Fig. S7b). MCM7 has previously been implicated in cancer stemness and metastasis⁶⁵, thus making it an interesting target for clustering studies. TNBC cells transfected with siMCM7 for gene KD showed a compromised clustering within 6 h (Supplementary Fig. S7c–d). Thus, our data demonstrates a mechanism by which PLXNB2 can promote clustering through influencing MCM7. However, MCM7 overexpression did not rescue the clustering nor the mammosphere formation of TNBC cells with PB2 KD or KO (Supplementary Fig. S7e–h), indicating that MCM7 is necessary but not sufficient to rescue the PLXNB2 signaling in clustering and self-renewal. Moreover, KD of CDC42, a known target of PLXNB2 signaling⁶⁴ and a component of the PLXNB2 protein complex (Fig. 4c), mimicked the KD of PLXNB2 and MCM7 in reducing efficiencies of mammosphere formation and tumor cell clustering in vitro (Supplementary Fig. S7i–k).

PLXNB2-dependent heterotypic clustering signals through SEMA4A on monocytes

In addition to MCM7, the global mass spectrometry analysis of PB2 KD cells revealed many biological processes at 10- to 40-fold enrichment, such as positive regulation of the immune response to tumor cells and astrocyte activation (Supplementary Table S1). Based on our CellSearch data that demonstrated PLXNB2 enrichment in the CTC-WBC heterotypic clusters (Fig. 1f-g), we hypothesized that PLXNB2 in tumor cells impacts interactions with immune cells. Since immune cells, especially myeloid cells, have been found in heterotypic CTC clusters driving cell cycle progression, proliferation, and/or immune evasion^19,66, we proposed to determine what, if any, role PLXNB2 plays in heterotypic CTC-immune cell clusters (Fig. 5a).

**Fig. 5: PLXNB2 binding to SEMA4A on monocytes drives heterotypic clustering.**

In co-culture suspension, we mixed WT or PB2-KO MDA-MB-231 cells with patient WBCs. The PB2 WT cells formed larger and more heterotypic clusters with WBCs than PB2 KO cells without affecting tumor cell viability (Fig. 5b–c, and Supplementary Fig. S4b, S8a), demonstrating that PLXNB2 promotes heterotypic tumor-WBC cluster formation in which tumor cells can possibly evade immune cell killing. To identify which ligand might be involved in CTC-WBC cluster formation, we used a computational ranking of surface proteins to determine which canonical ligands of PLXNB2, i.e., semaphorin interactors⁶⁷, were expressed in WBCs. The online single-cell RNA sequencing data suggested that among canonical PLXNB2 ligands, SEMA4A had the highest expression within peripheral blood derived mononuclear cells (PBMCs) (monocytes), whereas SEMA4C was minimally expressed (https://www.proteinatlas.org) (Supplementary Fig. S8b). Consistently, the published MS proteomic data⁶⁷ also revealed a higher abundance of SEMA4A protein than SEMA4D with undetectable SEMA4C in human blood monocytes (Supplementary Fig. S8b, and Supplementary Data 1). Furthermore, our flow cytometry analyses of the blood cells from patients with breast cancer demonstrated that, compared to granulocytes and lymphocytes, monocytes express a higher level of SEMA4A (Fig. 5d, and Supplementary Fig. S8c), and that the heterotypic CTC-WBC clusters double positive for EpCAM and CD45 have enriched co-expression of PLXNB2 and SEMA4A versus WBC-only clusters and single cells (Fig. 5e, Supplementary Fig. S8d-e), implying the possible contribution of PLXNB2 and SEMA4A to heterotypic CTC-WBC clustering.

To determine the protein interactions between monocyte SEMA4A and tumor cell PLXNB2, we utilized the THP1 monocytic cells which express high SEMA4A ( > 80%) (Supplementary Fig. S8f) as an alternative source of human monocytes for heterotypic cluster formation. Similar to patient WBCs, they do not impact the viability of PB2 KO tumor cells in co-culture (Supplementary Fig. S8g). Using human tumor cell-THP1 clusters at an optimized 1:4 ratio and anti-PLXNB2 antibody for co-immunoprecipitation, we also detected the enrichment of SEMA4A and CDC42 in the PLXNB2 protein complex, compared to monocytes only (Fig. 5f, g).

To determine the importance of PLXNB2 in tumor-monocyte clustering, we found that in mixed cell suspension, PLXNB2 control tumor cells formed clusters effectively with THP1 cells (1:4 ratio), whereas PB2 KO TNBC cells lost the capability for heterotypic cluster formation (Fig. 4h-i). Consistently, sorted primary PDX TNBC cells (TN1) with PLXNB2^high expression formed significantly larger heterotypic clusters with THP1 cells than PLXNB2^low TN1 cells, when co-cultured under the previously established clustering conditions on collagen-coated plates¹⁶ (Fig. 4j-k, Supplementary Fig. S8h).

To further determine the association of SEMA4A with heterotypic tumor-monocyte clustering, we sorted SEMA4A⁺ and SEMA4A⁻ THP1 cells and genetically knocked down SEMA4A in THP1 cells for co-culture with PB2 WT and KO TNBC cells. SEMA4A⁺ monocytes with PLXNB2⁺ tumor cells (double positive) formed the largest clusters with more cell counts per cluster and in highest numbers of clusters compared to single-negative or double-negative combinations of SEMA4A^−/+ monocytes with PLXNB2^+/− tumor cells in heterotypic clustering (Fig. 5l, m, Supplementary Fig. S8i, and S9a–b). Interestingly, loss of SEMA4A alone in monocytes was sufficient to reduce the size of heterotypic PLXNB2⁺ tumor clusters which is comparable with the effects of PB2 KO in tumor cells, suggesting that SEMA4A is the primary ligand on monocytes for PLXNB2-dependent tumor-monocyte clustering (Fig. 5a).

We then compared the outcomes of experimental colonization with TNBC (MDA-MB-231) control cells and PB2 KO cells which were pre-clustered for 4 h ex vivo prior to tail vein-injections. With minimal clustering capacity, the PB2 KO cells showed a significant reduction in tumor cell dissemination to the lungs after 12 h, and the phenotype was maintained for up to four days (Fig. 5n), suggesting that PLXNB2 enhances dissemination and metastatic colonization, coupled with tumor clustering and independent of proliferation effects. Similarly, after 4 h pre-clustering with unlabeled THP1 monocytes, heterotypic L2G⁺ tumor cell-THP1 clusters promoted tumor cell dissemination (12 h), in a PLXNB2-dependent manner (Fig. 5n, o). Meanwhile, the presence of THP1 monocytes promoted the metastatic colonization of both WT and PB2 KO tumor cells within 4 days (Fig. 5n, o), possibly through both PLXNB2-dependent clustering and clustering-independent factors, such as stemness⁶⁸. These results demonstrated that monocytes promote heterotypic CTC clustering, early dissemination, and lung colonization.

Loss of Plxnb2 from mouse tumor cells reduces metastasis and CTC-WBC formation in vivo

Finally, to determine the role of mouse Plxnb2 in CTC-immune cell cluster formation, we knocked down Plxnb2 gene expression in mouse 4T1 TNBC cells with siPlxnb2 transfections and found that loss of Plexin-B2 in tumor cells diminished the heterotypic 4T1-mouse WBC cluster formation (Fig. 6a–b). 6 h after the L2T⁺ tumor cells were inoculated into the tail vein of immune competent Balb/c mice, the blood CTC analyses also revealed that Plxnb2 KD reduced heterotypic CTC clustering with monocytes (Fig. 6c), which possibly provided opportunities for other WBCs (such as T cells) to interact with 4T1 tumor cells (Supplementary Fig. S9c-e).

**Fig. 6: Mouse *Plxnb2* depletion inhibits CTC clustering and spontaneous metastasis of 4T1 tumors in vivo.**

To assess the effects of Plexin-B2 depletion on spontaneous metastasis in immune competent mice, we further generated Plxnb2-KO 4T1 cells using CRISPR/Cas9 technologies and orthotopically implanted the WT control cells (Control) and KO cells (1.5 × 10⁶ cells) into the 4th mammary fat pads of Balb/c mice (Fig. 6d). By Day 9 when two groups of tumors developed into comparable sizes (tumor weight), the Plxnb2-KO tumor group significantly reduced spontaneous metastases to the lungs, coupled with decreased CTCs, both singles and CTC-WBC clusters, particularly the CTC-myeloid clusters (Fig. 6e–i), as analyzed by bioluminescence imaging and flow cytometry. To further confirm the lung metastasis measured via IVIS bioluminescence/fluorescence imaging (Fig. 6e), we performed histology analyses of the mouse lungs, which cross-validated that Plxnb2 depletion via CRISPR KO in 4T1 cells reduces the spontaneous micrometastic lesions of 4T1 tumors spread to the mouse lungs (Supplementary Fig S9f–g).

Taken together, our data demonstrates that mouse Plxnb2 and human PLXNB2 play a similar role in promoting CTC-monocyte cluster formation and driving spontaneous metastases in breast cancer.

Discussion

Our studies have established an unprecedented computational ranking of quantitative global MS-based relative protein abundance and tumor-specific expression and integrated it with clinical outcome analyses and experimental function determination. PLXNB2 is identified as a driver that promotes both homotypic and heterotypic CTC clustering through its interactions with Semaphorin ligands SEMA4C on tumor cells and SEMA4A on monocytes, in addition to PLXNB2 functions in proliferation and other known functions^{43,44,45,46,47,48,49,50}. It has been reported by Gurrapu et al. that PLXNB2 promotes breast cancer cell proliferation through the RhoA and MAPK signaling pathways⁵¹, which appears to be context-dependent and influenced by the expression levels of co-receptors such as MET and ErbB2. Furthermore, our studies reveal proteins such as MCM7 as downstream mediators of the PLXNB2 interaction network that are required for PLXNB2-mediated tumor cluster formation but might not be sufficient to rescue surface protein-mediated intercellular crosstalk. Future therapeutics aimed at blocking PLXNB2 and SEMA4C/4A interactions can potentially serve as targeting strategies in breast cancer, especially TNBC and complement existing treatments, such as pembrolizumab^69,70,71, Sacituzumab^72,73, and Palbociclib⁷⁴ for improved outcomes.

CTC clusters may contain multiple cell types. Homotypic tumor cell clusters are often enriched with tumor cells that have stem cell properties^16,26,37,75, whereas heterotypic clusters contain mixtures of tumor and immune cells in which the tumor cells may evade immune cell attack and gain proliferation advantages^19,76. Our previous work and that of others have identified a few molecular drivers of homotypic tumor cell clusters, including tumor-initiating cell markers and receptors^16,26,37 and molecules in cell junctions or adhesive structures¹¹. The observation that heterotypic CTC-monocyte clusters enhance metastatic colonization highlights a potential pro-metastatic role for monocytes in facilitating tumor cell dissemination and colonization. Similarly, neutrophils within heterotypic CTC clusters promote cell cycle progression and metastatic potential of CTCs in breast cancer¹⁹.

CTC clusters have up to 50-fold greater metastatic potential compared to single CTCs, making them a very crucial population to target. While it has long been hypothesized that CTCs can also cluster with immune cells, limited research has thus far been devoted to identifying the molecular markers of these heterotypic clusters. In addition to myeloid cells such as the monocytes specified in this work, neutrophils, MDSCs^19,77, and the plethora of other immune cell types need to be further investigated for potential interactions with and influences on CTCs in cancer metastasis.

We demonstrate a strong correlation between PLXNB2-mediated CTC cluster formation (homotypic and heterotypic) and lung metastasis. While these findings suggest that PLXNB2 plays a crucial role in promoting metastasis through CTC clustering, we acknowledge that metastasis is a highly complex and multifaceted process. Prior studies have shown that polyclonal metastases can also arise through sequential seeding events⁵². Metastasis is a series of complex steps by which tumor cells shed off the primary tumor, enter the vasculature (intravasate), circulate, and ultimately extravasate into secondary tissues^14,78,79. Previous studies demonstrate that myeloid cells such as macrophages promote migrating tumor cell contact with the vasculature and enhance vascular permeability for intravasation^80,81. Considering the tumor-myeloid interactions promoted by PLXNB2, PLXNB2 may also promote intravasation of tumor cells in addition to CTC dissemination.

From the mass spectrometry analyses, the unexpected identification of the tumor cell PLXNB2 signaling pathway in connection with astrocyte activation and immune response might imply its role in regulating the immune microenvironment of central nervous system (CNS) and other organs. PLXNB2 is known to regulate neuronal migration and synaptogenesis^46,82 and has previously been implicated in regulating cellular biomechanics, stemness, host microenvironment, and angiogenesis in brain cancer and other cancers^{43,51,54,62,83,84}. The newly identified role of PLXNB2 in CTC clustering in conjunction with its ligands SEMA4C in tumor cells and SEMA4A in monocytes promotes metastasis, and is correlated with lower OS and DMFS in the context of advanced breast cancer, TNBC in particular. It is likely that certain cancer types or host immunity^85,86 may confound the effects of PLXNB2⁸⁷. The context-dependent functions and biomarker association of PLXNB2 may be largely dependent on the ligands it interacts with and the microenvironmental conditions^36,88. For instance, other known ligands of PLXNB2, such as SEMA5A and SEMA4D, are either not expressed or not detectable in breast tumors at the primary site. However, the interaction of SEMA4D in disseminating breast cancer cells with PLXNB1 of brain endothelial cells can contribute to the brain-tropic metastasis of cancer⁸⁹. One of the future studies is to examine the association of PLXNB2 with CNS diseases and its contribution to regulating metastasis of breast cancer to the brain and other organs.

A recent study by Borrelli et al. has identified PLXNB2 as a crucial host-derived regulator of liver colonization in colorectal and pancreatic cancer through its interactions with class IV semaphorins on tumor cells, leading to the upregulation of KLF4 and promoting the acquisition of epithelial traits^86,86. These findings suggest that PLXNB2-semaphorin signaling plays a pivotal role in tumor metastasis across different cancer types, making it a promising therapeutic target for further investigation. Furthermore, SEMA4D interacts with PLXNB2 to mediate monocyte adhesion to endothelial cells⁹⁰. This interaction is significant because it suggests that PLXNB2 not only plays a role in tumor cell behavior but also influences immune cell recruitment and adhesion, which can impact the tumor microenvironment and metastatic niche. In cancer, this could mean that PLXNB2-semaphorin signaling may contribute to forming a more adhesive and supportive environment for both cancer cells and immune cells, facilitating metastasis. Understanding the dual role of PLXNB2 in both tumor cell adhesion and immune cell dynamics could offer new therapeutic strategies aimed at disrupting these processes to reduce metastasis and improve patient outcomes.

Ongoing studies remain necessary to overcome the limitations of our study. First, the RScore approach for ranking candidate proteins is a semi-quantitative, heuristic-based method that integrates multiple modalities of biological datasets and clinical relevance. It is designed with a modifiable constant factor (weight) for each data input to provide room and flexibility for continuous training and machine learning (ML)-based determination in following studies. It will be optimized through integration with expanded dataset inputs, both known and unknown, and experimental validation of candidate rankings to enhance predictive power. Second, the use of fluorescence-activated cell sorting (FACS) for analyses of CTCs and CTC clusters requires cross validation via other methods, including but not limited to multiplexing fluorescence imaging and/or single-cell sequencing of enriched CTCs from fixed or unfixed blood cells ex vivo (i.e., CellSearch, Parsortix, ImageStream, CellView, etc) as well as intra-vascular CTCs in tissue sections in situ (no blood processing).

Methods

Patient sample collections

Blood samples and breast tissue sections (tumors and normal adjacent) were collected from stage III-IV breast cancer patients under guidelines from the Institutional Review Board and Ethics committee at Northwestern University (IRB protocols STU00203283 and STU00214936) in compliance with NIH human subject studies guidelines. Blood samples were collected into CellSave tubes for CellSearch analyses or into EDTA tubes for flow cytometry of live cells. The CellSearch® platform for CTC detection (CD45⁻DAPI⁺Cytokeratin⁺) had one open immunofluorescence channel for Plexin B2 analysis. Prior to flow cytometry analysis, blood specimens in EDTA tubes underwent red blood cell lysis (Sigma #R7757). Breast tumor tissues and normal adjacent tissues were frozen in tissue freezing buffer OCT until sectioned for laser capture microdissection (LCM) for mass spectrometry analysis.

Animal studies

All animal procedures complied with the NIH Guidelines for the Care and Use of Laboratory Animals and were approved by the Northwestern University Institutional Animal Care and Use Committee (IACUC protocol IS00014098). The maximal tumor size/burden (2 cm in diameter) was not exceeded in any case. All mice used in this study were kept in specific pathogen-free facilities in the Animal Resources Center at Northwestern University.

CellSearch

CellSearch analysis processed 7.5 mL of patient blood using the CTC Epithelial Kit (CellSearch #7900001) and CXC Kit (CellSearch #7900017) to deplete immune cells by an EpCAM⁺ selection and identify specific markers. Cells were then stained for CK, DAPI, and CD45 (3 mL for each, Menarini # 7900001) and Plexin B2 (1:1000, Miltenyl Biotec #130-126-566).

Flow cytometry analysis

Mouse/human cells, PDXs, and patient cells were counted and resuspended in wash buffer (PBS + 2% FBS). They were blocked with mouse IgG (Sigma #I5381) for 10 minutes on ice and incubated with fluorescence-conjugated antibodies for 15 minutes on ice: Plexin B2 PE (phycoerythrin) (1:1,000, human, R&D Systems #FAB53291P), Plexin B2 APC (1:1,000, human, R&D Systems #FAB53291A), Plexin B2 FITC (1:1,000, mouse, R&D Systems #FAB6836G), CD45 (1:1,000, human, BD Bioscience #557748), EpCAM FITC (1:1,000, human, BD Bioscience #347197), SEMA4A PE (1:1,000, human/mouse, BioLegend #148404), SEMA4A APC (1:1,000, human/mouse Biolegend #148406). Cells were washed 2x in wash buffer and run for analysis on a fluorescence-associated cell sorting (FACS) LSR cytometer from BD Biosciences.

Cell sorting

Human/PDX/PBMC cells were resuspended in PBS + 2% FBS at a final concentration of 10 × 10⁶ cells/mL in 5 mL FACS tubes after filtering (Fisher Scientific #352235). Cells were blocked with mouse IgG (Sigma #I5381) for 10 minutes on ice and florescent antibody for 15 minutes on ice: Plexin B2 PE (1:1000, human, R&D Systems #FAB53291P), Plexin B2 APC (1:1000, human, R&D Systems #FAB53291A), Plexin B2 FITC (1:1000, mouse, R&D Systems #FAB6836G), SEMA4A PE (1:1000, human/mouse, BioLegend #148404). Cells were then run through BD FACSMelody Cell Sorter and cells were collected based on gated populations. Collected cells were washed 2x in PBS prior to downstream application.

Tissue and cell preparation for mass spectrometry analysis

Human breast cancer tissues were frozen with OCT and stored at −80 °C until being cryo-sectioned to a series of 10-µm thick sections using a blade temperature −35 °C and specimen temperature of −25 °C. The sections were then thawed and mounted onto PEN membrane slides for laser capture microdissection (LCM) and onto regular glass slides for H&E staining or immune-histochemistry staining. For LCM, the OCT was removed by immersing slides into 50% ethanol for 2 min, dehydration in100% ethanol (2 min) and xylenes (1 min), and then the sections dried out in a vacuum desiccator. After identified via H&E staining of adjacent sections, and the regions of tumor and non-tumor/normal adjacent were correspondingly identified in the unstained tissue sections for LCM on a PALM MicroBeam system (Carl Zeiss MicroImaging, Munich, Germany), at an energy level of 63 and with an iteration cycle of 1. Voxels were collected within a wet buffer droplet for the best protein recovery.

MDA-MB-231 breast cancer tumor cells underwent double transfection of siPB2 for KD of PB2 (Dharmacon #L-031513-01) and control siRNAs (Dharmacon #D-001810-10-50). Cells were then trypsinized and resuspended on PolyHema-coated plates and allowed to cluster for 4 h. Adherent cells were scraped from the plate at the zero hour time point. After clustering, cells were harvested and lysed using RIPA buffer (VWR Amresco #N653-100mL) supplemented 1:100 with protease inhibitor (Thermo Fisher #78440).

Protein digestion by S-Trap for proteomic analysis

Tissue voxels were collected in digested 10 µL of a cocktail of buffer containing 0.1% DDM in 50 mM TEABC, 20 ng of Lys-C, and 80 ng of trypsin. MDA-MB-231 cell lysates were resuspended in SDS buffer (the final concentration was 5%). The protein concentration was measured with a BCA protein assay (Thermo Fisher Scientific). A total of 50 mg of protein was reduced with 10 mM DTT for 15 minutes at 37 °C and subsequently alkylated with 50 mM iodoacetamide at 25 °C for 15 minutes in the dark. The samples were acidified by phosphoric acid (the final concentration was 2.5%) and then diluted with six volumes of “binding” buffer (90% methanol; 100 mM triethylammonium bicarbonate, TEAB; pH 7.1). After mixing, the protein solution was loaded onto an S-Trap filter from Protifi (Huntington, NY), spun at 10,000 g for 1 minute, and then the filter was washed with 150 μL of binding buffer three times. Proteins were digested with Lys-C (Wako) and sequencing-grade trypsin (Promega, V5117) (1 μg of each in 20 μL of 50 mM TEAB) in the filter at 37 °C for 16 h. To elute peptides, 40 μL of 50 mM TEAB, 40 μL of 0.2% formic acid (FA) in H₂O, and 40 μL of 50% acetonitrile (CAN) in H₂O were added sequentially. The peptide solutions were pooled for BCA assay to estimate peptide amounts. Totals of 20 μg of peptides were dried with a SpeedVac and stored at −80 °C until LC-MS/MS analysis.

LC-MS/MS analysis of tumor tissue voxels, patient CTCs, CTC-derived PDXs, and MDA-MB-231 cells: Lyophilized peptides were reconstituted in 200 μL of 0.1% TFA with 2% ACN containing 0.01% n-Dodecyl $\beta$-D-maltoside (DDM)⁹¹ to reach a concentration of 0.1 μg/μL, and 5 μL of the resulting sample was analyzed by LC-MS/MS using an Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Scientific) connected to a nanoACQUITY UPLC system (Waters) (buffer A: 0.1% FA with 3% ACN and buffer B: 0.1% FA in 90% ACN) as previously described⁹². Peptides were separated by a gradient mixture with an analytical column (75 μm i.d. × 20 cm) packed using 1.9-μm ReproSil C18 and with a column heater set at 50 °C using an LC gradient of 2–6% buffer B in 1 min, 6-30% buffer B in 84 min, 30-60% buffer B in 9 min, 60-90% buffer B in 1 min, and finally 90% buffer B for 5 min at 200 nL/min. The DIA-MS/MS scan was performed in the HCD mode with the following parameters: precursor ions from 350–1650 m/z were scanned at 120,000 resolution with an ion injection time of 60 ms and an AGC target of 1E6. The range of m/z (isolation window) of data-independent acquisition (DIA) windows from 377 (54), 419 (32), 448 (28), 473.5 (25), 497.5 (25), 520.5 (23), 542.5 (23), 564.5 (23), 587 (24), 610.5 (25), 635 (26), 660 (26), 685.5 (27), 712.5 (29), 741 (30), 771 (32), 803.5 (35), 838.5 (37), 877 (42), 921 (48), 972 (52), 1034.5 (71), 1133.5 (129), and 1423.5 (453) was scanned at 30,000 resolution with an ion injection time of 120 ms and an AGC target of 3E6. The isolated ions were fragmented with HCD at the 30% level.

Proteomic data analysis

The raw DIA data were processed with Fragpipe^93,94 and searched against a human UniProt database (FASTA file dated Jan. 05, 2022 with 40,818 sequences, which contained 20,409 decoys). Initial fragment mass tolerances were set to 20 ppm. A peptide search was performed with strict tryptic digestion (Trypsin) and allowed a maximum of two missed cleavages. Carbamidomethyl (C) was set as a fixed modification; acetylation (protein N-term) and oxidation (M) were set as variable modifications. DIA_SpecLib_Quant workflow was used for DIA quantitation (takes DIA data as input, builds a spectral library using MSFragger-DIA, then quantifies with DIA-NN).

Integrated ranking of proteins based on their relative abundance, tumor specificity, and significance in association with patient outcomes: The presence of proteins in chosen datasets was first determined. The integrated ranking of each protein is formulated as Rscore = $\sum {n}_{{{{\rm{i}}}}=1}({r}_{{{{\rm{i}}}}}{c}_{{{{\rm{i}}}}})/n$ which sums each protein’s individual ranks (r_i) multiplied with a constant factor (c_i) as its weight factor and is divided by n (number of integrated ranks). The individual ranks in this study include (but not limited to):

$$\begin{array}{cc}{R}_{{score}}=\frac{{r}_{1}{c}_{1}+{r}_{2}{c}_{2}+{r}_{3}{c}_{3}\ldots+{r}_{n}{c}_{n}}{n} & {c}_{i}={constantfactor}({weight}\; {of}\; {each}\; {rank})\end{array}$$

(1)

$${r}_{i}(p{{{\rm{a}}}})={rank}\left(\frac{{protein}\; {abundance}/{intensity}\; {in}\; {MS}\; {spectral}\; {counting}}{{{{\rm{candidate}}}}\; {{{\rm{protein}}}}\;{length}\left({{{\rm{a}}}}.{{{\rm{a}}}}.\right)/{median}\; {pr}.{length}\left(a.a.\right)}\right)$$

(2)

$${r}_{i}\left({sc}\right)={rank}({significant}\; {changes}:p,{ratio}\; {or}\; {fold}\; {change},{absolute}\; {change})$$

(3)

$${r}_{i}\left({ca}\right)={rank}({clinical}\; {association}:p,{hazard}\; {ratio})$$

(4)

(1) r_i (pa), ranks in relative protein abundance in multiple mass spectrometry (MS) datasets (breast tumors, cancer cell lines, and CTCs), calculated as protein intensity (total number of MS/MS spectra for a given protein) divided by the ratio of the tested protein length (# of amino acids) versus the median protein length (# of amino acids). The ranks of detected proteins (x) in a dataset ranged from the highest abundance (r_{i =}1) to the lowest (r_{i =}x).

(2) r_i (sc), ranks in significant changes (p values, ratio or fold change, and absolute change), such as tumor-specific expression of a chosen protein in TNBC voxels via laser capture microdissection vs normal adjacent tissues. The ranking scored from the lowest to highest in p values, from the highest to lowest in fold changes (ratios), and from the highest to lowest in absolute differences.

(3) r_i (ca), ranks in clinical association (p values, hazard ratios) of each protein with overall survival (OS) and distant metastasis-free survival (DMFS) among multiple datasets. The log rank P values are calculated based on the most significant cut-off values on multiple Cox regression tests to separate two groups of high and low expression as cited³⁵ (PMID: 34309564) unless specified. The ranking for oncogenic proteins scored from the lowest to highest in log rank p values (negative association) and from the highest to lowest in the hazard ratios.

The ranks of detected proteins (x) in a dataset ranged from the highest rank (r _i = 1) and to the lowest (r_i = x). The constant factor (c_i) of each individual rank reflects its contribution weight and can be determined by computational estimation and experimental analyses. The top R scores are of the smallest sum values. Lab-obtained proteomic datasets as well as published datasets were utilized for integrated analyses of adhesion proteins in the study.

Adhesion proteins (N = 608, derived from the Molecular Signature Database^27,28,29 of the Gene Ontology Biological Processes^30,31) were ranked according to tumor specific expression in the global MS proteomic datasets of laser capture microdissection voxels of human TNBC regions versus normal adjacent tissues (p values, ratio or fold change, absolute differences), relative protein abundances in 122 treatment-naive primary breast tumors of patient samples³², patient CTCs, patient-derived xenograft (PDXs), and TNBC cells such as MDA-MB-231³³ and Hs578T cells³⁴. Relative protein abundance in other datasets such as THP1 macrophage cells⁹⁵ and expression in extracellular vesicles (EVs) was also ranked. Patient outcome association analyses were conducted via KMplotter data in individual datasets (protein and mRNA expression) and ProteinAtlas data (mRNA expression). The p values and hazard ratios of each protein in negative or positive OS and DMFS were ranked (see Supplementary Data 1).

Extracellular vesicle isolation

EVs were isolated from human cell lines and PDX models from culture media after 72 h in culture. Media were then pooled and isolated using differential centrifugation. The first centrifugation step was at 2000 x g for 10 minutes. Media were then transferred into an appropriate ultracentrifuge tube (Beckman Coulter # 344058) and spun at 10,000 x g for 30 minutes. Supernatant was removed and transferred into a clean ultracentrifuge tube and spun at 100,000 x g for 70 minutes. The resulting EV pellet was washed with PBS and spun again at 100,000 x g for 70 minutes. The final EV pellet was resuspended in PBS and stored at −80 °C until analysis.

Cell culture

All cell lines used in this study were obtained from ATCC and periodically tested to be mycoplasma-negative using Lonza’s MycoAlert Mycoplasma Detection Kit (cat #LT07-218). Cell lines were maintained in complete supplemented media (10% FBS, 1% penicillin-streptomycin (Sigma-Aldrich P4333-100ML)). MDA-MB-231 and 4T1 cell lines were maintained in DMEM supplemented with high glucose (Corning #10-013-CV) for <20 passages in cell culture incubators at 37 °C, 5% CO₂. MDA-MB-468, HS578T, and THP1 cells were cultured in RPMI-1640 supplemented media (Fisher Scientific #SH30027.01) for <20 passages in cell culture incubators at 37 °C, 5% CO₂. No commonly misidentified cell lines were used in the study.

For tumor cell clustering analyses, we used multiple human breast cancer cell models, including HER2 + SKBR3 and TNBC MDA-MB-231 cells. MDA-MB-231 cells not only form tumor cell-tumor cell clusters, but also adhere with human WBCs and monocytes (THP1 cells) to form heterotypic clusters. In addition, mouse breast cancer line 4T1 has shown similar phenotypes as MDA-MB-231 cells. Therefore, we focused on using MDA-MB-231 and 4T1 for functional studies in vivo. For tail vein injection and spontaneous metastasis studies, we used MDA-MB-231 cells (WT or PLXNB2 KO) in female NSG mice (6–8 weeks old). For mouse Plxnb2 KO effect in lung metastasis, we used 4T1 cells (WT or Plxnb2 KO) in female Balb/c mice (6–8 weeks old).

Gene knockdown

A total of 2 × 10⁶ cells were plated in a 10 cm dishes one day prior to KD (day 0). On the day of KD, ON-TARGETplus siRNA at a final concentration of 50 nM per plate was incubated with Dharmafect 1 at 100 nmol/L (GE Dharmacon #T-2001-03) for 20 minutes at room temperature in reduced serum Opti-MEM Medium (Thermo Fisher Scientific #31985070). siRNA+Dharmafect mixture was added to cells in Opti-MEM for 16-24 h. Cells were passaged in complete media and reseeded at 2 × 10⁶ cells in a 10 cm dish, and the KD procedure was repeated once more as described. On day 4, cells were harvested, counted, and analyzed via flow cytometry and western blotting to check for sufficient KD of target protein. The following siRNAs were used: SMARTpool Human PLXNB2 (Dharmacon #L-031513-01), SMARTpool Mouse Plxnb2 (Dharmacon #L-040980-00-0010), SMARTpool Human SEMA4C (Dharmacon #L-015364-01-0010), ON-TARGETplus PLXNB2 Human siRNA-09 (Dharmacon #J-031513-09-0010), ON-TARGETplus PLXNB2 Human Individual siRNA-10 (Dharmacon #J-031513-10-0010), ON-TARGETplus PLXNB2 Human Individual siRNA-11 (Dharmacon #J-031513-11-0005), ON-TARGETplus PLXNB2 Human Individual siRNA-12 (Dharmacon #J-031513-12-0005), ON-TARGETplus Non-targeting Pool (Dharmacon #D-001810-10-50).

Gene overexpression

Cells were transfected with PLXNB2 full-length and mutant overexpression plasmids via a Lipofectamine 3000 Transfection Kit (Thermo Fisher #L3000015). Cells were plated at 300k cells/well in a 6-well plate the day prior to transfection. After transfection, cells were incubated for 48-72 h under standard growth conditions and then harvested to check expression of target protein via flow cytometry and western blotting. The following overexpression vectors were used: pLV-PLXNB2-mRBD (Addgene #86240), pLV-PLXNB2-dVTDL (Addgene 86239), pLV-PLXNB2-dECTO (Addgene #86238), PLXNB2 OHu01778C_pcDNA3.1( + ) N-Terminal Flag-Tag (GenScript #SC1626), PLXNB2 OHu01778D_pcDNA3.1 + /C- C-Terminal Flag-Tag (GeneScript #OHu01778D), PLXNB2 Untagged Construct (GeneScript #SC1625).

Gene knockout

Two individual pre-designed human PLXNB2 sgRNA CRISPR-Sanger clones and one non-targeting control vector (Sigma-Aldrich #0020) were ordered from Sigma as glycerol stock (Sigma-Aldrich #HS5000013567, #HS500013568). Bacteria were expanded for maxi prep according to kit protocols (Qiagen #12163). Plasmid was isolated and used to create lentivirus. Cells were infected with either of the sgPLXNB2 clones or the control sgRNA with Cas9-GFP virus (Sigma-Aldrich #0030) at 10 IFU/cell in Opti-MEM (Thermo Fisher Scientific #31985070) with 8 µg/mL supplemented Polybrene (Millipore Sigma #TR-1003-G). Cells were incubated for 4 h at 37 °C in 5% CO₂ upon which complete medium was added to the culture. Cells were incubated in normal growth conditions for 48-72 h and monitored for GFP expression. Cells were harvested and analyzed for sufficient KO of protein. KO cells were sorted for multiple pooled clones using the BD FACSMelody Automated Cell Sorter based on PLXNB2 expression.

For mouse Plxnb2 KO generation, we transfected two individual TrueGuide Synthetic gRNAs (CRISPR66859_SGM and CRISPR66855_SGM, Thermo Fisher) into Cas9-expressing 4T1 cells. The gRNAs were resuspended in 100 µM stock in TE buffer. One day before transfection, cells were seeded in a 6-well plate at 30% confluency. For each well, 37.5 pmol of gRNA was added to 125 µL Opti-MEM I Medium, and then 7.5 µL of Lipofectamine CRISPRMAX Cas9 Transfection Reagent was added to the Medium. The gRNA and Lipofectamine tubes were incubated for 5 minutes at RT, and then combined and incubated for 10 minutes at RT. Transfection complex was added to the wells, and cells were incubated at 37 °C for 2 days, and the Plxnb2 KO was validated by flow cytometry.

Clustering assay

In homotypic tumor cell clustering assays, 20,000 tumor cells were plated in Poly (2-hydroxyethyl methacrylate) (PolyHema)-coated 96-well flat bottom plates (Sigma-Aldrich #P3932-25G). Tumor cell lines were trypsinized from the plate and resuspended by pipetting up and down 10 times and lightly vortexing the samples for 5 seconds immediately before plating. This procedure was optimized by cell imaging under a microscope and/or in the Incucyte at Time 0 h to ensure cells were in a single cell suspension. Tumor cells from PDXs were dissociated from primary tumors (maintained in NSG mice) and seeded on collagen-coated plates for clustering (as terminal experiments) based on established protocols^16,96.

In heterotypic tumor cell-immune cell clustering assays, tumor cells and immune cells were plated at a 1:4 ratio in PolyHema-coated 96-well flat-bottom plates. Cells were monitored in the Incucyte Live Cell Imager (Sartorius) for 24 h and analyzed for average cluster size over time, with a cluster being defined as two or more cells.

Cluster assay analysis was performed using the built-in image analysis tool of IncuCyte software, based on pre-defined masks and cluster objects. The processing definition applied to cluster assays depends on the cell type used to account for individual differences in the cell size. Clusters in breast cancer cell lines, for example, are defined by their surface area from a minimum of 2 cells (object area >300-500 µm²). The pooled analyses of clusters are based on the calculated area of tumor cells and cluster counts. To help visualize the clustering phenotype, a cell cluster of approximately 2-3 cells are in 500 µm², 4-5 cells in 1,000 µm² and 6-12 cells in 1,500-3,000 µm² (PDX models), 2 cells in 750 µm², 4 cells in 1,500 µm², and 10-12 cells in 3,500 µm² (cell line models).

Immunoblotting (western blotting)

Cells were pelleted and then lysed in 1x RIPA Lysis Buffer (VWR Amresco #N653-100mL) supplemented with 1:100 protease inhibitor cocktail (Thermo Fisher #78440) and incubated on ice for 30 minutes. Lysates were then centrifuged for 10 minutes at 14,600 x g at 4 °C. Cell lysates were measured for protein concentration using Bradford analysis (Thermo Fisher #23209, Bio-Rad #500-0006). Totals of 10-30 µg of protein lysate and 10 µL of dual color protein standard ladder (Bio-Rad #161-0374) were loaded onto 4-20% Mini-PROTEAN gels (Bio-Rad #4568094) and then transferred to nitrocellulose membranes using the Bio-Rad TurboTransfer system and kits (Bio-Rad #1704270). Membranes were blocked with 5% milk in Tris-buffered saline and 0.1% Tween 20 (TBST) for 60 minutes and then washed in 0.1% TBST. Primary antibody was incubated on membranes in 5% milk in TBST for 1.5 h at room temperature or at 4 °C overnight. Secondary HRP-conjugated antibodies were added at a dilution of 1:10,000 in 5% milk in TBST and incubated for 60 minutes at room temperature (Anti-Mouse: Promega #W402B; Anti-Rabbit: Promega #W401B). The Pierce SuperSignal West Pico PLUS chemiluminescent substrate (Thermo Scientific #34577) was added one minute prior to imaging using the Bio-Rad Chemidoc imaging system. Primary antibodies used include: Anti-Plexin B2 (1:500, Protein Tech #10602-1-AP), Anti-β-Actin (1:1,000, Abcam #AB8224), Anti-SEMA4A (1:1,000, Thermo Fisher #PA5-101258), Anti-SEMA4C (1:1,000, Ray Biotech #102-11819), Anti-SEMA4G (1:1,000, Santa Cruz Biotech #sc-515644), Anti-SEMA4D/CD100 (1:1,000, Santa Cruz Biotech #sc-39065), Anti-SEMA7C (1:1,000, Santa Cruz Biotech #sc-376149).

Co-immunoprecipitation

The co-immunoprecipitation protocol was done as directed in the Dynabeads Co-Immunoprecipitation Kit (Thermo Scientific #14321D). Plexin B2 (Protein Tech #10602-1-AP) and control rabbit IgG (Protein Tech #3000-0-AP) were pre-conjugated to Dynabeads at 7 µg antibody/mL of beads. Cells were lysed using immunoprecipitation-lysis buffer and incubated overnight at 4 °C with 7.5 mg of pre-conjugated Dynabeads for every 1-15 g protein. Beads were washed, and protein was eluted using SDS loading buffer for downstream applications of western blotting and mass spectrometry analysis.

PDX mouse models

Multiple PDX models of human breast cancer were previously established in the lab^13,16. Cells from TNBC patient tumors or pleural effusion were used to establish tumors that propagated in immunodeficient NSG mice. PDX models were labeled with either Luc2-tdTomato (L2T) (red) or Luc2-eGFP (L2G) (green) reporters to track tumor growth and measure metastasis using bioluminescence imaging and fluorescence analyses (microscopy or flow cytometry). Models were maintained in the lab through tumor harvesting, dissociation, and re-implantation of tumor cells into the mammary fat pads of female NSG mice (6-8 weeks old). PDX models were routinely checked for expression/lack of expression of key markers to monitor phenotype. PDX models were sorted for PB2^high/low cells for downstream in vitro experiments and analysis.

Lung colonization via tail vein injection and bioluminescence imaging

All mice used in this study were female NSG mice 6–12 weeks of age (Jackson Laboratory) and housed in the pathogen-free barrier facility in the Animal Resources Facility at Northwestern University. All animal studies were completed under approved protocols (IS00014098) and adhered to all procedures and regulations outlined by the NIH Guidelines for the Care and Use of Laboratory Animals. Animals were randomized by age and weight and excluded from experiments if presenting conditions unrelated to tumors. Mice were injected with 200,000 L2G or L2T-labeled MDA-MB-231 cells (WT or PLXNB2 KO) via the tail vein, and lung colonization was monitored upon intraperitoneal injection of luciferin (Gold Bio #LUCK-1G 115144-35-9). The bioluminescence signals of metastasis burden were imaged with the IVIS Spectrum Imager, using same imaging time (acquisition times ranged from 5 s to 5 min) across all groups with identical region of interest quantified for comparison. Mice were kept for four days post-injection or until the survival endpoint. At the experimental endpoint, lungs were harvested, imaged in PBS ex vivo for bioluminescence on black-wall 24 well plates, and then preserved in formalin (Fisher Scientific #SF98-4) for H & E staining. The bioluminescence imaging (BLI) signals for regions of interest were measured as total flux (photons/second, p/s). When applicable, the tumor growth signals were normalized and presented as fold change in comparison to that of Day 0. To assess spontaneous lung metastasis, the BLI signals (total flux, p/s) were normalized by tumor weight to minimize the effects of tumor burden.

Spontaneous metastasis in vivo

Female NSG mice at the age of 6–8 weeks were injected with 10,000 L2T/L2G-labeled human breast tumor MDA-MB-231 cells (WT or PLXNB2 KO) into each lower mammary fat pad (L4 and R4) for 6–10 weeks of monitoring. For mouse L2T/eGFP-labeled 4T1 models, female Balb/c mice at age of 6–8 weeks received 1.5 − 2.0 × 10⁶ tumor cells per mammary fat pad (L4 and R4) and monitored for 9 days before harvest. Tumors were monitored at least weekly for growth, and lung metastasis was monitored using bioluminescence imaging. Human tumors grew for 8 weeks and mouse 4T1 grew for 9 days (prior to anti-L2T/eGFP immunity development) or until survival endpoint, at which point tumors, lungs, and blood were collected and analyzed ex vivo. Lungs were imaged under bioluminescence and microscopy and preserved in formalin (Fisher Scientific #SF98-4) for hematoxylin-eosin (H&E) staining and immunohistochemistry (IHC) analysis. Tumors were weighed and preserved in formalin for H&E/IHC analysis. Blood was collected, and red blood cells were lysed using RBC lysis buffer (Sigma #R7757) followed by analysis via flow cytometry for L2G/L2T-positive tumor cells.

Tumorigenesis (Limiting dilution assay)

Serial dilutions of cells were prepared to achieve 1000, 100, and 10 cells per injection in a total volume of 100 µL mixture consisting of a 1:1 mixture of Matrigel and PBS. Female NSG mice (6–8 weeks old) were injected with L2G-labeled human breast tumor MDA-MB-231 cells (WT or PLXNB2 KO) into the mammary fat pads. Each dilution group included three mice, and each mouse received injections of both WT and PLXNB2 KO cells from same dilutions, two injection per flank. Mice were monitored for tumor formation over 6 weeks. Tumor growth was assessed twice weekly by measuring bioluminescence signals using the IVIS Spectrum Imager with consistent imaging acquisition times (5 min per session).The frequency of tumor-initiating cells (TICs) was calculated using extreme limiting dilution analysis (ELDA) software (http://bioinf.wehi.edu.au/software/elda/).

Orthotopic injection and cell cycle analysis of CTCs

L2G- or L2T-labeled MDA-MB-231 cells (WT and PLXNB2 KO) were mixed in equal numbers, with a total of 2 × 10⁵ cells resuspended in a 1:1 mixture of Matrigel and PBS (50 μL each). The cell suspension was injected into both lower mammary fat pads of each mouse. Each group included five female NSG mice (6–8 weeks old). Tumor growth was monitored for 8–10 weeks. At the endpoint, tumors were dissected and weighed, and lung metastases were imaged using bioluminescence with the IVIS Spectrum Imager. Blood samples were processed with red blood cell lysis buffer and stained with Hoechst dye (Thermo Fisher, Cat# 33342) diluted 1:1000 for cell cycle analysis of CTCs. CTCs were identified based on their L2G or L2T fluorescence signals.

Tumor microarray (TMA) and IHC

A total of 89 formalin-fixed paraffin-embedded breast tumor tissues were included in the tumor TMA, with selected tumor regions guided by H&E-stained images. To make a TMA that allows microscopic comparison of the staining characteristics of different blocks while preventing exhaustion of pathological material, a core of paraffin was removed from a “recipient” paraffin block (one without embedded tissue), and the remaining empty space was filled with a core of paraffin-embedded tissue from a “donor” block. An H&E-stained recipient block that is representative of the tissue remaining in the donor block was used to select the sample core with a color marker corresponding to tumor, benign, etc. Matched blocks were pulled out and a recipient TMA block was made and trimmed with the face of the block even with the size of a 1.5 mm core by using the semi-automatic Veridiam Tissue Microarrayer VTA-100. The created TMA block was sectioned for staining. In this TMA, 19 cases from the STU00203283, 9 ER-negative cases, 30 triple-negative cases, 27 ER-positive cases, and 4 normal breast cases (see Supplementary Data 3) were selected and used to construct the recipient block. Detailed patient information can be found in Supplementary Data table. IHC was performed with the help of Bella Shmaltsuyeva of the Robert H. Lurie Comprehensive Cancer Center Pathology Core Facility with Plexin B2 antibody (1:500, Protein Tech #10602-1-AP).

The definition of PlexinB2 high (positive) and low (weak positive and negative) in the TMA analyses followed pathologist recommendations that apply to IHC analyses with both protein staining intensity (positive vs negative) and areas (%): any sample with > 10% positive tumor regions was called positive (high), anything between 1-10% was called low positive and anything <1% was called negative. The relatively low intensity of overall PlexinB2 staining as shown in Supplementary Fig. S2a (low control) was also considered weak positive and included in the PlexinB2 low group.

H&E staining

Mouse lungs and tumors were paraffin-embedded, and sections were mounted on slides and processed with H&E staining with help from the Mouse Histology and Phenotyping Laboratory at Northwestern University.

Mammosphere assay

Tumor cells were plated at 1000 cells per well on PolyHema-coated 12-well plates in Prime-XV Tumorsphere SFM media (Irvine Scientific #91130). Cells were monitored for up to 10 days and the total number of mammospheres per well was counted. Mammospheres were defined as groups with 25 or more cells originating from a single cell.

Quantification and statistical analysis

All data unless otherwise specified is displayed as mean ± standard deviation (SD). For comparing two groups, we employed Student’s t-tests. When comparing more than two groups, we utilized ANOVA and other methods for analysis. Statistical analysis was conducted using Microsoft Excel and GraphPad to calculate p-values, with significance set at p = <0.05. The Kaplan-Meier curves were generated using the Kaplan-Meier plotter (https://kmplot.com/). The Cox-Mantel (log-rank) test was used to evaluate the significance of differences between two patient cohorts, assessing whether the overall survival patterns differed throughout the entire study period. The difference between the cohorts is quantified by the hazard ratio (HR), which reflects the differential decline in survival between the two groups.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The RAW global MS data and the identified results generated in this study have been deposited in the Japan ProteOme STandard Repository (jPOST: https://repository.jpostdb.org/) under accession code JPST002098 for jPOST and PXD041009 for ProteomeXchange (https://repository.jpostdb.org/entry/JPST002098). Raw data supporting the findings of this study are available in the Source Data files, which include uncropped blot images, raw data corresponding to each figure. Source data are provided with this paper.

References

Min, S., Lee, B. & Yoon, S. Deep learning in bioinformatics. Brief. Bioinform 18, 851–869 (2017).
PubMed Google Scholar
Miotto, R., Wang, F., Wang, S., Jiang, X. & Dudley, J. T. Deep learning for healthcare: review, opportunities and challenges. Brief. Bioinform 19, 1236–1246 (2018).
Article PubMed Google Scholar
Xu, J. et al. Interpretable deep learning translation of GWAS and multi-omics findings to identify pathobiology and drug repurposing in Alzheimer’s disease. Cell Rep. 41, 111717 (2022).
Article PubMed PubMed Central CAS Google Scholar
Stahlschmidt, S. R., Ulfenborg, B. & Synnergren, J. Multimodal deep learning for biomedical data fusion: a review. Brief Bioinform. 23, bbab569 (2022).
Kang, M., Ko, E. & Mersha, T. B. A roadmap for multi-omics data integration using deep learning. Brief Bioinform 23, bbab454 (2022).
Wang, R. et al. The Clinicopathological features and survival outcomes of patients with different metastatic sites in stage IV breast cancer. BMC Cancer 19, 1091 (2019).
Article PubMed PubMed Central Google Scholar
DeSantis, C. E. et al. Breast cancer statistics, 2019. CA Cancer J. Clin. 69, 438–451 (2019).
PubMed Google Scholar
Loibl, S., Poortmans, P., Morrow, M., Denkert, C. & Curigliano, G. Breast cancer. Lancet 397, 1750–1769 (2021).
Article PubMed CAS Google Scholar
Takada, M. & Toi, M. Neoadjuvant treatment for HER2-positive breast cancer. Chin. Clin. Oncol. 9, 32 (2020).
Article PubMed Google Scholar
Chen, W., Hoffmann, A. D., Liu, H. & Liu, X. Organotropism: new insights into molecular mechanisms of breast cancer metastasis. NPJ Precis Oncol. 2, 4 (2018).
Article PubMed PubMed Central Google Scholar
Aceto, N. et al. Circulating tumor cell clusters are oligoclonal precursors of breast cancer metastasis. Cell 158, 1110–1122 (2014).
Article PubMed PubMed Central CAS Google Scholar
Dashzeveg, N. K. et al. New advances and challenges of targeting cancer stem cells. Cancer Res 77, 5222–5227 (2017).
Article PubMed PubMed Central CAS Google Scholar
Liu, H. et al. Cancer stem cells from human breast tumors are involved in spontaneous metastases in orthotopic mouse models. Proc. Natl Acad. Sci. USA 107, 18115–18120 (2010). [pii].
Article ADS PubMed PubMed Central CAS Google Scholar
Massague, J. & Obenauf, A. C. Metastatic colonization by circulating tumour cells. Nature 529, 298–306 (2016).
Article ADS PubMed PubMed Central CAS Google Scholar
Cristofanilli, M. et al. The clinical use of circulating tumor cells (CTCs) enumeration for staging of metastatic breast cancer (MBC): International expert consensus paper. Crit. Rev. Oncol. Hematol. 134, 39–45 (2019).
Article PubMed Google Scholar
Liu, X. et al. Homophilic CD44 Interactions mediate tumor cell aggregation and polyclonal metastasis in patient-derived breast cancer models. Cancer Discov. 9, 96–113 (2019).
Article PubMed Google Scholar
Pineiro, R., Martinez-Pena, I. & Lopez-Lopez, R. Relevance of CTC clusters in breast cancer metastasis. Adv. Exp. Med Biol. 1220, 93–115 (2020).
Article PubMed CAS Google Scholar
Schuster, E. et al. Better together: circulating tumor cell clustering in metastatic cancer. Trends Cancer 7, 1020–1032 (2021).
Article PubMed PubMed Central CAS Google Scholar
Szczerba, B. M. et al. Neutrophils escort circulating tumour cells to enable cell cycle progression. Nature 566, 553–557 (2019).
Article ADS PubMed CAS Google Scholar
Clotilde Costa, L. M.-R. et al. Analysis of a real-world cohort of metastatic breast cancer patients shows circulating tumor cell clusters (CTC-clusters) as predictors of patient outcomes. Cancers 12, 1111(2020).
Jansson, S., Bendahl, P. O., Larsson, A. M., Aaltonen, K. E. & Ryden, L. Prognostic impact of circulating tumor cell apoptosis and clusters in serial blood samples from patients with metastatic breast cancer in a prospective observational cohort. BMC Cancer 16, 433 (2016).
Article PubMed PubMed Central Google Scholar
Mu, Z. et al. Prospective assessment of the prognostic value of circulating tumor cells and their clusters in patients with advanced-stage breast cancer. Breast Cancer Res Treat. 154, 563–571 (2015).
Article PubMed CAS Google Scholar
Murlidhar, V. et al. Poor prognosis indicated by venous circulating tumor cell clusters in early-stage lung cancers. Cancer Res 77, 5194–5206 (2017).
Article PubMed PubMed Central CAS Google Scholar
Wang, C. et al. Longitudinally collected CTCs and CTC-clusters and clinical outcomes of metastatic breast cancer. Breast Cancer Res Treat. 161, 83–94 (2017).
Article PubMed Google Scholar
Cheung, K. J. et al. Polyclonal breast cancer metastases arise from collective dissemination of keratin 14-expressing tumor cell clusters. Proc. Natl Acad. Sci. USA 113, E854–E863 (2016).
Article PubMed PubMed Central CAS Google Scholar
Taftaf, R. et al. ICAM1 initiates CTC cluster formation and trans-endothelial migration in lung metastasis of breast cancer. Nat. Commun. 12, 4867 (2021).
Article ADS PubMed PubMed Central CAS Google Scholar
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Article ADS PubMed PubMed Central CAS Google Scholar
Liberzon, A. et al. Molecular signatures database (MSigDB) 3.0. Bioinformatics 27, 1739–1740 (2011).
Article PubMed PubMed Central CAS Google Scholar
Liberzon, A. et al. The molecular signatures database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015).
Article PubMed PubMed Central CAS Google Scholar
Ashburner, M. et al. Gene ontology: tool for the unification of biology. Gene Ontol. Consort. Nat. Genet 25, 25–29 (2000).
Article CAS Google Scholar
Gene Ontology, C. The gene ontology resource: enriching a GOld mine. Nucleic Acids Res 49, D325–D334 (2021).
Article Google Scholar
Krug, K. et al. Proteogenomic landscape of breast cancer tumorigenesis and targeted therapy. Cell 183, 1436–1456 e1431 (2020).
Article PubMed PubMed Central CAS Google Scholar
Kim, H. et al. Quantitative proteomics reveals knockdown of CD44 promotes proliferation and migration in claudin-low MDA-MB-231 and Hs 578 T breast cancer cell lines. J. Proteome Res 20, 3720–3733 (2021).
Article PubMed CAS Google Scholar
Velloso, F. J., Campos, A. R., Sogayar, M. C. & Correa, R. G. Proteome profiling of triple negative breast cancer cells overexpressing NOD1 and NOD2 receptors unveils molecular signatures of malignant cell proliferation. BMC Genomics 20, 152 (2019).
Article PubMed PubMed Central Google Scholar
Lanczky, A. & Gyorffy, B. Web-based survival analysis tool tailored for medical research (KMplot): development and implementation. J. Med Internet Res 23, e27633 (2021).
Article PubMed PubMed Central Google Scholar
Uhlen, M. et al. A pathology atlas of the human cancer transcriptome. Science 357, eaan2507 (2017).
Ramos, E. K. et al. Machine learning-assisted elucidation of CD81-CD44 interactions in promoting cancer stemness and extracellular vesicle integrity. Elife 11, e82669 (2022).
Chen, I. H. et al. Phosphoproteins in extracellular vesicles as candidate markers for breast cancer. Proc. Natl. Acad. Sci. USA 114, 3175–3180 (2017).
Article ADS PubMed PubMed Central CAS Google Scholar
Dai, J. et al. Exosomes: key players in cancer and potential therapeutic strategy. Signal Transduct. Target Ther. 5, 145 (2020).
Article PubMed PubMed Central CAS Google Scholar
Sandfeld-Paulsen, B. et al. Exosomal proteins as diagnostic biomarkers in lung cancer. J. Thorac. Oncol. 11, 1701–1710 (2016).
Article PubMed Google Scholar
Zhou, B. et al. Application of exosomes as liquid biopsy in clinical diagnosis. Signal Transduct. Target Ther. 5, 144 (2020).
Article PubMed PubMed Central CAS Google Scholar
Artigiani, S. et al. Functional regulation of semaphorin receptors by proprotein convertases. J. Biol. Chem. 278, 10094–10101 (2003).
Article PubMed CAS Google Scholar
Yu, W. et al. Plexin-B2 mediates physiologic and pathologic functions of angiogenin. Cell 171, 849–864 e825 (2017).
Article PubMed PubMed Central CAS Google Scholar
Zielonka, M., Xia, J., Friedel, R. H., Offermanns, S. & Worzfeld, T. A systematic expression analysis implicates Plexin-B2 and its ligand Sema4C in the regulation of the vascular and endocrine system. Exp. Cell Res 316, 2477–2486 (2010).
Article PubMed CAS Google Scholar
Atkin-Smith, G. K. et al. Plexin B2 is a regulator of monocyte apoptotic cell disassembly. Cell Rep. 29, 1821–1831 e1823 (2019).
Article PubMed CAS Google Scholar
Deng, S. et al. Plexin-B2, but not Plexin-B1, critically modulates neuronal migration and patterning of the developing nervous system in vivo. J. Neurosci. 27, 6333–6347 (2007).
Article PubMed PubMed Central CAS Google Scholar
Junqueira Alves, C. et al. Plexin-B2 orchestrates collective stem cell dynamics via actomyosin contractility, cytoskeletal tension and adhesion. Nat. Commun. 12, 6019 (2021).
Article ADS PubMed PubMed Central CAS Google Scholar
Li, Y. et al. Macrophages facilitate peripheral nerve regeneration by organizing regeneration tracks through Plexin-B2. Genes Dev. 36, 133–148 (2022).
Article PubMed PubMed Central CAS Google Scholar
Zhou, X. et al. Microglia and macrophages promote corralling, wound compaction and recovery after spinal cord injury via Plexin-B2. Nat. Neurosci. 23, 337–350 (2020).
Article PubMed PubMed Central CAS Google Scholar
Maier, V. et al. Semaphorin 4C and 4 G are ligands of Plexin-B2 required in cerebellar development. Mol. Cell Neurosci. 46, 419–431 (2011).
Article PubMed CAS Google Scholar
Gurrapu, S., Pupo, E., Franzolin, G., Lanzetti, L. & Tamagnone, L. Sema4C/PlexinB2 signaling controls breast cancer cell growth, hormonal dependence and tumorigenic potential. Cell Death Differ. 25, 1259–1275 (2018).
Article PubMed PubMed Central CAS Google Scholar
Maddipati, R. & Stanger, B. Z. Pancreatic cancer metastases harbor evidence of polyclonality. Cancer Discov. 5, 1086–1097 (2015).
Article PubMed PubMed Central CAS Google Scholar
Junqueira Alves, C., Yotoko, K., Zou, H. & Friedel, R. H. Origin and evolution of plexins, semaphorins, and Met receptor tyrosine kinases. Sci. Rep. 9, 1970 (2019).
Article ADS PubMed PubMed Central Google Scholar
Huang, Y. et al. Plexin-B2 facilitates glioblastoma infiltration by modulating cell biomechanics. Commun. Biol. 4, 145 (2021).
Article PubMed PubMed Central CAS Google Scholar
Charafe-Jauffret, E. et al. Breast cancer cell lines contain functional cancer stem cells with metastatic capacity and a distinct molecular signature. Cancer Res 69, 1302–1313 (2009).
Article PubMed PubMed Central CAS Google Scholar
Grillet, F. et al. Circulating tumour cells from patients with colorectal cancer have cancer stem cell hallmarks in ex vivo culture. Gut 66, 1802–1810 (2017).
Article PubMed CAS Google Scholar
Yang, M. H., Imrali, A. & Heeschen, C. Circulating cancer stem cells: the importance to select. Chin. J. Cancer Res 27, 437–449 (2015).
PubMed PubMed Central CAS Google Scholar
Malik, M. F., Ye, L. & Jiang, W. G. The Plexin-B family and its role in cancer progression. Histol. Histopathol. 29, 151–165 (2014).
PubMed CAS Google Scholar
Worzfeld, T. & Offermanns, S. Semaphorins and plexins as therapeutic targets. Nat. Rev. Drug Discov. 13, 603–621 (2014).
Article PubMed CAS Google Scholar
Worzfeld, T. et al. Genetic dissection of plexin signaling in vivo. Proc. Natl Acad. Sci. USA 111, 2194–2199 (2014).
Article ADS PubMed PubMed Central CAS Google Scholar
Conrotto, P., Corso, S., Gamberini, S., Comoglio, P. M. & Giordano, S. Interplay between scatter factor receptors and B plexins controls invasive growth. Oncogene 23, 5131–5137 (2004).
Article PubMed CAS Google Scholar
Le, A. P. et al. Plexin-B2 promotes invasive growth of malignant glioma. Oncotarget 6, 7293–7304 (2015).
Article PubMed PubMed Central Google Scholar
Roney, K. E. et al. Plexin-B2 negatively regulates macrophage motility, Rac, and Cdc42 activation. PLoS One 6, e24795 (2011).
Article ADS PubMed PubMed Central CAS Google Scholar
Xia, J. et al. Semaphorin-plexin signaling controls mitotic spindle orientation during epithelial morphogenesis and repair. Developmental cell 33, 299–313 (2015).
Article PubMed CAS Google Scholar
Qu, K. et al. MCM7 promotes cancer progression through cyclin D1-dependent signaling and serves as a prognostic marker for patients with hepatocellular carcinoma. Cell Death Dis. 8, e2603 (2017).
Article PubMed PubMed Central CAS Google Scholar
Wei, C. et al. Crosstalk between cancer cells and tumor associated macrophages is required for mesenchymal circulating tumor cell-mediated colorectal cancer metastasis. Mol. Cancer 18, 64 (2019).
Article PubMed PubMed Central Google Scholar
Ravenhill, B. J., Soday, L., Houghton, J., Antrobus, R. & Weekes, M. P. Comprehensive cell surface proteomics defines markers of classical, intermediate and non-classical monocytes. Sci. Rep. 10, 4560 (2020).
Article ADS PubMed PubMed Central CAS Google Scholar
Zhang, X., Zhu, M., Hong, Z. & Chen, C. Co-culturing polarized M2 Thp-1-derived macrophages enhance stemness of lung adenocarcinoma A549 cells. Ann. Transl. Med 9, 709 (2021).
Article PubMed PubMed Central CAS Google Scholar
Food, U. & Administration, D. FDA approves pembrolizumab for high-risk early-stage triple-negative breast cancer. US Food and Drug Administration (2021).
Shah, M. et al. FDA approval summary: pembrolizumab for neoadjuvant and adjuvant treatment of patients with high-risk early-stage triple-negative breast cancer. Clin. Cancer Res. 28, 5249–5253 (2022).
Article PubMed CAS Google Scholar
Food, U. & Administration, D. FDA grants accelerated approval to pembrolizumab for locally recurrent unresectable or metastatic triple negative breast cancer. News release. Silver Spring, MD: US Food and Drug Administration (2020).
Bardia, A. et al. Sacituzumab govitecan-hziy in refractory metastatic triple-negative breast cancer. N. Engl. J. Med. 380, 741–751 (2019).
Article PubMed CAS Google Scholar
Bardia, A. et al. Sacituzumab govitecan in metastatic triple-negative breast cancer. N. Engl. J. Med. 384, 1529–1541 (2021).
Article PubMed CAS Google Scholar
Clark, A. S. et al. Combination Paclitaxel and Palbociclib: Results of a Phase I Trial in Advanced Breast Cancer. Clin. Cancer Res 25, 2072–2079 (2019).
Article PubMed CAS Google Scholar
Gkountela, S. et al. Circulating Tumor Cell Clustering Shapes DNA Methylation to Enable Metastasis Seeding. Cell 176, 98–112 e114 (2019).
Article PubMed PubMed Central CAS Google Scholar
Hurtado, P., Martinez-Pena, I., & Piñerio, R. Dangerous Liasons: Circulating Tumor Cells (CTCs) and Cancer-Associated Fibroblasts (CAFs). Cancers 12 https://doi.org/10.3390/cancers12102861 (2020).
Sprouse, M. L. et al. PMN-MDSCs Enhance CTC Metastatic Properties through Reciprocal Interactions via ROS/Notch/Nodal Signaling. Int J Mol Sci 20 https://doi.org/10.3390/ijms20081916 (2019).
Aceto, N., Toner, M., Maheswaran, S. & Haber, D. A. En Route to Metastasis: Circulating Tumor Cell Clusters and Epithelial-to-Mesenchymal Transition. Trends Cancer 1, 44–52 (2015).
Article PubMed Google Scholar
Yu, M. et al. Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition. Science 339, 580–584 (2013).
Article ADS PubMed PubMed Central CAS Google Scholar
Roh-Johnson, M. et al. Macrophage contact induces RhoA GTPase signaling to trigger tumor cell intravasation. Oncogene 33, 4203–4212 (2014).
Article PubMed CAS Google Scholar
Harney, A. S. et al. Real-Time Imaging Reveals Local, Transient Vascular Permeability, and Tumor Cell Intravasation Stimulated by TIE2hi Macrophage-Derived VEGFA. Cancer Discov. 5, 932–943 (2015).
Article PubMed PubMed Central CAS Google Scholar
Duan, Y. et al. Semaphorin 5 A inhibits synaptogenesis in early postnatal- and adult-born hippocampal dentate granule cells. Elife 3 https://doi.org/10.7554/eLife.04390 (2014).
Li, S., Goncalves, K. A., Lyu, B., Yuan, L. & Hu, G. F. Chemosensitization of prostate cancer stem cells in mice by angiogenin and plexin-B2 inhibitors. Commun. Biol. 3, 26 (2020).
Article PubMed PubMed Central CAS Google Scholar
Borrelli, C. et al. In vivo interaction screening reveals liver-derived constraints to metastasis. Nature 632, 411–418 (2024).
Article PubMed PubMed Central CAS Google Scholar
Ito, D. & Kumanogoh, A. The role of Sema4A in angiogenesis, immune responses, carcinogenesis, and retinal systems. Cell Adh Migr. 10, 692–699 (2016).
Article PubMed PubMed Central CAS Google Scholar
Liu, X. et al. Sema4A Responds to Hypoxia and Is Involved in Breast Cancer Progression. Biol. Pharm. Bull. 41, 1791–1796 (2018).
Article PubMed CAS Google Scholar
Karki, A. et al. Plexin-B2 and Semaphorins Do Not Drive Rhabdomyosarcoma Proliferation or Migration. Sarcoma 2022, 9646909 https://doi.org/10.1155/2022/9646909 (2022).
Uhlen, M. et al. Proteomics. Tissue-based map of the human proteome. Science 347, 1260419 (2015).
Article PubMed Google Scholar
Klotz, R. et al. Circulating Tumor Cells Exhibit Metastatic Tropism and Reveal Brain Metastasis Drivers. Cancer Discov. 10, 86–103 (2020).
Article PubMed CAS Google Scholar
Luque, M. C. A., Gutierrez, P. S., Debbas, V., Kalil, J. & Stolf, B. S. CD100 and plexins B2 and B1 mediate monocyte-endothelial cell adhesion and might take part in atherogenesis. Mol. Immunol. 67, 559–567 (2015).
Article PubMed CAS Google Scholar
Tsai, C. F. et al. Surfactant-assisted one-pot sample preparation for label-free single-cell proteomics. Commun. Biol. 4, 265 (2021).
Article PubMed PubMed Central CAS Google Scholar
Tsai, C. F. et al. An Improved Boosting to Amplify Signal with Isobaric Labeling (iBASIL) Strategy for Precise Quantitative Single-cell Proteomics. Mol. Cell Proteom. 19, 828–838 (2020).
Article CAS Google Scholar
Teo, G. C., Polasky, D. A., Yu, F. & Nesvizhskii, A. I. Fast Deisotoping Algorithm and Its Implementation in the MSFragger Search Engine. J. Proteome Res 20, 498–505 (2021).
Article PubMed CAS Google Scholar
Demichev, V. et al. dia-PASEF data analysis using FragPipe and DIA-NN for deep proteomics of low sample amounts. Nat. Commun. 13, 3944 (2022).
Article ADS PubMed PubMed Central CAS Google Scholar
Zhang, T. et al. A proteome-wide assessment of the oxidative stress paradigm for metal and metal-oxide nanomaterials in human macrophages. NanoImpact 17 https://doi.org/10.1016/j.impact.2019.100194 (2020).
Dashzeveg, N. K. et al. Dynamic Glycoprotein Hyposialylation Promotes Chemotherapy Evasion and Metastatic Seeding of Quiescent Circulating Tumor Cell Clusters in Breast Cancer. Cancer Discov. 13, 2050–2071 (2023).
Article PubMed PubMed Central CAS Google Scholar

Download references

Acknowledgements

This work could not have been completed without the gracious contributions of the core facilities at Northwestern University including the Robert H. Lurie Comprehensive Cancer Center CTC Core, the Flow Cytometry Core, the Pathology Core, the Mouse Histology and Phenotyping Laboratory, and the Center for Comparative Medicine. We are deeply grateful to our collaborators and the Mass Spectrometry Core at Pacific Northwest National Laboratory. We would also like to thank our collaborators and all other Liu laboratory members at Northwestern University Feinberg School of Medicine. This project has been partially supported by Department of Defense grants W81XWH-16-1-0021 and W81XWH-20-1-0679 (H. Liu); NIH/NCI grants R01CA245699 (H. Liu and E.K. Ramos), R01AI167272, and R01CA298232 (H. Liu), R01GM139858 (T. Shi), and UG3CA256967 (T. Shi and H. Liu); the Lurie Cancer Center Lynn Sage Breast Cancer Research Foundation (H. Liu and N. Dashzeveg); American Cancer Society CSCC-Team-23-980420-01-CSCC (H. Liu); Chan Zuckerberg AGMT (H. Liu), a Northwestern University Pharmacology start-up grant (H. Liu); and NIH Fellowships T32 CA009560 (E.K. Ramos), T32 CA080621-15 and the Julius Kahn Fellowship (R. Taftaf), and T32GM008061 (E.J. Schuster).

Author information

These authors contributed equally: Emma Schuster, Nurmaa K. Dashzeveg, Fangjia Tong.
These authors jointly supervised this work: Chia-Feng Tsai, Tujin Shi, Huiping Liu.

Authors and Affiliations

Department of Pharmacology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
Emma Schuster, Nurmaa K. Dashzeveg, Fangjia Tong, Yuzhi Jia, Lamiaa El-Shennawy, Andrew D. Hoffman, Golam Kibria, Joshua R. Squires, Erika Ramos, Rokana Taftaf, David Scholten, Hannah F. Almubarak, Valery Adorno-Cruz, William Purev-Ochir & Huiping Liu
Driskill Graduate Program in the Life Sciences, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
Emma Schuster, David Scholten, Hannah F. Almubarak & Allegra C. Minor
Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA, USA
Tong Zhang, Reta Birhanu Kitata, Chia-Feng Tsai & Tujin Shi
ExoMira Medicine Inc, Chicago, IL, USA
Andrew D. Hoffman
Division of Hematology and Oncology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
Youbin Zhang, Sabina Spahija, Leonidas C. Platanias, Ami Shah, William J. Gradishar & Huiping Liu
Center for Artificial Intelligence in Drug Discovery, Case Western Reserve University, Cleveland, OH, USA
Chunlei Zheng & Rong Xu
Department of Pathology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
David P. Sullivan, Kalliopi P. Siziopikou & William A. Muller
Division of Hematology and Medical Oncology, Department of Medicine, Weill Cornell School of Medicine, New York, NY, USA
Carolina Reduzzi & Massimo Cristofanilli
Department of Biochemistry and Molecular Genetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
Allegra C. Minor, Kalliopi P. Siziopikou, Leonidas C. Platanias, Ami Shah, William A. Muller, William J. Gradishar & Huiping Liu
Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
Huiping Liu
Chan Zuckerberg Biohub Chicago, Chicago, IL, USA
Huiping Liu

Authors

Emma Schuster
View author publications
Search author on:PubMed Google Scholar
Nurmaa K. Dashzeveg
View author publications
Search author on:PubMed Google Scholar
Fangjia Tong
View author publications
Search author on:PubMed Google Scholar
Yuzhi Jia
View author publications
Search author on:PubMed Google Scholar
Lamiaa El-Shennawy
View author publications
Search author on:PubMed Google Scholar
Tong Zhang
View author publications
Search author on:PubMed Google Scholar
Andrew D. Hoffman
View author publications
Search author on:PubMed Google Scholar
Reta Birhanu Kitata
View author publications
Search author on:PubMed Google Scholar
Golam Kibria
View author publications
Search author on:PubMed Google Scholar
Youbin Zhang
View author publications
Search author on:PubMed Google Scholar
Joshua R. Squires
View author publications
Search author on:PubMed Google Scholar
Chunlei Zheng
View author publications
Search author on:PubMed Google Scholar
Erika Ramos
View author publications
Search author on:PubMed Google Scholar
Rokana Taftaf
View author publications
Search author on:PubMed Google Scholar
David Scholten
View author publications
Search author on:PubMed Google Scholar
Hannah F. Almubarak
View author publications
Search author on:PubMed Google Scholar
Valery Adorno-Cruz
View author publications
Search author on:PubMed Google Scholar
David P. Sullivan
View author publications
Search author on:PubMed Google Scholar
Carolina Reduzzi
View author publications
Search author on:PubMed Google Scholar
Allegra C. Minor
View author publications
Search author on:PubMed Google Scholar
William Purev-Ochir
View author publications
Search author on:PubMed Google Scholar
Sabina Spahija
View author publications
Search author on:PubMed Google Scholar
Rong Xu
View author publications
Search author on:PubMed Google Scholar
Kalliopi P. Siziopikou
View author publications
Search author on:PubMed Google Scholar
Leonidas C. Platanias
View author publications
Search author on:PubMed Google Scholar
Ami Shah
View author publications
Search author on:PubMed Google Scholar
William A. Muller
View author publications
Search author on:PubMed Google Scholar
William J. Gradishar
View author publications
Search author on:PubMed Google Scholar
Massimo Cristofanilli
View author publications
Search author on:PubMed Google Scholar
Chia-Feng Tsai
View author publications
Search author on:PubMed Google Scholar
Tujin Shi
View author publications
Search author on:PubMed Google Scholar
Huiping Liu
View author publications
Search author on:PubMed Google Scholar

Contributions

E.S., N.D., F.T., and H.L. conceived the idea presented here. H.L., T.S., and C-F.T. supervised experimental planning and implementation. N.D. performed an initial analysis of patient survival association with protein candidates identified from multiple datasets and breast cancer cells. E.S. drove the development of the presented storyline and data acquisition. E.S. and N.D. planned and carried out cell culture, clustering, imaging, flow cytometry/sorting, immunoprecipitation, and mass spectrometry sample preparation experiments. F. T. helped with the clustering experiments and data analyses. E.S., N.D., and F.T. planned and completed all animal work in vivo with the help of Y.J. and, later, L.E., D.P.S., and W.A.M. N.D. and G.K. carried out the EV mass spectrometry screen, including all necessary EV isolation and sample preparation. C-F.T. and R.B.K. performed mass spectrometry proteomic analyses of breast cancer cells and the co-immunoprecipitation samples, analyzed and deposited data. C.Z. and R.X. further helped analyze the mass spectrometry proteomic data of EVs. T.Z. and A.H. helped run the computational analysis across multiple cancer data sets to rank surface protein expression in breast cancer. M.C., A.S, W.J.G., L.C.P., C.R., and S.S. supported blood sample collection from breast cancer patients by enrolling patients, collecting blood work regularly for analysis, and managing patient data. E.S., N.D., Y.J., E.R., R.T., L.E-S., D.S., and V.A.C. all helped to process and analyze patient blood samples via flow cytometry. Y.Z. processed and analyzed patient blood samples via CellSearch® with the help of S.S., T.S., R.B.K., and C.F., ran and analyzed multiple mass spectrometry analyses on breast cancer CTCs and clusters. K.P.S. aided in the analysis of TMA data. J.S. helped to analyze images of CellSearch. H.A., A.M., and W.P. helped with Western blots. D.S. performed ImageStream analysis to validate the presence of CTC clusters and immune cells. E.S. and H.L. wrote the manuscript. F.T., N.K.D., T.S., Z.T., R.B.K., L.E-S., K.P.S. and V.A.C. edited the text.

Corresponding authors

Correspondence to Tujin Shi or Huiping Liu.

Ethics declarations

Competing interests

Huiping Liu and Andrew D. Hoffmann are scientific co-founders and equity shareholders of ExoMira Medicine Inc., whose business is not currently related to the content of this manuscript. Massimo Cristofanilli has the following financial relationships to disclose: (1) serving as consultant for AZ, Celcuity, Menarini-Stemline, Repare Therapeutics, Olaris, Syantra, BriaCell, Datar Cancer Genomics, Biotheryx; (2) received grant or research support from AZ, Celcuity; (3) received honoraria from AZ, Merck, Iylon, Menarini-Silicon Biosystem, Datar Cancer Genomics; and (4) participated in Speaker Bureau of Pfizer. Carolina Reduzzi receives research support from Menarini Silicon Biosystems, ANGLE, Qiagen, BioRad, Tethis. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Movie 1

Supplementary Movie 2

Supplementary Movie 3

Supplementary Movie 4

Supplementary Movie 5

Supplementary Movie 6

Supplementary Movie 7

Supplementary Movie 8

Supplementary Movie 9

Supplementary Movie 10

Supplementary Movie 11

Supplementary Movie 12

Supplementary Movie 13

Supplementary Movie 14

Supplementary Movie 15

Supplementary Movie 16

Supplementary Movie 17

Supplementary Movie 18

Supplementary Movie 19

Supplementary Movie 20

Supplementary Movie 21

Supplementary Movie 22

Supplementary Movie 23

Supplementary Movie 24

Supplementary Movie 25

Supplementary Movie 26

Supplementary Movie 27

Supplementary Movie 28

Supplementary Movie 29

Supplementary Movie 30

Supplementary Movie 31

Supplementary Movie 32

Supplementary Movie 33

Supplementary Movie 34

Supplementary Movie 35

Supplementary Movie 36

Reporting Summary

Transparent Peer Review file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Schuster, E., Dashzeveg, N.K., Tong, F. et al. Computational ranking identifies Plexin-B2 in circulating tumor cell clustering with monocytes in breast cancer metastasis. Nat Commun 16, 7649 (2025). https://doi.org/10.1038/s41467-025-62862-z

Download citation

Received: 21 February 2025
Accepted: 02 August 2025
Published: 16 August 2025
Version of record: 16 August 2025
DOI: https://doi.org/10.1038/s41467-025-62862-z

This article is cited by

CTC cluster in breast cancer: synergetic metastasis promotion mechanism and transformation path from platelet cloaking to leukocyte escort
- Cangtai Guan
- Liangyu Hao
- Lixiang Zheng
Clinical and Translational Oncology (2025)