Netrin-1 blockade inhibits tumour growth and EMT features in endometrial cancer

Netrin-1 is upregulated in cancers as a protumoural mechanism1. Here we describe netrin-1 upregulation in a majority of human endometrial carcinomas (ECs) and demonstrate that netrin-1 blockade, using an anti-netrin-1 antibody (NP137), is effective in reduction of tumour progression in an EC mouse model. We next examined the efficacy of NP137, as a first-in-class single agent, in a Phase I trial comprising 14 patients with advanced EC. As best response we observed 8 stable disease (8 out of 14, 57.1%) and 1 objective response as RECIST v.1.1 (partial response, 1 out of 14 (7.1%), 51.16% reduction in target lesions at 6 weeks and up to 54.65% reduction during the following 6 months). To evaluate the NP137 mechanism of action, mouse tumour gene profiling was performed, and we observed, in addition to cell death induction, that NP137 inhibited epithelial-to-mesenchymal transition (EMT). By performing bulk RNA sequencing (RNA-seq), spatial transcriptomics and single-cell RNA-seq on paired pre- and on-treatment biopsies from patients with EC from the NP137 trial, we noted a net reduction in tumour EMT. This was associated with changes in immune infiltrate and increased interactions between cancer cells and the tumour microenvironment. Given the importance of EMT in resistance to current standards of care2, we show in the EC mouse model that a combination of NP137 with carboplatin-paclitaxel outperformed carboplatin-paclitaxel alone. Our results identify netrin-1 blockade as a clinical strategy triggering both tumour debulking and EMT inhibition, thus potentially alleviating resistance to standard treatments.

Article before and after treatment, for translational research. Preliminary data from this trial were disclosed at the 2019 ESMO meeting 14 , but the trial is still ongoing, with patients continuing to receive treatment, and results will be fully reported after final analysis. In this article we report translational data generated in parallel with the Phase 1 study including patients with endometrial cancer and provide a series of preclinical and biopsy data demonstrating that NP137 not only reduces tumour cell number but also triggers inhibition of epithelial-to-mesenchymal transition (EMT) features, which ultimately increases tumour sensitivity to chemotherapy.

Netrin-1 and endometrial adenocarcinomas
We analysed netrin-1 expression by quantitative PCR with reverse transcription (RT-qPCR) in a cohort of 72 human endometrial tumours (Supplementary Table 1a). As shown in Extended Data Fig. 1a, netrin-1 is significantly upregulated in endometrial adenocarcinoma (EC) without specific change in grades or subtypes (Extended Data Fig. 1b,c). Its main receptor, UNC5B, was also found to be expressed more in tumour tissues than in normal endometrium (Extended Data Fig. 1a-c) while DCC, another netrin-1 receptor, is expressed neither in normal endometrium nor endometrial cancer. Netrin-1 (and UNC5B) positivity was monitored in most endometrial tumours by immunohistochemistry (IHC) (Extended Data Fig. 1d).
We thus moved to a preclinical model recapitulating the development of EC, using the genetically engineered, temporally controlled Pten f/f-deleted mouse model (namely, the tamoxifen-inducible CAG-CreERT +/− promoter), which has been shown to develop EC in situ rapidly, as well as thyroid hyperplasia 15 (Fig. 1a). We thus treated control and Pten-deleted mice for 3-4 weeks with NP137 (10 mg kg -1 , three times per week) and analysed netrin-1 expression and tumour progression. As shown in Fig. 1b,c, netrin-1 was upregulated in mouse tumours following deletion of Pten, and this upregulation was decreased following NP137 treatment. Of interest, NP137 was associated with decreased development of endometrial tumours (Extended Data Fig. 2a) and increased survival ( Fig. 1d; note that there was no extension of the study after 7 weeks following Pten deletion because most mice showed breathing difficulties or other ailments due to development of thyroid tumours). Pathologists observed a decreased number of cancer cells following NP137 treatment (Fig. 1e) and healthier endometrial tissue (Fig. 1f). Similarly, antitumour activity was also observed in the thyroids of mice treated with NP137 (Extended Data Fig. 2b-e). These results indicate that targeting netrin-1 in EC inhibits tumour progression.

Objective response in a patient with EC
Based on previous results suggesting that netrin-1 blockade is a viable therapeutic strategy in cancer, a netrin-1-blocking antibody was developed for clinical use 13 and is undergoing Phase I/II evaluation. We extracted efficacy data for 14 patients with EC from the ongoing Phase 1 study (Supplementary Table 1b,c). In this study, NP137 was administered once every 2 weeks (Q2W) as monotherapy until clinical/ radiological progression. As shown in Fig. 2a and Extended Data induce Pten deletion into CAG-CreERT2 +/− Pten f/f mice using tamoxifen injection and either treatment with NP137 or control. Mice were euthanized if they experienced breathing difficulties 15 . b, Relative messenger RNA expression of netrin-1 as defined by RT-qPCR in the endometrium of CreERT2 -/animals (n = 12) and in tumours induced following Pten deletion (tamoxifen injection) in CreERT2 +/− Pten f/f mice intraperitoneally treated with NP137 (10 mg kg -1 ) (n = 12) and in control (n = 5   Table 1b). The original diagnosis for this patient showed an endometrioid origin, a microsatellite stable phenotype, and expression of CK7, PAX8 and oestrogen receptors but no expression of CK20 or progesterone receptor. Before administration of NP137 she had received multiple therapeutic attempts, including adjuvant radiotherapy and carboplatin-paclitaxel followed by lurbinectedin as first-line treatment for metastatic disease, and a second attempt with carboplatin-paclitaxel, but had progression of liver metastases despite these therapeutics.

Article
A positron emission tomography-computed tomography (PET-CT) scan, performed at inclusion, confirmed intense uptake of fluorodeoxyglucose (before C1D1). She received 14 mg kg -1 intravenous NP137 Q2W and underwent a PET-CT scan at 6 weeks (that is, post cycle 3 of NP137), showing partial response according to RECIST v.1.1 with a 51% reduction in target liver lesions (Fig. 2b,c and Extended Data Table 1b). Partial response was confirmed on PET-CT scan at 3 months (Fig. 2c) and then again at 6 months, when size reduction of target lesions reached 55% (Extended Data Table 1b). This patient eventually experienced disease progression after 17 cycles of NP137 and went on to receive additional therapy, including letrozole A, immunotherapy and tamoxifen, without experiencing additional objective response.

Netrin-1 blockade inhibits tumour EMT
To gain insight into the underlying mechanisms that link netrin-1 blockade to tumour growth inhibition we first analysed, based on the hypothesized mode of action of netrin-1 blockade, whether tumour growth inhibition was associated with tumour cell death in Pten f/f tumours treated with NP137. As shown in Fig. 3a,b, NP137 increased cancer apoptosis as measured by IHC on active caspase-3. We next performed RNA-seq of Pten f/f mouse tumours treated with NP137 versus untreated. Among the pathways/genes modulated following NP137 administration, we noted a decrease in EMT-related genes following treatment. Several in vitro studies have suggested the involvement of netrin-1 in EMT [16][17][18][19] , often associated with the PI3K/AKT pathway, which is frequently altered in endometrial cancer 17,18 . We then investigated whether NP137 might impact tumour EMT in the Pten f/f mouse model. We first assessed EpCAM epithelial marker expression in control versus NP137-treated tumours, and observed a statistically significant increase of this epithelial marker in NP137-treated tumours (Fig. 3c,d). To gain a more general view of tumour gene expression we then utilized a commonly used pancancer EMT signature 20 that can be employed to determine EMT score 20 . We observed that treatment with NP137 decreased EMT score ( Fig. 3e), associated with decreased expression of mesenchymal genes and increased expression of epithelial genes (Fig. 3f). These preclinical data support the view that netrin-1 blockade has a dual action on tumour cells: triggering of cancer cell death and inhibition of EMT features, rendering overall NP137-treated tumours more epithelial. We next determined whether the inhibition of EMT features observed in the preclinical models also occurs in patients treated with NP137. In the NP137 Phase 1 trial, among patients with EC described above, paired biopsies were collected at inclusion (C1D1) and after 1 month of treatment with the anti-netrin-1 compound (that is, just before the third infusion of NP137 (C3D1)). Bulk RNA-seq was successfully performed on 12 paired pre-and on-treatment biopsies (Supplementary Table 1b) and, as shown in Fig. 4a,b, two injections of NP137 were found sufficient to trigger a significant decrease in the pancancer EMT score described above, indicating an overall more epithelial phenotype of the tumour in patients after 1 month of treatment with NP137. The shift toward a more epithelial phenotype was confirmed by increased EpCAM IHC staining in tumour samples from patients treated with NP137 ( Fig. 4c,d). We also observed a decrease in the proportion of tumour cells coexpressing pancytokeratin and vimentin (Fig. 4e).
Of interest, when analysing changes in EMT score between baseline and on-treatment biopsies we observed that samples from patients who had disease progression as their best response on NP137 (progressive disease at first evaluation at 6 weeks) failed to show variation in EMT score. Conversely, there was a statistically significant decrease in EMT score in samples from patients who had disease control (at least stable disease at 6 weeks; Extended Data Fig. 3a).
To demonstrate more formally that the bulk change in EMT score seen with NP137 treatment in patients was occurring specifically in cancer cells, as a first approach we used the 10x Visium spatial gene expression system compatible with formalin-fixed, paraffin-embedded (FFPE) tissues. Two pairs of C1D1/C3D1 FFPE sections from two patients with EC were analysed (Extended Data Fig. 3b,c) and spatial gene expression profiling was performed in pathologist-selected regions of interest where only cancer cells could be identified by haematoxylin and eosin (H&E) staining. In both cases, EMT score was strongly decreased at C3D1 compared with C1D1 (Fig. 4f). Thus in patients, after 1 month of treatment with NP137, remaining cancer cells showed decreased EMT features compared with their pretreatment status.
To extend these studies to the single-cell level we performed 3′ single-cell RNA-seq (10x Genomics Next GEM 3′ kit) directly on fresh biopsies obtained at inclusion (C1D1, 9,216 cells) in the NP137 trial and after 1 month of treatment (C3D1, 7,159 cells) from a patient with advanced EC (patient no. 01-040; Supplementary Table 1b

Article
Unsupervised clustering showed the presence of different cell populations including tumour cells (expressing EpCAM, PGR and TFF3-all markers of ECs 21 ), immune cells (marked by CD45 (PTPRC) expression), cancer-associated fibroblasts (CAFs, marked by αSMA (ACTA2)) and endothelial cells (PECAM1 positive) ( Fig. 5b and Extended Data Fig. 4a,b). Treatment with NP137 led to a statistically significant decrease in the tumour cell compartment (Fig. 5b,c and Extended Data Fig. 4b).
The proportion of tumour cells following anti-netrin-1 treatment was 3.19 times lower after two cycles of NP137 (Extended Data Fig. 4b).
Of interest, in addition to the net decrease in cancer cells we noted a statistically significant decrease in EMT score in the whole tumour compartment, indicating an overall more epithelial phenotype associated with NP137 treatment (Fig. 5d). We also performed unbiased differential expression and pathway analysis only in tumour cells (Extended Data Table 2). Remarkably, EMT was the most significant term, with downregulation following exposure to NP137 (Extended Data Table 2). Of interest, while following NP137 treatment, most of the subcompartments of tumour cells decreased in a similar way and we noted that tumour cluster 2, which had the strongest decrease, also showed a high EMT score (Fig. 5e-h; note that clusters 5-7, with high EMT score, were not detectable following NP137 treatment but we cannot exclude the possibility that they were not captured, because these clusters already had very few cells at C1D1).
The change occurring in the tumoural compartment was also associated with a change in stromal cells (Figs. 5b,c and 6 and Extended Data Figs. 4c-f, 5 and 6a). Of note, NP137 treatment clearly appeared to have an impact on immune cells (Fig. 6a,b and Extended Data Fig. 5a). More specifically, following NP137 treatment we noted an increase in lymphocytes endowed with cytotoxic functions (CD8 + T cells and natural killer (NK) cells; Fig. 6c and Extended Data Fig. 5b). A similar increase in CD8 + cells was also noted in Pten f/f mice treated with NP137 (Extended Data Fig. 5c,d). Of interest, following NP137 treatment we noted both in the single-cell analysis from patient no. 01-40 and in the spatial transcriptomic data from patient nos. 01-034 and 01-039 an increase in both the number and strength of interaction between T cells and tumour cells (Fig. 6d,e). In particular, single-cell RNA-seq data showed an increase in the number and strength of interactions between CD8 + T cells and tumour cells (Fig. 6d). A decrease in M2 macrophages was detected (Extended Data Fig. 5e,f), together with an increase in major histocompatibility class I/II antigen presentation (Extended Data Fig. 6a,b). NP137 treatment is associated with more efficient antigen-presenting cells (APCs) because we observed a clear switch from monocytes in C1D1 to dendritic cells in C3D1 interacting with cancer cells (Fig. 6d,e and Extended Data Fig. 6b).

NP137 inhibits chemotherapy resistance
Because of a large body of literature describing EMT as a major cause of chemotherapy resistance 2,22 and because our data suggest that targeting of netrin-1 inhibits EMT, we assessed whether the addition of NP137 to carboplatin-paclitaxel (carbotaxol), the standard-of-care chemotherapy used in clinical settings in endometrial cancer, might enhance the efficacy of carbotaxol alone in the Pten f/f mouse model of endometrial adenocarcinoma. As shown in Extended Data Fig. 7, NP137/carbotaxol treatment proved superior to carbotaxol alone, leading to complete responses in mice. Altogether, these preclinical and clinical data demonstrate that a clinical-stage drug both induces tumour cell debulking and triggers a general inhibition of EMT features, which offers the possibility of alleviating resistance to conventional therapies.

Discussion
We provide here documentation of a clinical activity of netrin-1 titration using NP137 as monotherapy. Whereas the interest in targeting netrin-1 in the cancer field is relatively recent 13 , this report indicates that targeting netrin-1 in endometrial cancer may be effective. In addition to the generation of a series of data, including analysis of a human cohort and preclinical experiments in mice, we have confirmed a key mechanistic role for netrin-1 in endometrial cancer resistance and progression. We also report a partial response detected in a patient with EC receiving an anti-netrin-1 mAb. This partial response in a human patient, together with decrease in tumour cell counts in mice and marked decrease in tumour cells observed in single-cell RNA-seq analysis of a patient treated with NP137, further support the initial view of the mode of action of the netrin-1 mAb as an inducer of tumour cell death 1,23 . Indeed, netrin-1 was previously considered to be an embryonic secreted molecule re-expressed in cancer settings to promote tumour cell survival 1,8 . NP137, by blocking netrin-1, is theorized as unleashing the pro-death activity of netrin-1 receptor as observed in various preclinical models [11][12][13]24 . In patients with EC or preclinical models of EC treated with NP137, this is expected to be translated into a decrease in tumour cells. However, we have shown here that, not only does NP137 induce tumour cell death, but it also appears to impact tumour EMT. Only a few previous studies have suggested that netrin-1 may be implicated in EMT, and these showed only in vitro data [16][17][18][19] , thus only weakly supportive of a major regulatory effect of netrin-1 on EMT 25 . Tumour EMT, in which tumour cells lose their epithelial characteristics and acquire mesenchymal features, appears to be a key driver of tumour heterogeneity and has been associated with different steps in tumourigenesis such as tumour initiation, progression, metastasis and, more recently, resistance to chemotherapy or immunotherapy 2 . Whereas great progress in the understanding of the role and mechanisms by which EMT regulates these different tumour functions has been achieved, there is still virtually no pharmacological intervention that allows the clinician to alleviate EMT in tumours. It is also fair to say that, to date, because there are no clinical-stage drugs impacting only on EMT features, there is no clinical demonstration that EMT is clinically important. Here we demonstrate, using both preclinical models and pre-and on-treatment biopsies from patients with EC, that systemic treatment with NP137 led to a decrease in tumour EMT features. This inhibition of EMT features is associated with an overall more epithelial phenotype. Given the extreme complexity of tumour heterogeneity, we have not yet demonstrated whether the effect of N137 on tumour EMT is mediated by a direct effect of netrin-1 blockade on cancer cells or whether this effect is due to an indirect effect triggered by changes in the tumour microenvironment. It is likely that these effects are combined, because we have shown using single-cell analysis that changes in EMT features are associated with changes in the tumour microenvironment. Although further confirmation is required, we observed a decrease in cancer-associated fibroblasts, which usually serve as primary source of EMT-inducing molecules 26 , and a decrease in protumourigenic M2-like macrophages, which also contribute to EMT by multiple mechanisms 27 . Moreover, we observed an increase in lymphocytes endowed with cytotoxic functions and an increase in both the number and strength of interactions between immune cells and tumour cells associated with more highly efficient APCs. Together this supports the view that NP137, possibly by impacting on EMT, enhances tumour immune response. Whatever the mechanism, because there is an increasing literature describing EMT as a major player in resistance to chemotherapy and immune checkpoint inhibitors 2,28 , the observation that treatment with NP137 inhibits features of tumour EMT argues for the clinical assessment of combinations of the anti-netrin-1 mAb with conventional therapies to interfere with tumour progression. This is currently investigated in the Phase 2 GYNET trial (NCT04652076) assessing the safety and efficacy of combining NP137 with carboplatin-paclitaxel and/or pembrolizumab (anti-PD1 mAb) in patients with endometrial or cervical cancer.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06367-z.

Genetically modified mouse model and NP137 administration
Floxed homozygous Pten (C;129S4-Ptentm1Hwu/J, hereafter called Pten f/f) Cre:ER (B6.Cg-Tg(CAG-CRE/Esr1* 5Amc/J) mice were obtained from the Jackson Laboratory. Cre:ER +/− Pten f/f mice were bred in a mixed background (C57BL6; 129S4) by crossing Pten f/f and Cre:ER +/− mice. To obtain mice carrying both Pten floxed alleles (Pten f/f) and a single Cre:ER (Cre:ER +/− ), Cre:ER +/− Pten f/ + mice were backcrossed with Pten f/f mice. To induce deletion of floxed alleles, tamoxifen (Sigma-Aldrich) was dissolved in 100% ethanol at 100 mg ml -1 . Tamoxifen solution was emulsified in corn oil (Sigma-Aldrich) at 10 mg ml -1 by vortexing. To induce Pten deletion, adult mice (4-5 weeks old) were given a single intraperitoneal injection of 0.5 mg of tamoxifen emulsion (30-35 µg mg -1 body weight). Three weeks after tamoxifen injection, mice were treated via intraperitoneal injection of 100 µl of NP137, or its isotypic control NP001 diluted in PBS, at 10 mg kg -1 every 2 days. Animal care and housing were in accordance with institutional European guidelines from the CEEA local Ethical committee of Lleida University concerning Pten mouse experiments. General behaviour and weight were monitored three times per week and animals were killed in the event of strong alteration or weight loss under 20%. All mice were always killed before terminal tumour progression, and endometrial tissues were analysed blind by a pathologist.

RT-qPCR
Endometrial samples collected in the Biomedical Research institute of Lerida (Spain) were frozen and sent to the Cancer Research Center of Lyon (France) in dry ice. Samples were cryoground to obtain tumour powder, which was processed for total RNA extraction using the Nucleospin RNA Plus kit (Machery-Nagel) according to the manufacturer's instructions. Expression of mRNA was measured using a NanoDrop1000 (Themo Scientific). RNA was retrotranscribed using the T100 ThermoCycler (Bio-Rad) and the iScript cDNA Synthesis Kit (Bio-Rad) according to the manufacturer's instructions. RT-qPCR was performed using LC480 qPCR (Roche) and OneGreen Fast qPCR Premix (Ozyme) according to the manufacturers' instructions.
Bulk RNA-seq Patient analysis: patient microbiopsy RNA was extracted with the RNA easy FFPE kit (Qiagen). The RNA-seq library was produced from 100 ng of RNA with the Illumina TruSeq Exome kit (RNA Library Prep for Enrichment & TruSeq RNA Enrichment) according to the manufacturer's instructions and then sequenced with an Illumina NovaSeq 6000. FASTQ files were then processed with STAR (v.2.7.10a). Briefly, FASTQ files were mapped to the human reference genome (gencode.v.27) and aligned reads were converted for counting with STAR. The quality of FASTQ files was also checked, by FATSQC (v.0.11.9). RNA-seq analysis was performed with R (v.4.0.3) and the DESeq2 package (v.1.30.1). log 2 -Transformed transcripts per million were calculated, and we performed EMT score calculation as previously described 20 .
Murine endometrial cancer model: tumours were collected, snap-frozen and cryoground. RNA was extracted with a standard kit (Macherey-Nagel). The RNA-seq library was produced from RNA with Illumina TruSeq Stranded Total RNA Library Prep Human/Mouse/Rat, according to the manufacturer's instructions, then sequenced with Illumina NOVASeq. FASTQ files were processed as described above, except for the reference genome which was Mus_musculus.GRCm38 (GENE-CODE release 25). Analysis was done with DESeq2 (v. 1

Single-cell RNA-seq
Endometrial metastasis biopsies were dissociated for single-cell RNA-seq using the Tumor Dissociation kit by Miltenyi Biotec (no. 130-095-929). Briefly, the biopsy was placed in RPMI medium in a Petri dish on ice and cut into small pieces (2-4 mm) after removal of necrotic tissue. Pieces were then infused with the RPMI/enzyme mix (Miltenyi Biotec), transferred to a gentleMACS C tube containing RPMI/enzyme mix, attached to the sleeve of the gentleMACS Octo Dissociator and run using the programme 37_h-TDK1. After completion of the programme the cells were spun down at 300g for 7 min at 4 °C, resuspended in RPMI, passed through a 70 µm strainer and centrifugation was repeated. The cell pellet was treated with 500 µl of ACK solution for 5 min at room temperature and lysis then stopped with 5 ml of RPMI/10% FBS. After centrifugation, cells were resuspended in 100 µl of RPMI. The number of live cells was determined with a Luna-FL Dual fluorescence cell counter (Logos Biosystems) to obtain an expected cell recovery population of 10,000 cells per channel, loaded on a 10x G chip and run on the Chromium Controller system (10x Genomics) according to manufacturer's instructions. Single-cell RNA-seq libraries were generated with the Chromium Single Cell 3′ v.3.1 kit (10x Genomics, no. PN-1000121) and sequenced on the NovaSeq 6000 platform (Illumina) to obtain around 50,000 reads per cell.
Filtered barcoded matrices from single-cell RNA-seq data were imported into R using the Seurat package (v.4.1.1). Doublets were detected with DoubletFinder (v.2.0.3) and filtered out, together with cells showing a low number of features (nFeature_RNA < 500) or a high percentage of mitochondrial genes (above 25%). Seurat functions were used for normalization (SCTransform), merging, dimensional reduction and clustering. Initial cell type identification was based on consensus from several automated cell annotation packages (SciBet' v.1.0, SingleR v.1.10.0 and scType (https://github.com/IanevskiAleksandr/sc-type/ blob/master/README.md)). T cell subtypes were visually inspected and manually curated after further annotation with ProjecTILs (v. 3 Spatial RNA-seq matrices and images were imported into R using the Load10X_Spatial function of the Seurat package. For each sample, only those spots with more than 1,000 features were kept for downstream preprocessing, including SCTransform normalization, dimensional reduction and clustering. Most analyses were performed independently for each sample (without merging or integration). Two strategies were used for cell type annotation: label transfer following integration of the single-cell RNA-seq data described above, and manual annotation using known markers for major cell types.
Inference of cell-cell communication was done with CellChat (v.1.6.0), for both single-cell and spatial RNA-seq data.

Histology and IHC analysis
Cre:ER +/− Pten f/f mice were euthanized by cervical dislocation after 3 weeks of NP137 treatment. Endometrial samples were collected and formalin fixed overnight at 4 °C. Tumours were paraffin embedded for further histologic analysis. Paraffin blocks were sectioned at 3 µm and dried for 1 h at 65 °C before the pretreatment procedures of deparaffinization, rehydration and epitope retrieval in the pretreatment module at 95 °C for 20 min in 50× Tris/EDTA buffer. Before staining of sections, endogenous peroxidase was blocked. Representative images were taken with a Leica DMD108 microscope.

Multiplex IHC
Sequential chromogenic multiplex IHC for vimentin/pancytokeratin (panCK), as previously described, was performed on tumour sections from patients included in the NP137 clinical trial and that were collected at C1D1 and C3D1. Dewaxed 4-µm-thick, paraffin-embedded tissue sections were subjected to two successive steps of IHC on a Ventana discovery XT platform (Ventana, Roche Diagnostics) using the REDMap and DABMap detection systems according to the manufacturer's recommendations. In a first step for vimentin expression, slides were incubated with mouse monoclonal anti-human vimentin for 1 h (mouse, clone V9, Leica, no. NCL-L-VIM-V9, dilution 1:100) and incubated with rabbit monoclonal anti-mouse secondary antibody for 20 min (clone M1gG51-4, abcam, no. 125913, dilution 1:750). The slides were then incubated with biotinylated anti-rabbit secondary antibody for 24 min (Vector Laboratories, dilution 1:200) followed by the addition of the streptavidin-alkaline phosphatase complex. Immunostaining was detected by incubation with naphthol and Fast red. Tissue sections were counterstained with Gill's haematoxylin, dehydrated and mounted. Whole histological slides were digitized at ×20 magnification using a Hamamatsu 2.0 HT scanner. After removal of coverslips, slides were incubated in 100% ethanol until complete erasure of red colour. In a second step, to show panCK expression, the slides were incubated with mouse monoclonal anti-panCK antibody for 1 h (mouse, clone CKAE1AE3, Dako Belgium, no. M351529-2, dilution 1:150) and then with rabbit monoclonal anti-mouse secondary antibody for 20 min. The slides were then incubated with biotinylated anti-rabbit secondary antibody for 28 min (Vector Laboratories, dilution 1:200) followed by the addition of streptavidin-alkaline phosphatase. PanCK immunostaining was detected by incubation with naphthol and Fast red. The IHC slides were counterstained with Gill's haematoxylin, dehydrated, mounted and again digitized. Image processing and analysis for Hynrid EMT score computation were performed using Visiomorph DP 2018.4 to determine vimentin and panCK co-expression in each tissue slide. Briefly, each pair of vimentin-and panCK-labelled virtual slides, which were acquired from the same tissue section, was subjected to image registration. Vimentin-and panCK-positive areas were automatically detected in the aligned virtual slides to evidence their co-expression in tumour cells. This co-expression was evaluated on whole slides at ×10 magnification to take into account potential imperfections. Manual corrections were carried out to exclude irrelevant sample parts, such as necrosis. Cell nuclei negative for both markers were also excluded, to focus only on cytoplasmic areas where colocalization could occur.
Spatial transcriptomics using Visium FFPE technology FFPE tissue sections were placed on Visium slides and prepared according to the 10x Genomics protocols. After H&E staining, imaging and de-crosslinking steps, tissue sections were incubated with human-specific probes targeting 17,943 genes (10x Genomics, Visium Human Transcriptome Probe Set v.1.0). Probes hybridized on mRNA were captured on Visium slides and a gene xpression library prepared following the provided protocol and sequenced on an Illumina NovaSeq 6000 with 50,000 reads per spot targeted sequencing depth.
For each FFPE section, FASTQ files and histology images were processed using 10x Space Ranger v.2.0 to obtain the gene expression matrix associated with each spot.
Seurat v.4 (https://satijalab.org/seurat/) in R 4.1 was used to perform the analysis. Briefly, filtered matrices were loaded, merged per patient and spots with fewer than 1,000 detected genes were removed. Following SCTransform normalization we subset the tumoural spot according to pathologist spot identification then calculated the EMT gene set enrichment score (escape R package) with the UCell method.

NP137 clinical trial
NP137 is a first-in-human Phase I trial with a dose-escalation part followed by extension cohorts (NCT02977195) conducted in adult patients with advanced or metastatic solid tumours. The dose-escalation part was initiated using an accelerated dose titration with one patient per dose level until the occurrence of a grade 2 or higher drug-related adverse event. Following occurrence of a grade 2 NP137-related adverse event, patients were enrolled in a slower dose-escalation design with at least three patients per dose level using a modified continual reassessment method. Additional patients were enrolled in three biomarker cohorts, at dose levels that had been declared safe, and underwent paired biopsies for pharmacodynamics purposes. In the dose-escalation part, 19 patients were enrolled in seven dose levels (1-20 mg kg -1 , intravenous, Q2W). No dose-limiting toxicities were observed and 11 (58%) patients had infusion-related reactions of grade 12 severity, all at doses of 4 mg kg -1 and above 14 . Based on available data, 14 mg kg -1 Q2W was selected as the recommended Phase 2 dose. Two extension cohorts were opened, including one in patients with hormone receptor-positive EC (enrolment closed in October 2021).
The trial was conducted according to Good Clinical Practice guidelines, the Declaration of Helsinki and relevant French and European laws and directives. All patients provided written informed consent.

Statistical analysis
Statistical analyses were performed on Prism (GraphPad Software). In the figure legends, n denotes the total number of replicates. All statistical tests were two-sided. For mouse experiments, statistical methods were not used to predetermine necessary sample size but sample size was chosen based on pilot experiments applying appropriate statistical tests that could return significant results. Survival curves were analysed using the log-rank (Mantel-Cox) test. Population ratios were analysed by chi-squared tests. qPCR expression, caspase-3 IHC and thyroid weights were compared by Mann-Whitney test. For data involving patients, gene expressions and EMT scores were compared by t-test.

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

Data availability
The raw RNA-seq, single-cell RNA-seq and spatial transcriptomic sequencing data of this study have been deposited in Gene Expression Omnibus with accession no. GSE225691. For version control, sharing and reproducibility, all bioinformatic code related to the single-cell analysis is retained in a GitHub repository (https://github.com/ hernandezvargash/NP137_single.cell). Source data are provided with this paper.