For RNA sequencing, patients were included if they had histologically confirmed ccRCC with VTT. The ccRCC thrombus-tumor-normal tissue triples of 5 cases were obtained following nephrectomy and tumor thrombus resection ( Supplementary Table 1 ). For mass spectrometry, 12 patients with histological-type ccRCC undergoing nephrectomy and 11 healthy donor volunteers from the same period were included ( Supplementary Table 2 ). Their samples of the second urine in the morning were collected before surgery in sterile tubes containing 1 mM of phenylmethanesulfonyl fluoride (Sigma, St. Louis, MO) to inhibit proteases. In addition, 54 urine samples from an independent cohort of consecutive ccRCC patients were also collected for ELISA analysis ( Supplementary Table 3 ). Figure 1 shows a workflow summary of the transcriptomic and proteomic research that revealed characteristics of ccRCC with VTT and developed a urine-based prognostic classifier to predict ccRCC prognosis. The study was approved by the ethics committee of Changhai Hospital, Naval Medical University, and written informed consent was obtained from all participants prior to study enrollment.

Total RNA of thrombus, tumor and normal tissue from ccRCC patients was extracted using the mirVana miRNA Isolation Kit (Ambion, TX, USA) following the manufacturer’s instructions. RNA purity was checked using a NanoPhotometer spectrophotometer (IMPLEN, CA, USA). The TruSeq Stranded mRNA LTSample Prep Kit (Illumina, CA, USA) was used to build the libraries. Then these libraries were sequenced on the Illumina sequencing platform (HiSeqTM 2500 or Illumina HiSeq X Ten) and 150 bp paired-end reads were generated.

The urine samples were centrifugated to collect the supernatant, and then the protein extract in urine supernatant was digested into peptides with trypsin. The peptides were subjected to capillary source followed by the timsTOF Pro (Bruker Daltonics) mass spectrometry. The electrospray voltage applied was 1.60 kV. Precursors and fragments were analyzed at the TOF detector, with a MS/MS scan range from 100-1700 m/z. The timsTOF Pro was operated in parallel accumulation serial fragmentation (PASEF) mode. Precursors with charge states 0 to 5 were selected for fragmentation, and 10 PASEF-MS/MS scans were acquired per cycle. The dynamic exclusion was set to 30s.

The analyses of differentially expressed genes (DEGs) and differentially expressed proteins (DEPs) were performed using the “limma” package of R statistical software.

DEGs were divided among three groups: RCC vs. normal renal tissue (NRT), VTT vs. NRT, VTT vs. RCC. The DEGs which co-expressed in RCC vs. NRT and VTT vs. NRT and those in VTT vs. RCC were defined as thrombus invasion-associated genes. Furthermore, DEPs were selected based on their different levels between urinary samples of ccRCC patients and healthy controls. DEGs/DEPs were defined by |log2 FC|>2 and P<0.05. For the public single-cell RNA sequencing data, the transcriptional profiles from all ccRCC patients and samples were visualized via uniform manifold approximation and projection. Then, the normalized expressions of DEGs were presented in all single-cell clusters and compared among tissues of ccRCC tumor, adjacent normal kidney, and lymph node. Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were performed using the “clusterProfiler” package of R statistical software.

Using the survival package, the univariate Cox regression analysis was carried out to targeted proteins linked to OS. (version 3.3.1; https://github.com/therneau/survival). The optimal prognostic protein set for OS was further screened on the basis of SVM-RFE method using the e1071 (version 1.7.1; https://cran.r-project.org/web/packages/e1071) and caret packages (version 6.0.76; https://cran.r-project.org/web/packages/caret). The SVM classifier was then built to predict OS according to the expression levels of optimal prognostic protein set. Additionally, the results of the SVM classification analysis were validated using data from The Cancer Genome Atlas-Kidney Renal Clear Cell Carcinoma (TCGA-KIRC) dataset.

The multivariate Cox regression analysis was performed to extract independent prognostic genes for OS using survival package (version 3.3.1; https://github.com/therneau/survival). Afterwards, a risk score model of prognostic makers was established according to following formula: risk score = ∑βDEPs × ExpDEPs. The βDEPs represented the estimated contribution coefficient of independent prognostic proteins in multivariate Cox regression analysis and ExpDEPs denoted the level of independent prognostic genes. Then, all patients were divided into high- or low-risk groups with the median of risk scores as the cutoff.

All data processing and statistical tests were performed using R 4.1.2 and further visualized using GraphPad Prism 6. The continuous parametric variables were displayed as mean ± standard deviation and compared using Student’s t-Test. The hazard ratios (ORs) and corresponding 95% confidence intervals (CIs) of the selected predictors of survival were also presented. The difference in survival between two groups was shown with Kaplan-Meier curves, and the receiver operating characteristic curve (ROC) for pathological grades and stages were drawn to obtain the area under the curve (AUC) values. Statistically significant P value was set at 0.05 with two sides.

The transcriptomic analysis of 5 matched RCC, VTT and NRT tissues found 1131, 1258, and 63 transcripts differentially expressed in RCC vs. NRT, VTT vs. NRT, and VTT vs. RCC groups, respectively. Among them, 856 DEGs were obtained as thrombus invasion-associated genes, of which there were 382 up-regulated and 474 down-regulated genes ( Figure 2A ). In addition, mass spectrometry analysis of urinary samples between 12 ccRCC patients and 11 healthy subjects showed 133 DEPs, with 85 up-regulated and 48 down-regulated proteins ( Figure 2B ).

The integrative analysis of transcriptomic landscape and urinary signature reveals six urinary detectable proteins (VSIG4, C3, GAL3ST1, TGFBI, AKR1C3, P4HB) displaying upregulated abundance changes consistent with corresponding genes in transcriptomic profiling ( Figure 2C ). Among them, expressions of TGFBI, AKR1C3, and P4HB increased consecutively from NRT to RCC and then to VTT, indicating that they had a consistent promoting effect in the processes of tumorigenesis and thrombus invasion ( Figure 2D ). The expressions of the targeted proteins in urine samples of ccRCC patients were over 1.5-time higher than those of healthy controls. However, only the expressions of C3, GAL3ST1, TGFBI, and P4HB achieved statistically significant difference between two groups ( Figure 2E ).

We obtained the transcriptional and follow-up data from TCGA and evaluated the correlation between expressions of targeted proteins and prognosis of ccRCC patients. First, the significant higher mRNA levels of all the six proteins in tumor compared to matched normal renal tissue were verified ( Figure 3A ; Supplementary Figure 1A ). Second, in the TCGA cohort of ccRCC patients, increased mRNA levels of VSIG4, TGFBI, P4HB were associated with higher pathological grades (all p<0.01) and later pathological stages (all p<0.05) ( Figures 3B-E ). While mRNA levels of C3, AKR1C3, GAL3ST1 were not completely correlated with tumor pathological grades and stages ( Supplementary Figures 1B-E ). Third, significant expression differences of VSIG4, TGFBI, and P4HB could be seen between patients with different OS events (366 alive vs. 173 dead). They were significantly overexpressed in patients with shorter survival and might be independent prognostic factors for ccRCC patients (all p<0.05) ( Figure 3F ). However, the expression differences of C3, AKR1C3, and GAL3ST1 were not seen in ccRCC patients with different prognosis ( Supplementary Figure 1F ).

The qRT-PCR and immunohistochemistry (IHC) experiments were respectively conducted to evaluate the mRNA and protein expression levels of VSIG4, TGFBI, and P4HB in ccRCC thrombus-tumor-normal tissue triples. The qRT-PCR analysis showed that mRNA levels of these three molecules were the highest in VTT, and then their levels in RCC were significantly higher than those in NRT ( Figure 4A ). The IHC assay further confirmed that protein expressions of VSIG4, TGFBI, and P4HB increased consecutively from normal kidney to renal tumor and then to tumor thrombus ( Figure 4B ).

The three proteins highly associated with survival (VSIG4, TGFBI, and P4HB) were used to establish a prognostic classifier ( Figure 4C ). We calculated the risk score of survival in each case from TCGA database according to expression levels of these three proteins, and then divided patients into high- or low-risk groups ( Figure 4D ). It demonstrated that ccRCC patients in high-risk group had shorter OS time (HR=2.0, p<0.01) and disease-free survival (DFS) time (HR=1.6, p<0.001) ( Figure 4E ).

The ELISA analysis was conducted in 54 urine samples from an independent cohort of ccRCC patients. As for the tumor pathological characteristics, the WHO/ISUP grade was I in two cases, II in 41 cases, III in nine cases, and IV in two cases. Urinary detectable TGFBI and P4HB, but not VSIG4, were demonstrated to be higher expressed in patients with III-IV grade tumor than those with I-II grade tumor ( Figure 4F ). The T stage was T1a in 36 cases, T1b in nine cases, T2 in three cases, and T3-4 in six cases. Urinary detectable VSIG4 and TGFBI, but not P4HB, were demonstrated to be higher expressed in patients with pathological T2-4 stage than those with pathological T1 stage ( Figure 4G ). Finally, it confirmed the satisfactory predictive power of the prognostic classifier for pathological grade (AUC=0.795, p<0.001) ( Figure 4H ) and stage (AUC=0.894, p<0.001) ( Figure 4I ) in ccRCC patients.

To determine the key roles of selected proteins in processes of tumorigenesis and thrombus invasion, we analyze the single-cell RNA-sequencing data obtained from research by Krishna et al (14). Louvain clustering revealed 31 clusters across tissues spanning lymphoid, myeloid, epithelial cells, and cancer cells based on the single-cell RNA-sequencing of 167,283 cells from multiple tumor regions, lymph node, normal kidney of ccRCC patients ( Figure 5A ). VSIG4 was indicated to be a characteristic marker for tumor-associated macrophage populations, while TGFBI and P4HB were showed to be broadly expressed in ccRCC tumor and its immune microenvironment. Furthermore, the average expression level of P4HB in ccRCC tumor and renal epithelium was the highest among 31 single-cell clusters ( Figure 5B ). After dividing single-cell transcriptomes into ccRCC tumor, adjacent normal kidney, and lymph node subgroup according to the different sources of each cell. As we can see, the macrophage-expressed VSIG4 in lymph node was higher than that in ccRCC tumor and adjacent normal kidney ( Figure 5C ), whereas the epithelium-expressed TGFBI and P4HB in ccRCC tumor were higher than those in adjacent normal kidney ( Figures 5D, E ). In addition, the GO and KEGG enrichment analyses disclosed that those selected proteins were predominantly related to the central carbon metabolism, ferroptosis, ECM-receptor interaction, and platinum drug resistance ( Supplementary Figure 2 ).