The complement system plays a complex role in cancer. In clear cell renal cell carcinoma (ccRCC), local production of complement proteins drives tumor progression, but the mechanisms by which they do this are poorly understood. We found that complement activation, as reflected by high plasma C4d or as C4d deposits at the tumor site, was associated with poor prognosis in two cohorts of patients with ccRCC. High expression of the C4-activating enzyme C1s by tumor cells was associated with poor prognosis in three cohorts. Multivariate Cox analysis revealed that the prognostic value of C1s was independent from complement deposits, suggesting the possibility of complement cascade–unrelated, protumoral functions for C1s. Silencing of C1s in cancer cell lines resulted in decreased proliferation and viability of the cells and in increased activation of T cells in in vitro cocultures. Tumors expressing high levels of C1s showed high infiltration of macrophages and T cells. Modification of the tumor cell phenotype and T-cell activation were independent of extracellular C1s levels, suggesting that C1s was acting in an intracellular, noncanonical manner. In conclusion, our data point to C1s playing a dual role in promoting ccRCC progression by triggering complement activation and by modulating the tumor cell phenotype and tumor microenvironment in a complement cascade–independent, noncanonical manner. Overexpression of C1s by tumor cells could be a new escape mechanism to promote tumor progression.

See related Spotlight by Magrini and Garlanda, p. 855.

See article by Daugan et al., p. 909 (40).

The complement system is a first line of immune defense against pathogens (1). In cancer, the complement system plays a context-dependent role, either killing tumor cells or generating local chronic inflammation, thereby promoting tumor growth (2). Analysis of the prognostic impact of intratumoral expression of complement genes has revealed the existence of tumors with “protective” or “aggressive” complement signatures, illustrating the context-dependent role of the complement system in cancer. Key genes overexpressed in the “aggressive” signature are components of the classical complement pathway encoding C1q, C1r, C1s, C4, and C3. C1q interacts with C1r and C1s to form the complement-initiating C1 complex (3). Upon recognition of immune complexes by C1q, C1s becomes activated and cleaves the components C4 and C2 to initiate the cascade (4–7). In clear cell renal cell carcinoma (ccRCC), tumor cells that produce C1s and C1r can hijack C1q generated by macrophages to initiate the classical complement pathway locally on IgG, deposited on the tumor cell surface (8). Classical pathway activation, as measured by deposits of C4 activation fragments in tumor sections, is associated with a poor prognosis in non–small cell lung cancer (NSCLC; ref. 9) and ccRCC (8), suggesting a canonical complement cascade–related mechanism. Moreover, the level of C4d, a cleavage product of C4, is elevated in the plasma of patients with NSCLC (9, 10) and malignant pleural mesothelioma (11) and can predict patient survival. Overexpression of C1s also independently predicts poor disease-specific and metastasis-free survival rates in urothelial carcinoma (12), although the mechanism responsible remains unknown.

Mouse fibroblasts, which do not produce C1s and do not form tumors when injected in vivo, gain tumorigenic capacity when transfected with the C1s gene (13). This tumorigenic potential is dependent on the enzymatic activity of C1s (14). C1s knockdown in cutaneous squamous cell carcinoma (cSCC; ref. 15) and glioblastoma (16) cell lines promotes apoptosis and significantly suppresses growth of human cSCC xenograft tumors in vivo. In contrast, C1s overexpression correlates with decreased survival in glioblastoma (16). On the basis of these results, we hypothesized the existence of protumoral, cascade-independent functions of C1s, which could be considered as noncanonical. However, the extent to which the canonical and noncanonical functions of complement proteins contribute to tumor progression is unknown.

Here, we have shown that a fraction of cancer cells in ccRCC tumors exhibited a complement-rich phenotype, including high levels of the initiating enzyme C1s. C1s overexpression by tumor cells was associated with shorter survival. The protumoral effect of C1s was found to involve three major mechanisms: (i) a canonical function of cleaving C4 and initiating the classical pathway, (ii) a function that includes noncanonical, intracellular activity that directly affects the tumor cell transcriptome, phenotype, and capacity to (iii) activate T cells. Thus, C1s can promote tumor cell proliferation and escape from immunosurveillance through modulation of the tumor microenvironment (TME).

Patients

The patients included in this study signed an informed consent form prior to inclusion. The research was approved (no. CEPAR 2014-001, CPP Ile-de-France II no. 2016-07-08) by the medical ethics board and conducted according to the recommendations in the Declaration of Helsinki. Three independent cohorts of patients with primary ccRCC undergoing nephrectomy were evaluated. The first cohort (cohort 1) was a retrospective cohort of 82 patients who underwent surgery at Necker-Enfants Malades Hospital (Paris, France) between 1999 and 2003 (17). The second cohort (cohort 2) was a prospective cohort of 26 patients for whom plasma-EDTA samples were collected at time of surgery when the tumor size was greater than 4 cm at Hospital Européen Georges Pompidou (Paris, France) since 2017. The third cohort (cohort 3) included 92 patients since 2014 whose tumors were collected at the time of surgery at Institut Mutualiste Montsouris (IMM; Paris, France). Among these patients, we assessed 58 plasma-EDTA samples. Histologic subtype of ccRCC, all tumor–node–metastasis (TNM) stages and tissue with good quality for the analysis constituted the inclusion criteria. Histopathologic features such as Fuhrman grade, TNM stages, and size were available for the majority of the patients (Supplementary Table S1). As controls for C4d quantification by ELISA, 31 plasma-EDTA samples from healthy donors were used for cohort 3 and six for cohort 2. Sections from one NSCLC tumor were available from the Departments of Pathology and Thoracic Surgery, Cochin Hospital, Paris, France.

C4d quantification in plasma

A commercial kit was used to measure C4d in plasma (Svar Life Science, catalog no. COMPL C4d RUO). Plasma-EDTA samples were stored at −80°C until use. The quantification was done according to manufacturer's instructions using the reagents of the kit. In brief, plasma samples were diluted 1:50 and transferred to microtiter wells precoated with the C4d-specific mAb for 1 hour. After washing, incubation with the secondary antibody labeled with horseradish peroxidase (HRP) was performed for 30 minutes. After washing, the substrate was added for 30 minutes, and the color was measured using an ELISA plate reader Infinite 200 PRO (Tecan). Concentrations were calculated as described by the manufacturer of COMPL C4d RUO, taking into account the 1:50 dilution.

ELISA for C1s quantification

An ELISA assay for quantification of C1s was set up in house using C1s-specific IgG, purified using Protein G resin (GE Healthcare, catalog no. 17061802) from polyclonal goat anti-human C1s antiserum (Quidel, catalog no. A302), which were left native or were biotinylated with EZ link NHS Biotin according to the manufacturer's instructions (Thermo Fisher Scientific, catalog no. 20217). Briefly, Maxisorb ELISA plates were coated with native anti-C1s IgG for 2 hours at room temperature and then blocked with PBS 5% BSA. Pure supernatants and cell lysates were added and incubated for 1 hour at room temperature. After washing with PBS 0.05% Tween 20, biotinylated anti-C1s IgG was applied for 1 hour at room temperature followed by strepavidin-HRP (Dako, catalog no. P0397). Binding was detected with TMB Microwell Peroxidase (SeraCare, catalog no. 5120-0075) and stopped after 10 minutes at room temperature with 1 mol/L sulphuric acid. The absorbance was read using a Tecan infinite M200 Pro. Purified C1s (Complement Technology, catalog no. A103) was used as a standard.

Immunofluorescence/IHC

Formalin-fixed paraffin-embedded human ccRCC or NSCLC tumors were cut into 3-μm–thick sections and stained for C1s, C1q, C4, C3, C5aR1, cytokeratin, vimentin, MHC class I, αSMA, CD31, CD68, CD163, CD66b, CD20, CD3, or CD8. The primary and secondary antibodies used, as well as the antigen retrieval and staining conditions, are summarized in Supplementary Table S2. Human liver sections (Geneticist) were used as a positive control for C1s, C4, and C3. Tonsil (Geneticist) was used as a positive control for C1q, cytokeratin, αSMA, CD31, CD163, CD20, and CD3. First, tissues were deparaffinized with three baths of Clearene (Leica Biosystems, catalog no. 3803600E) and rehydrated with an ethanol gradient (100%, 90%, 70%, 50%). Second, PT-link (Dako, PT200) was used for antigen retrieval with heat for 30 minutes at 97°C in EnVision FLEX Retrieval Solution (Dako, pH Low catalog no. K8005, pH High catalog no. K8004) at high or low pH depending of the antibody (Supplementary Table S2). A 3% H2O2 solution (Gilbert, catalog no. 6553036) was added for blocking of endogenous peroxidase, and nonspecific binding was inhibited by protein blocking (Dako, catalog no. X0909). For IHC, 3-amino-9-ethylcarbazole substrate (Vector Laboratories, catalog no. SK4200) was used. Slides were then counterstained with hematoxylin (Dako, catalog no. CS700), mounted with glycergel (Dako, catalog no. C0563) and scanned with a Nanozoomer (Hamamatsu). For immunofluorescence (IF), the nucleus was stained with DAPI; slides were mounted with Prolong Glass antifade reagent and scanned with AxioScan (Zeiss). Calopix software was used to quantify PD1, CD8, and CD163. In all cases, isotype controls were used. The whole slide was always analyzed, avoiding the bias of choosing particular areas.

Staining specificity

A competition test on liver sections was performed to verify the specificity of C1s staining (Supplementary Fig. S1A and S1B). In this test, preincubation at different molar ratios of anti-C1s (Abcam, catalog no. ab134943) and purified C1s protein for 1 hour at 37°C was performed before addition to tissues. Purified C1r (Complement Technology, catalog no. A101) served as a negative control.

Patient stratification

Among the patients in cohort 1, 82 were analyzed for C1s and 77 for C4 considering sample availability and staining specificity. The same was done for cohort 3. The tumors were classified into three groups in a semiquantitative manner according to the percentage of C1s-positive tumor cells: 0 (<5%), 1 (5%–30%), and 2 (>30%). For the classification, only the staining on tumor cells was considered. To avoid bias, the stratification was realized by three independent observers (M.V. Daugan, M. Revel, G. Lacroix, or A. Bougouin). As a result of their similar effect on survival, groups 0 and 1 were combined to form a “Low” group and compared with the “High” group, with a score of 2. The same method was applied for the classification of anti-C4d and anti-C3d staining, as described previously (8, 18). Briefly, the staining with anti-C3d and anti-C4d showed two patterns: membranous staining, identified as deposits of complement-activation fragments, and intracellular staining. The two patterns were distinguished visually by the observers and quantified separately. Independent scores for the “deposits” and for the “intracellular” staining were calculated.

Cell lines

Human lung adenocarcinoma cells (A549), human ccRCC cells (Caki-1 and A498), and human colon carcinoma cells (HCT116) were purchased from ATCC in 2016 and cultured in a humidified atmosphere of 5% CO2 at 37°C until 90% confluence. HCT116 and Caki-1 cells were cultured in McCoy's medium (Gibco, catalog no. 16600082) + 10% heat-inactivated FCS (Biowest, catalog no. S1810-500) + 1× penicillin/streptomycin (Gibco, catalog no. 10378016). A498 cells were cultured in Eagle Minimum Essential Medium (EMEM; ATCC, catalog no. 30-2003) + 10% heat-inactivated FCS + 1× penicillin/streptomycin. A549 cells were cultured in DMEM nutrient mixture F12 (Gibco, catalog no. 11320033) +10% heat-inactivated FCS + 1× penicillin/streptomycin + 1× l-glutamine (Gibco, catalog no. 25030081) + 1× HEPES (Gibco, catalog no. 15630080) + 1× nonessential amino acids solution (Gibco, catalog no. 11140050). All cell lines were routinely tested for Mycoplasma with a kit (Ozyme, catalog no. LT07-710) and used only if negative. The cells were authenticated by morphology before each experiment.

For some experiments, Caki-1 cell lines were cultured for 48 hours with or without IFNγ (PeproTech, catalog no. 300-02) at different concentration (10 or 100 ng/mL).

Silencing of C1s

A 30-minute preincubation of four siRNAs against C1s (Qiagen Hs_C1S_1_Flexitube siRNA catalog no. SI00027356; Qiagen Hs_C1S_2_Flexitube siRNA catalog no. SI00027363; Qiagen Hs_C1S_3_Flexitube siRNA catalog no. SI00027370; and Hs_C1s_5_Flexitube siRNA catalog no. SI030322873) or siRNA control (Qiagen AllStars negative control siRNA, catalog no. SI03650318) at 50 nmol/L with lipofectamine RNAimax (Thermo Fisher Scientific, catalog no. 13778030) in OptiMEM (Gibco, catalog no. 31985070) was performed before adding this mixture to cells at 50% confluence in 6-well plates. The transfection was stopped after 24 hours by replacing the medium containing siRNA with the standard medium in which the cells were routinely cultured (see “Cell lines”) without antibiotics. Photographs and functional assays were performed on the silenced cells 72 hours after transfection. The cells and supernatants were collected to assess the efficacy of silencing by qRT-PCR and Western blotting. Cells treated with siRNA against C1S were designated siC1s and the cells treated with a control siRNA against an irrelevant sequence were designated siC.

Recombinant C1s-expressing stable cell line HCT116

Bacteria transformation and plasmid amplification

The pcDNA3.1/Zeo C1s plasmid was obtained by insertion of C1s PCR amplicon of pFastBac1/C1s in pcDNA3.1/Zeo between Nhel and EcoRI. The C1s coding or empty plasmid pcDNA3.1 (Thermo Fisher Scientific, catalog no. V79520) was amplified by bacteria transformation in E. coli (Thermo Fisher Scientific, catalog no. C4040-03), cultured in LB agar, supplemented with 100 μg/mL of Ampicillin (Duchefa Biochemie, catalog no. A0104.0025), overnight at 37°C. Clones were selected and amplified in LB medium + Ampicillin. Plasmids were then extracted using ZymoPure Plasmid Maxiprep kit according to the manufacturer's protocol (Zymo Research, catalog no. D4202), the gene sequence was then verified (Eurofins Genomics).

Stable transfection in HCT116 cell line

The transfection was performed in OptiMEM using a mix containing lipofectamine 3000 (dilution 1:100, Thermo Fisher Scientific, catalog no. L3000015) and the plasmid (2.5 μg/mL, empty or with C1s gene), for 24 hours at 37°C, 5% CO2. The transfection was stopped by washing and cells were selected using EMEM with Zeocin (50 μg/mL, Thermo Fisher Scientific, catalog no. R25005). The validation of the transfection was performed by C1s quantification in the supernatant (see “ELISA for C1s quantification”).

Western blotting

C1s was evaluated by Western blotting using cell culture supernatants or cell lysates. Tumor cell lines, silenced or not for C1s, were cultured in OptiMEM for 48 hours, after which the supernatant was concentrated with Amicon Ultracel 3K units (UFC, catalog no. 900324). NuPAGE LDS sample buffer (4×, Thermo Fisher Scientific, catalog no. NP0008) was added to the concentrated supernatants or to the lysate, and the samples were denatured at 80°C for 10 minutes. Proteins were separated by NuPAGE 10% Bis-Tris gel (Thermo Fisher Scientific) electrophoresis and transferred to a nitrocellulose membrane. The membranes were blocked with TBS/5% BSA and incubated overnight on a rocking chair platform with a goat polyclonal antiserum against human C1s diluted 1:500 in TBS/5% BSA. After washing with TBS/0.1% Tween, an HRP-conjugated rabbit anti-goat IgG (Covalab, catalog no. lab0048) was added at 1:1,000 dilution for 1 hour at room temperature under agitation. The chemiluminescent signal was assessed with an iBright CL100 (Thermo Fisher Scientific) after incubation with ECL reagent (Pierce, catalog no. 32106).

In addition, a Western blot analysis was performed on normal human plasma from two healthy donors (diluted 1:10) with the C1s-specific antibody (Abcam, catalog no. ab134943) to verify that it recognizes bands corresponding to C1s and that there are no nonspecific bands. Purified C1s proenzyme and enzyme (Complement Technology, catalog nos. A103 and A104) were used as positive controls.

MHC class I was detected by Western blot analysis on lysate of tumor cells, silenced or not for C1s. NuPAGE LDS sample buffer with reducing agent (DTT, 1 mol/L) was added and the samples were heated at 80°C for 10 minutes. Proteins were separated on a NuPAGE 10% Bis-Tris gel. The proteins were transferred onto a nitrocellulose membrane using iBlot (Invitrogen). The membrane was incubated overnight with the primary antibodies (anti–MHC class I A/B/C, Abcam, catalog no. 70328) diluted 1:500 in TBS/5% BSA. After washing with TBS/1%Tween and incubation for 1 hour with HRP-conjugated goat anti-mouse IgG diluted 1:5,000 (Covalab, lab0252), visualization was performed by chemiluminescence using a substrate for HRP (SuperSignal WestDuraLuminol, Thermo Fisher Scientific, catalog no. 1856145), detected by iBright Western Blot Imaging System (iBright FL1500 Thermo Fisher Scientific).

qRT-PCR

Extraction of RNA from silenced cells or cells treated for 48 hours in medium-free serum with IFNγ (100 or 1,000 ng/mL) was performed with a Maxwell cell 16LEV simplyRNA purification kit according to the manufacturer's protocol (Promega, catalog no. AS1270). For all conditions, three replicates were included a 2100 Bioanalyzer (Agilent) and an Agilent RNA 6000 Nano Assay Kit (catalog no. 5067-1511) were used to determine the quality and quantity of the RNA. Reverse transcription was carried out with the Applied Biosystem High-capacity cDBNA reverse transcription kit according to the manufacturer's protocol (Applied Biosystem, catalog no. 4368814). Gene expression of C1S was assessed by Taqman using primers for human C1S (Thermo Fisher Scientific, catalog no. Hs00156159_m1), compared to GAPDH (Thermo Fisher Scientific, catalog no. Hs00266705_g1) and ACTB (Thermo Fisher Scientific, catalog no. Hs01060665_g1) genes for normalization.

Analysis of differentially expressed genes

For RNA sequencing (RNA-seq) analysis, three biological replicates for each condition (siC1s vs. siC) were used. The libraries were prepared with NEBNext Ultra II Directional RNA Library Prep Kit (New England BioLab, catalog no. E7760) according to the manufacturer's protocol: polyA purification using poly-T oligo attached magnetic beads, followed by 300-bp fragmentation with divalent cations under elevated temperature, synthesis of double-stranded cDNA, and amplification by PCR. Sequencing was then carried out using a paired-end 75 bp of lllumina HiSeq 4000 instrument. The quality of FASTQ files was verified by using FASTQC and then aligned to a reference genome using STAR tool. Once the data were mapped on the genome, HTSeqCount was used to count and assign reads to a given exon and generate a count table. Before differential analysis, nonexpressed genes with a count equal to 0 were removed. R software and the DESeq2 package (19) was used to examine differential expression between siC1s- and siC-transfected cells. Results were considered statistically significant at an Padjusted < 0.05. The top 50 most significantly upregulated or downregulated genes were sorted by the Padjusted; volcano plots were produced, as was a heatmap using heatmap.2 and EnhancedVolcano packages in R. Gene Ontology (GO) enrichment analysis was performed with WEV-based Gene SeT AnaLysis Toolkit (WebGestalt) with the GO biological process category. The raw data are available under the Gene Expression Omnibus accession number ID PRJNA718732.

Proliferation and mortality

Cells were washed with PBS before the addition of carboxyfluorescein diacetate succinimidyl ester (CFSE) reagent 1:1,000 (Invitrogen, catalog no. C34554) for 20 minutes at 37°C. Staining was stopped with the addition of complete culture medium. The cells were then washed and cultured in 6-well plates for 48 hours in the presence of the standard culture medium for each cell type. Cell proliferation and mortality were compared in the presence of culture medium with or without 48 hours treatment with C3a, C5a, or C1s enzyme or C1s proenzyme (each protein was used at the concentration of 500 ng/mL, protein from Complement Technology, catalog no. A118, catalog no. A144). Supernatants containing dead cells and adherent cells were recovered and stained for DAPI before analysis by flow cytometry (BD LSRFortessa; see “Flow cytometry”).

Cell morphology assessment

Silenced or control cells were cultured for 24 hours on a round cover glass in the bottom of a 24-well plate in the presence of the standard medium in which the cells were routinely cultured (see “Cell lines”). After fixation with 4% PFA for 30 minutes at room temperature, the cells were stained for the actin cytoskeleton with Phalloidin-IFluor 488 reagent (Abcam, catalog no. ab176753) and for nuclei with DAPI. The cover glass was recovered and mounted on a microscope slide with Prolong Glass antifade reagent and scanned with AxioScan (Zeiss).

Migration

A wound healing assay was performed on a cell monolayer of Caki-1, A498, or HCT116 cells to evaluate migration capacity. The monolayer was obtained when cells reached a confluence of 90%, allowing a good monolayer for the 12 next hours. A scratch was generated with a tip, and photographs were taken under a microscope immediately or after 12 hours. The wound healing tool (http://dev.mri.cnrs.fr/projects/imagej-macros/wiki/Wound_Healing_Tool) in ImageJ was used to calculate the percentage of recovery. The migration capacity of siC versus siC1s cells was compared in the presence of culture medium with or without C3a, C5a, or C1s enzyme or C1s proenzyme. The migration capacity of siC and siC1s cells was compared in presence of culture medium with or without 24 hours treatment of C3a, C5a, or C1s enzyme or C1s proenzyme (each protein was used at the concentration of 500 ng/mL).

Flow cytometry

Caki-1 cells untreated or siC or siC1s were cultured for 24 hours in the presence or absence of IFNγ (500 ng/mL). Expression of MHC class I, C3aR, and C5aR on tumor cells was studied by flow cytometry. Single-cell suspensions were resuspended in PBS + 10% heat-inactivated FCS. To block nonspecific antibody binding to Fcγ receptors, with Fc block (BD Pharmingen, catalog no. 564219). Then cells were incubated with a viability marker (Invitrogen, Live/Dead fixable yellow dead cell, catalog no. L34968), MHC class I ABC-FITC (BD Pharmingen, catalog no. 555552), C3aR-PE (BioLegend, catalog no. 345804), and C5aR (BioLegend, catalog no. 344310) for 30 minutes at 4°C. After a washing step with PBS + 10% FCS, cells were fixed with a solution of 0.4% PFA.

Flow cytometry data were acquired using a BD LSRFortessa X-20 (BD Biosciences) and analyzed using FlowJow software. Dead cells were excluded of the analysis. Surface protein expression was quantified by the median fluorescence intensity (MFI).

Coculture between tumor cells and T cells

Peripheral blood mononuclear cells (PBMC) from healthy donors were cultured in RPMI (Gibco, catalog no. 61870036) + 10% FCS + 1% penicillin and streptomycin, at 37°C, 5% CO2 in a humidified atmosphere. Single-cell suspensions were incubated with Fc block to prevent nonspecific binding to Fcγ receptors. Then PBMCs were incubated with a viability marker, CD45-APC Cy7 (clone 2D1, BD Biosciences, catalog no. 561863), CD3-AF700 (clone UCHT1, BD Biosciences, catalog no. 561027), CD4-eF450 (clone OKT4, eBioscience, catalog no. 48-0048-42), and CD8-PerCP Cy5.5 (clone RPA-T8, eBioscience, catalog no. 45-0088-42). CD4+ and CD8+ T cells were sorted from PBMCs with a purity superior to 95% on FACS Aria III cell sorter (BD Biosciences), in RPMI + 10% FCS. Dead cells were excluded of the experiment, we gated on CD45+CD3+ cells and then selected either CD4+ cells or CD8+ cells. After 24 hours of activation by CD3/CD28 beads (Dynabeads Human T-Activator CD3/CD28, Gibco, catalog no. 11161D), T cells were seeded in 96-well plates and cocultured with Caki-1 cells, untreated or siC or siC1s, at ratios 1:1 or 1 T-cell for 5 tumor cells (1:5) and in the presence or absence of plasma-purified C1s enzyme and proenzyme (1:1 mixture, concentration of 500 ng/mL for each protein). The coculture was done in RPMI + 10% FCS + 1% penicillin and streptomycin. T-cell activation was assessed in each condition by flow cytometry after 24 hours of culture using viability marker, CD45-APC, CD3-AF700, CD4-eF450, CD8-PerCP Cy5.5, CD38 FITC (clone HIT2, BD Pharmingen, catalog no. 555459), CD69-PE (clone FN50, BD Biosciences, catalog no. 555531), and PD-1-APC (clone EH12.2H7, BioLegend, catalog no. 329907). Flow cytometry was performed as described in “Flow cytometry.” We excluded dead cells from the analyses, gated on the CD45+CD3+ cells, and then either on the CD4+ or CD8+ cells to study their activation markers. Surface protein expression was quantified by MFI.

Statistical analysis

The package “survival” in R software was used to generate Kaplan–Meier curves, with censoring at 2,500 days. A log-rank test was applied to examine the difference between curves. Association between the distributions of two semiquantitative variables was assessed by the χ2 test; association between one semiquantitative variable and one quantitative variable was evaluated by the Kruskal–Wallis test and correlation between two quantitative variables by the Pearson test. The difference between siC1s and siC cells with regard to proliferation and survival was assessed by the paired t test; the unpaired t test is used for cell migration. The Mann–Whitney test was applied to assess the difference in C4d level between patients with ccRCC and healthy donors as well as patients with high or low levels of C4d deposits inside the tumor core. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Bioinformatic analysis

Bulk RNA-seq

C1S gene expression, as measured as transcripts per million (TPM), was determined by using RNA-seq data from 28 The Cancer Genome Atlas (TCGA) cohorts of patients with different cancer types available at cBioPortal (20, 21). No patients were excluded. In this database, TPM values were generated with the RSEM method (RNA-Seq by Expectation Maximization). Tumor types were arranged according to median expression values. Hepatocellular carcinoma was excluded from the analysis to avoid hiding the potential differences between tumor types linked to the very high expression of complement genes in the liver. Survival analysis was conducted with the Protein Atlas Interactive survival scatter plot and Survival analysis tool (https://www.proteinatlas.org/) using the optimal cutoff and the Gene Expression Profiling Interactive Analysis (GEPIA) tool (http://gepia.cancer-pku.cn/); the results are represented as the log2 HR of C1s for overall survival (OS), with an optimal cutoff or median cutoff calculated with the Cox pH Model, respectively. A log-rank test was applied to evaluate differences between the curves; surrounded boxes represent a log-rank P < 0.05. Because of the low number of survival events, the testicular germ cell tumor cohort was also excluded from the survival study. The KIRC cohort of TCGA was used to compare expression (log2(1+TPM)) of complement genes (C1QA, C1R, C1S, C2, C3 and C4A) among patients. Heatmaps were generated with R software 3.4.2 and the heatmap.2 package. Differential expression of C1S between tumors versus paired normal tissue from TCGA normal and GTEx normal databases (n = 100) were investigated with GEPIA. Because of the absence of normal tissue for uveal melanoma, this cancer type was not included in the analysis. Kaplan–Meier curves showing the OS of patients with high (>median) or low (<median) expression of C1S in different TCGA cohorts were generated with GEPIA. C1s expression in log2(1+TPM) in the different TNM stages of patients with ccRCC (KIRC) are represented by a violin plot generated with GEPIA. The Pearson coefficient of the correlation between C1s expression (TPM) and that of other complement genes in KIRC are represented as heatmap, as prepared using R software.

Association between C1S expression in TCGA ccRCC and immune infiltrate genes are based on immune signature of the MCP counter (22).

Single-cell RNA-seq

The single-cell RNA-seq (scRNA-seq) count matrix, genes, and barcode files were from Young and colleagues (23; Supplementary Data). A total of two biopsy samples from 1 patient with ccRCC were used in this study. Seurat (version 3.0.1; ref. 24) was employed for quality control, and single cells with less than 200 unique molecular identifiers (UMI) or with more than 25% mitochondrion-derived UMI counts were considered low-quality cells and removed. Ultimately, 3140 single cells were subjected to further investigation. The FindClusters function (resolution = 1) of Seurat was used to identify cell clusters. The clusters were visualized using two-dimensional uniform manifold approximation and projection (UMAP). The FindAllMarkers function was utilized to list the markers of each cell cluster. The major cell types were then recognized on the basis of the markers obtained from a previous study (25). The C1s+ cells are defined by a C1S expression level greater than zero.

Plasma C4d is a noninvasive, negative prognostic factor for ccRCC

In our retrospective cohort 1 (Supplementary Table S1), C4d deposits conferred poor prognosis (8). Therefore, we evaluated the level of plasma C4d. It was significantly higher in patients with ccRCC than in healthy controls in two independent prospective cohorts, cohort 2 (Fig. 1A) and cohort 3 (Fig. 1B). The patients had normal C4 levels and no elevated sC5b-9 in plasma.

Figure 1.

Complement activation is associated with poor prognosis in ccRCC. A, Plasma C4d concentration determined by ELISA in cohort 2; mean + 2SD represents the cutoff to determine a high (i.e., above the normal range) versus low (i.e., within the normal range) expression. B, Plasma C4d concentration (ng/mL) determined by ELISA in cohort 3; mean + 2SD represents the cutoff to determine a high versus low expression. C, Progression-free survival (PFS) of patients with ccRCC according to C4d plasma level with mean + 2SD cutoff for cohort 2. D, PFS of patients with ccRCC according to C4d plasma level with mean + 2SD cutoff for cohort 3. E, Association between C4d plasma level and C4d deposits within the tumor in cohort 3; Mann–Whitney test is applied to test the difference between two groups. F, PFS of patients with ccRCC of cohort 3 according to C4d deposition level within the tumor. G, IF staining of C1q (green) and C1s (red). H, IF staining of C4d (green) and C1s (red). I, PFS of patients with ccRCC according to C4d plasma level and C4d deposits within the tumor in cohort 3.

Figure 1.

Complement activation is associated with poor prognosis in ccRCC. A, Plasma C4d concentration determined by ELISA in cohort 2; mean + 2SD represents the cutoff to determine a high (i.e., above the normal range) versus low (i.e., within the normal range) expression. B, Plasma C4d concentration (ng/mL) determined by ELISA in cohort 3; mean + 2SD represents the cutoff to determine a high versus low expression. C, Progression-free survival (PFS) of patients with ccRCC according to C4d plasma level with mean + 2SD cutoff for cohort 2. D, PFS of patients with ccRCC according to C4d plasma level with mean + 2SD cutoff for cohort 3. E, Association between C4d plasma level and C4d deposits within the tumor in cohort 3; Mann–Whitney test is applied to test the difference between two groups. F, PFS of patients with ccRCC of cohort 3 according to C4d deposition level within the tumor. G, IF staining of C1q (green) and C1s (red). H, IF staining of C4d (green) and C1s (red). I, PFS of patients with ccRCC according to C4d plasma level and C4d deposits within the tumor in cohort 3.

Close modal

We observed a trend for patient survival, with progression events only occurring in the group of patients with ccRCC in cohort 2 who had plasma C4d levels above normal range (>mean + 2SD of the C4d level of healthy controls; Fig. 1C). The difference did not reach statistical significance because of the short follow-up (starting 2017) in this cohort; most patients were free from progression/death at the time of last follow-up. In cohort 3, which includes patients with longer follow-up (starting in 2014), we observed the negative prognostic impact of high C4d levels on progression-free survival (PFS) with a cutoff corresponding to the mean + 2SD of the simultaneously tested 31 healthy controls (Fig. 1D). The increased plasma C4d levels did not correlate with available clinical parameters (sex, Furman grade, or TNM stage), immune infiltration or the activation status of the infiltrating immune cells in the tumor (Supplementary Table S3). Patients that displayed local complement activation inside the tumor, as revealed by high level of C4d deposition, had a trend toward increased plasma C4d (Fig. 1E). The negative prognostic impact of the C4d deposits within the tumor was confirmed in the extended prospective cohort 3 (Fig. 1F; Supplementary Fig. S1C) of 92 patients for whom we had ccRCC tumor paraffin-embedded samples. The C4d deposition was driven mainly by the classical pathway, because we detected colocalization of C1q with C1s (Fig. 1G) and C4d with C1s (Fig. 1H) deposits at the tumor cell surface. The patients who had both plasma and in situ C4d had the poorest prognosis, although the double negative showed no progression in the studied time frame (Fig. 1I). The patients with elevated C4d in plasma but not in the tumor, had intermediate prognosis; those who had C4d only in the tumor showed no progression. Thus, plasma C4d may be a noninvasive biomarker of poor prognosis in ccRCC.

A subgroup of patients with ccRCC displays a complement-rich phenotype

Local production of complement proteins is key for intratumoral complement activation. In the ccRCC (KIRC) cohort of TCGA, approximately half of the patients displayed a complement-rich tumor environment with simultaneous expression of genes related to the classical complement pathway (C1QA, C1R, C1S, C2, C3, and C4A; nonsupervised classification; Fig. 2A). All these genes were expressed at higher levels in ccRCC tumors, compared with normal tissue, according to TCGA database (P < 10−10). Analysis of the available scRNA-seq dataset of a fresh tumor sample (Fig. 2B; Supplementary Fig. S2; ref. 23) showed that C1S, C1R, C2, C3, and some extent C1Q genes were expressed in tumor cells but C1QA, (also C1QB, C1QC) and C3 were expressed by macrophages (Fig. 2C). Individual tumor cells that expressed C1S also expressed C1R and C2 and C3 (Fig. 2D), and this was confirmed for C1s, C4, and C3 by IF (Fig. 2E and F) and IHC (Fig. 2G). C1s+ and C4+ tumor cells exhibited intermediate/low staining of cytokeratin and C1s+ tumor cells were also vimentin negative, suggesting that they are not mesenchymal cells (Supplementary Fig. S3).

Figure 2.

Complement-rich phenotype of ccRCC tumors. A, Heatmap with a column scaled and centered representing the expression (log2(1+TPM)) of complement genes (C1QA, C1R, C1S, C2, C3, and C4A) in each patient with ccRCC from the ccRCC TCGA cohort (20, 21). Two clusters of patients could be distinguished according to the expression of all the complement genes. B, UMAP representing Seurat cluster (resolution = 1, Dim 1:10) in two samples of 1 patient with ccRCC, analyzed by scRNA-seq (23). Treg, regulatory T cell. C, UMAP plot representing the expression of C1QA, C1R, C1S, C2, and C3 in cells of the patient with ccRCC, analyzed by scRNA-seq and zoom in the tumor cell clusters. In this analysis, C4A and C4B were not included because a very low number of cells expressed these genes in the tested patient. D, Expression level of C1R, C2, and C3 in tumor cells positive for C1S or negative for C1S, based on the scRNA-seq data. If C1s counts > 0, the cell is considered positive. E, IF staining of C1s (red) and C4 (green). F, IF staining of C1s (red) and C3 (green). G, Consecutive sections of a ccRCC tumor stained by IHC for C1s, C4, and C3.

Figure 2.

Complement-rich phenotype of ccRCC tumors. A, Heatmap with a column scaled and centered representing the expression (log2(1+TPM)) of complement genes (C1QA, C1R, C1S, C2, C3, and C4A) in each patient with ccRCC from the ccRCC TCGA cohort (20, 21). Two clusters of patients could be distinguished according to the expression of all the complement genes. B, UMAP representing Seurat cluster (resolution = 1, Dim 1:10) in two samples of 1 patient with ccRCC, analyzed by scRNA-seq (23). Treg, regulatory T cell. C, UMAP plot representing the expression of C1QA, C1R, C1S, C2, and C3 in cells of the patient with ccRCC, analyzed by scRNA-seq and zoom in the tumor cell clusters. In this analysis, C4A and C4B were not included because a very low number of cells expressed these genes in the tested patient. D, Expression level of C1R, C2, and C3 in tumor cells positive for C1S or negative for C1S, based on the scRNA-seq data. If C1s counts > 0, the cell is considered positive. E, IF staining of C1s (red) and C4 (green). F, IF staining of C1s (red) and C3 (green). G, Consecutive sections of a ccRCC tumor stained by IHC for C1s, C4, and C3.

Close modal

We stratified the tumors from the different cohorts according to C1s staining in tumor cells by IHC (Supplementary Fig. S4). Most (15/17) of the C1s-high tumors of cohort 1 also expressed high levels of C4 or C3, whereas only half of the C1s-low tumors expressed high levels of C4 or C3 (37/65 and 30/65, respectively). This reflects the positive association between a high density of C1s staining in tumor cells and staining for C4 (P = 0.003) and C3 (P < 0.0001; Supplementary Fig. S5A and S5B). An association between local production of C4 and C1s by tumor cells, as determined by IHC, and the presence of C4d deposition (P = 0.038 and P = 0.017, respectively) suggested that the local production was in part responsible for complement activation (Supplementary Fig. S5C and S5D).

C1s staining within the tumors was found not only in the tumor cells, but also in some stromal cells (Supplementary Fig. S3A). Nontumoral cells within the TME that expressed C1s included endothelial cells (CD31+) and rare macrophages (CD68+; Supplementary Fig. S6).

C1s predicts a poor prognosis in ccRCC

Having shown that ccRCC tumor cells produce C1s, we investigated whether it has a prognostic impact by analyzing TCGA data. C1S gene expression was increased in ccRCC tumors compared with normal kidney (Fig. 3A), especially in advanced-stage patients (Fig. 3B). C1S gene expression in patients with ccRCC from TCGA was associated with a poor prognosis (Fig. 3C). The negative prognostic value of C1s was validated at the protein level by IHC in two independent cohorts. A high number of C1s-producing cells (group 2) were associated with poor outcomes in cohort 1 (Fig. 3D and E) and cohort 3 (Fig. 3F). When patients who were in the C1s-high group were compared with those who were in the C1s-low group (groups 0 and 1), the effects on survival were still significant, with decreased survival (OS and PFS) in the high group compared with the low group (Supplementary Fig. S7A and S7B; cohort 1 PFS P = 0.00403, OS P = 0.00604, cohort 2 PFS P = 4.37e−5). Like gene expression, a correlation between TNM stage and C1s protein level was observed (Fig. 3G). C1s protein also correlated with the presence of advanced Fuhrman grade and metastasis (Fig. 3H and I; Supplementary Fig. S7C). However, in a bivariate Cox model on PFS, C1s retained this independent prognostic value with regard to TNM stage (P = 0.027) and metastasis status (P = 0.014) but not Fuhrman grade (P = 0.11). The negative prognostic impact of C1s was observed for stage I and II of the ccRCC cohort in TCGA (PFS P = 0.0084, Fig. 3J; P = 0.0074, Supplementary Fig. S7D), but not for the advanced stages III and IV (P = 0.35 and P = 0.34). The decreased PFS of C1s-high patients with stage I ccRCC was confirmed at the protein level in our cohorts 1 and 3 (PFS P = 0.0042 and P = 0.0302, respectively; Fig. 3K and L). Not enough patients were available at stage II and no significant difference was found in advanced stages (PFS: stage III P = 0.592 and stage IV P = 0.226 for cohort 1 and stage III P = 0.164 for cohort 3). No difference was found also in the JAVELIN cohort of 886 RCC patients with stages III and IV, for whom transcriptomic data is available (26). The absence of a prognostic value of C1s by Kaplan–Meier analysis in advanced-stage patients could be related to the strong association between C1s expression and TNM stage. This suggests a potential action of C1s in the transition from localized to metastatic tumors. Thus, once tumors are already in advanced stage, C1s may no longer be involved.

Figure 3.

C1s is associated with poor prognosis in ccRCC. A,C1S expression (log2TPM) in ccRCC tumors versus normal kidney from TCGA normal and GTEx database [GTEx Analysis Release V8 (dbGaP Accession phs000424.v8.p2)]. B,C1S expression (log2TPM) in different TNM stages of ccRCC tumors generated by GEPIA. C, Kaplan–Meier curves of OS according to C1S gene expression level using median cutoff on ccRCC TCGA cohort, using Protein Atlas. Kaplan–Meier curves of PFS (D) and OS (E) according to the C1s classification in three groups in cohort 1: group 0, very low % of cells (<5%); group 1, intermediate % of cells (5%–30%); and group 2, high % of cells (>30%). F, Kaplan–Meier curves of PFS according to the C1s classification in three groups in cohort 3. G, Correlation between stage and C1s in cohort 1. H, Correlation between Fuhrman grade and C1s in cohort 1. I, Correlation between metastatic status (0: negative; 1: positive) and C1s in cohort 1. J, Kaplan–Meier curves according to C1s gene expression level using optimal cutoff on stage I patients from the KIRC TCGA cohort, visualized using Protein Atlas tool. Kaplan–Meier curves of PFS according to C1s protein staining on stage I patients from cohort 1 (K) and cohort 3 (L). For the Kaplan–Meier curves, log-rank test was used and P < 0.05 was considered significant.

Figure 3.

C1s is associated with poor prognosis in ccRCC. A,C1S expression (log2TPM) in ccRCC tumors versus normal kidney from TCGA normal and GTEx database [GTEx Analysis Release V8 (dbGaP Accession phs000424.v8.p2)]. B,C1S expression (log2TPM) in different TNM stages of ccRCC tumors generated by GEPIA. C, Kaplan–Meier curves of OS according to C1S gene expression level using median cutoff on ccRCC TCGA cohort, using Protein Atlas. Kaplan–Meier curves of PFS (D) and OS (E) according to the C1s classification in three groups in cohort 1: group 0, very low % of cells (<5%); group 1, intermediate % of cells (5%–30%); and group 2, high % of cells (>30%). F, Kaplan–Meier curves of PFS according to the C1s classification in three groups in cohort 3. G, Correlation between stage and C1s in cohort 1. H, Correlation between Fuhrman grade and C1s in cohort 1. I, Correlation between metastatic status (0: negative; 1: positive) and C1s in cohort 1. J, Kaplan–Meier curves according to C1s gene expression level using optimal cutoff on stage I patients from the KIRC TCGA cohort, visualized using Protein Atlas tool. Kaplan–Meier curves of PFS according to C1s protein staining on stage I patients from cohort 1 (K) and cohort 3 (L). For the Kaplan–Meier curves, log-rank test was used and P < 0.05 was considered significant.

Close modal

Canonical mode of action of C1s in ccRCC tumors

The most evident mechanism to explain the negative prognostic impact of C1s would be its canonical functions in the complement cascade. When the patients were stratified at optimal cutoff and using fragments per kilobase million (FPKM; as in Protein Atlas), C5AR1 conferred poor prognosis in ccRCC in TCGA cohort, whereas C3AR was associated with a favorable outcome (Fig. 4A). This significance was lost when the analysis was made with TPM data normalization and therefore should be interpreted with caution. ccRCC tumors are rich in C3AR-expressing macrophages and C5AR1-expressing neutrophils and macrophages, making them responsive to the generated C3a and C5a in the TME (Fig. 4B).

Figure 4.

C3aR and C5aR1 may impact survival in ccRCC by action on macrophage infiltration or tumor cells. A,C3AR and C5AR1 gene expression impacts survival in the ccRCC TCGA cohort, at optimal cutoff. B, scRNA-seq shows that macrophages and neutrophils as well as some rare tumor cells express C3AR and C5AR1. C, IF staining of ccRCC tumors (CD163 in green, C5aR1 in red) D, Density of CD163+ cells in each group of C1s classification determined by IHC in cohort 1. E, Kaplan–Meier curve of PFS according to the combination of C1s and CD163 high/low classification in cohort 1. F, IHC for C5aR1 on human tonsil (a positive control for immune cell staining) and ccRCC tumors (shown are two representative ccRCC tumors from n = 11). Immune infiltrating cells were detected in every tumor, but C5aR1-positive tumor cells were found only in 1 of 11 tumors. Scale bars (left to right), 250 μm, 100 μm, 100 μm. G, Caki-1 cells express low levels of C3aR and C5aR1, as detected by flow cytometry. WT, wild-type. H–J, Incubation of Caki-1 with purified C3a or C5a does not impact viability of the cells, detected by flow cytometry (error bars represent mean with SD; H); morphology, detected by actin staining, revealed by immunocytofluorescence (I); or proliferation, detected by flow cytometry (error bars represent mean with SD; J). I, Scale bars, 100 μm. K and L, C3a but not C5a impacts the migration of Caki-1 cells, as evaluated by a scratch test; error bars represent mean with SD.

Figure 4.

C3aR and C5aR1 may impact survival in ccRCC by action on macrophage infiltration or tumor cells. A,C3AR and C5AR1 gene expression impacts survival in the ccRCC TCGA cohort, at optimal cutoff. B, scRNA-seq shows that macrophages and neutrophils as well as some rare tumor cells express C3AR and C5AR1. C, IF staining of ccRCC tumors (CD163 in green, C5aR1 in red) D, Density of CD163+ cells in each group of C1s classification determined by IHC in cohort 1. E, Kaplan–Meier curve of PFS according to the combination of C1s and CD163 high/low classification in cohort 1. F, IHC for C5aR1 on human tonsil (a positive control for immune cell staining) and ccRCC tumors (shown are two representative ccRCC tumors from n = 11). Immune infiltrating cells were detected in every tumor, but C5aR1-positive tumor cells were found only in 1 of 11 tumors. Scale bars (left to right), 250 μm, 100 μm, 100 μm. G, Caki-1 cells express low levels of C3aR and C5aR1, as detected by flow cytometry. WT, wild-type. H–J, Incubation of Caki-1 with purified C3a or C5a does not impact viability of the cells, detected by flow cytometry (error bars represent mean with SD; H); morphology, detected by actin staining, revealed by immunocytofluorescence (I); or proliferation, detected by flow cytometry (error bars represent mean with SD; J). I, Scale bars, 100 μm. K and L, C3a but not C5a impacts the migration of Caki-1 cells, as evaluated by a scratch test; error bars represent mean with SD.

Close modal

Indeed, a fraction of the macrophages and neutrophils (Fig. 4C), but to a lesser extent T cells but not B cells (Supplementary Fig. S8A) were positive for C5aR1. C1S expression in tumors correlated with macrophage infiltration in TCGA ccRCC cohort (R = 0.45, P = 2.2 × 10−16; Supplementary Fig. S8B) and at the protein level (CD163+) in cohort 1, as measured by IHC (Fig. 4D). A composite score reflecting expression of C1s and infiltration by CD163+ macrophages revealed a group of high/high patients with a particularly poor outcome (Fig. 4E). In TCGA, no correlation was found between C1S and neutrophils (R = 0.1; P = 0.0335; Supplementary Fig. S8C).

Other nontumor cells, which express little if any C3aR1 or C5aR1, also correlated with C1S. A correlation was found with the B cells (R = 0.41; P < 2.2 × 10−16), T cells (R = 0.53; P < 2.2 × 10−16), and the fibroblasts (R = 0.48; P < 2.2 × 10−16), while R was below 0.3 for the natural killer cells (R = 0.21; P = 2.843 × 10−6) and endothelial cells (R = −0.01; P = 0.03235; Supplementary Fig. S8D–S8H).

Very few tumor cells expressed C3aR1 or C5aR1 (Fig. 4B). Only 1 of 11 tested tumor samples showed low levels of C5aR1 on tumor cells, although infiltrating stromal cells positive for C5aR1 were abundant (Fig. 4F). The ccRCC cell lines Caki-1 and A498 expressed low levels of C3aR and C5aR1 by flow cytometry (Fig. 4G for Caki-1) and qRT-PCR. Caki-1 and A498 cells did not generate detectable concentrations of C5a (detection limit 0.01 ng/mL), whereas C3a was at the detection limit of the kit, at 0.015 ng/mL in our experimental conditions. This makes an autocrine effect of C5a less likely. Addition of purified C3a or C5a did not modify the viability, morphology, or proliferation of resting Caki-1 cells (Fig. 4HJ). Silencing of C1s also did not make the Caki-1 cells proliferation sensitive to C3a or C5a. C3a did reduce the migration capacity of Caki-1 cells, but C5a had no effect on this (Fig. 4K and L). Taken together, these results suggest that C3a, but not C5a, may have an effect on tumor cells. The generation of C3a and C5a by the canonical role of C1s may result in remodeling of the TME.

Cox multivariate analysis in cohort 1 highlighted that the prognostic value of C1s was independent of complement activation markers [C1q deposits (P = 0.00148), C4d deposits (P = 0.00741), and C3d deposits (P = 0.00488)], suggesting a potential cascade-independent role for C1s.

C1s modulates the transcriptional program of ccRCC tumor cells

Considering that the prognostic value of C1s may be independent of complement activation, we investigated C1s effects on tumor cells. At the single-cell level [in a dataset from a fresh ccRCC human tumor (23)], we searched for genes that discriminated tumor cells expressing C1S from those that did not (Fig. 5A). GO analysis showed that genes enriched in C1s+ tumor cells are implicated in various pathways such as vesicle-mediated transport, response to oxygen-containing compounds, response to hormones, myeloid activation, neutrophil activation and degranulation, response to endoplasmic reticulum stress, extracellular matrix organization, complement proteins, regulation of IGF transport and uptake, and folding assembly and peptide loading of MHC class I (Fig. 5B and C). To study the direct effects of C1s on tumor cells, we performed in vitro studies on Caki-1 and A498 cells with or without C1S silencing. Both cell lines produced and secreted the proenzyme and enzyme forms of C1s (Supplementary Fig. S9A and S9B). After validation of the efficacy of C1S silencing in Caki-1 and A498 cells by qRT-PCR and Western blotting (Supplementary Fig. S9C and S9D and S9E and S9F, respectively), we performed RNA-seq for Caki-1 (Fig. 5D) and A498 (Supplementary Fig. S10A) cells. Among all genes, we identified 56 downregulated genes and 119 upregulated genes in Caki-1 siC1s versus siC cells and 41 downregulated genes and 112 upregulated genes in A498 siC1s versus siC cells by considering a fold change >2 in absolute value and an Padjusted <0,05 (Fig. 5E; Supplementary Fig. S10B). Enrichment analysis with GO:biological process and Reactome highlighted that these genes are involved in different pathways, some previously highlighted by the single-cell analysis (Fig. 5F and G; Supplementary Fig. S10C and S10D). Among the upregulated pathways in siC1s Caki-1 cells, we found regulation of development and growth, regulation of chemotaxis, and extracellular matrix organization. Downregulated genes were implicated in extracellular matrix organization, IGF transport and uptake, and endosomal/vacuolar pathway. For A498 cells, we observed upregulation of genes involved in regulation of motility, extracellular matrix organization, and cellular response to erythropoietin as well as downregulation of genes involved in endocytosis, vesicle-mediated transport, angiogenesis, and transcriptional regulation by RUNX1.

Figure 5.

Modulation of tumor cell gene expression by C1s. A, Heatmap representing the most variable genes between C1S+ tumor cells and C1S tumor cells. Differential genes are determined using FindAllmarker function from Seurat package in R. Genes with a log2 fold change > 0.5, min.pct = 0.25, min.diff.pct = 0.25, and a Padjusted < 0.05 are shown. Red represents high expression and blue low expression. The genes are ordered according to the adjusted P value. B, GO analysis of functional annotations for upregulated or downregulated genes in C1S+ tumor cells in comparison with C1S tumor cells. All genes with min.pct = 0.25 and Padjusted < 0.05 are included in the analysis C, Reactome analysis of upregulated or downregulated genes in C1S+ tumor cells in comparison with C1S tumor cells. D, Top 50 differentially expressed genes after silencing with siC1s compared with siC in Caki-1 cells. The most differentially expressed genes (log2 counts) were selected on adjusted P value. Expression across each gene has been row scaled, with relatively high expression of a given gene marked in red and samples with relatively low expression marked in blue. Lighter shades and white represent genes with intermediate expression levels. E, Volcano plot depicting the results of differential expression obtained with DESeq2 analysis between Caki-1 siC1s (n = 3) and Caki-1 siC (n = 3). The volcano plot was generated with R software and EnhancedVolcano package. Red points represent genes with a log2 fold change >1 and Padjusted < 0.05. F, GO analysis of functional annotations for upregulated or downregulated genes by C1s silencing versus Caki-1 siC. G, Reactome analysis of upregulated or downregulated genes by C1s silencing versus Caki-1 siC.

Figure 5.

Modulation of tumor cell gene expression by C1s. A, Heatmap representing the most variable genes between C1S+ tumor cells and C1S tumor cells. Differential genes are determined using FindAllmarker function from Seurat package in R. Genes with a log2 fold change > 0.5, min.pct = 0.25, min.diff.pct = 0.25, and a Padjusted < 0.05 are shown. Red represents high expression and blue low expression. The genes are ordered according to the adjusted P value. B, GO analysis of functional annotations for upregulated or downregulated genes in C1S+ tumor cells in comparison with C1S tumor cells. All genes with min.pct = 0.25 and Padjusted < 0.05 are included in the analysis C, Reactome analysis of upregulated or downregulated genes in C1S+ tumor cells in comparison with C1S tumor cells. D, Top 50 differentially expressed genes after silencing with siC1s compared with siC in Caki-1 cells. The most differentially expressed genes (log2 counts) were selected on adjusted P value. Expression across each gene has been row scaled, with relatively high expression of a given gene marked in red and samples with relatively low expression marked in blue. Lighter shades and white represent genes with intermediate expression levels. E, Volcano plot depicting the results of differential expression obtained with DESeq2 analysis between Caki-1 siC1s (n = 3) and Caki-1 siC (n = 3). The volcano plot was generated with R software and EnhancedVolcano package. Red points represent genes with a log2 fold change >1 and Padjusted < 0.05. F, GO analysis of functional annotations for upregulated or downregulated genes by C1s silencing versus Caki-1 siC. G, Reactome analysis of upregulated or downregulated genes by C1s silencing versus Caki-1 siC.

Close modal

C1s modulates the proliferation and survival of renal and lung tumor cells

In culture, we observed visual phenotypic changes in siC1s cells, with a longer time to reach confluence. Functionally, silencing of C1S in Caki-1 and A498 cells induced a decrease in viability (Fig. 6A; Supplementary S10E) and proliferation capacity (Fig. 6B and C; Supplementary S10F and S10G) but no effect on cell morphology (Fig. 6D; Supplementary S10H) or migration capacity (Supplementary Fig. S10I and S10J).

Figure 6.

C1s expression modulates proliferation and viability of Caki-1 cells. A, Viability assessment by the percentage of DAPI-positive dead cells, determined by flow cytometry after Caki-1 siC1s or Caki-1 siC were cultured for 72 hours. Data shown are from 11 independent experiments; error bars represent mean with SD. B, Histogram representing CFSE intensity in Caki-1 siC1s and Caki-1 siC after 72 hours of proliferation (gray = unstained Caki-1 cells; orange = Caki-1, taken at the time of CFSE staining T0; red = Caki-1 siC after 72 hours of proliferation; and blue = Caki-1 siC1s after 72 hours of proliferation). C, Fold change of CFSE MFI in Caki-1 siC1s versus Caki-1 siC in four independent experiments. D, IF staining of phalloidin (white) in Caki-1 siC1s and Caki siC. E, Fold change of CFSE mean fluorescence (MFI) in Caki-1 siC versus siC1s versus siC1s with addition of purified protein C1s enzyme (enz) or proenzyme (proenz).

Figure 6.

C1s expression modulates proliferation and viability of Caki-1 cells. A, Viability assessment by the percentage of DAPI-positive dead cells, determined by flow cytometry after Caki-1 siC1s or Caki-1 siC were cultured for 72 hours. Data shown are from 11 independent experiments; error bars represent mean with SD. B, Histogram representing CFSE intensity in Caki-1 siC1s and Caki-1 siC after 72 hours of proliferation (gray = unstained Caki-1 cells; orange = Caki-1, taken at the time of CFSE staining T0; red = Caki-1 siC after 72 hours of proliferation; and blue = Caki-1 siC1s after 72 hours of proliferation). C, Fold change of CFSE MFI in Caki-1 siC1s versus Caki-1 siC in four independent experiments. D, IF staining of phalloidin (white) in Caki-1 siC1s and Caki siC. E, Fold change of CFSE mean fluorescence (MFI) in Caki-1 siC versus siC1s versus siC1s with addition of purified protein C1s enzyme (enz) or proenzyme (proenz).

Close modal

Expression of C1s by tumor cells was not limited to ccRCC. Analysis of C1S expression in 28 cohorts of the PanCancer study of TCGA (27) revealed heterogeneous expression according to tumor type (Supplementary Fig. S11A). The almost 100-fold higher expression of C1S in pancreatic adenocarcinoma and sarcoma than in kidney chromophobe represents the two extremes. NSCLC subtypes lung adenocarcinoma and lung squamous cell carcinoma were among the cancer types in which C1S was strongly expressed. We also found that a fraction of the cells (defined by cytokeratin staining) in an NSCLC tumor were positive for C1s protein (Supplementary Fig. S11B and S11C). Moreover, silencing of C1S affected the phenotype of the lung cancer cell line A549 in similar ways to that seen for the ccRCC cell lines (Supplementary Fig. S11D and S11E).

C1s in the extracellular milieu does not impact the phenotype of cancer cells

To find out whether the phenotypic alterations of the siC1s Caki-1 cells could be mediated by the absence of secreted C1s, we added purified C1s proenzyme or enzyme in the supernatant. No effect on the proliferation and viability was detected (Fig. 6E). Furthermore, we attempted to “transplant” the phenotype of enhanced proliferation to a naturally C1s-deficient cell line, HCT116, by transfecting it with a C1S-expressing plasmid (Supplementary Fig. S12). The recombinant C1s was efficiently expressed, but it was entirely released in the supernatant, without intracellular retention (Supplementary Fig. S12A and S12B). Two C1s-expressing clones were selected and both showed no difference in proliferation and migration compared with untreated cells or cells transfected with empty vector (Supplementary Fig. S12C–S12E).

C1s interplay with T cells

C1s expression in fibroblasts is induced by IFNγ (28), produced by tumor-infiltrating T cells. We found that IFNγ stimulated C1S expression in ccRCC tumor cell lines in vitro (Fig. 7A; Supplementary Fig. S13). Expression of IFNG and C1S correlated positively in the ccRCC cohort of TCGA (R = 0.26, P = 1.5 × 10−9). One of the main target genes of IFNγ in tumor encodes MHC class I. C1s was suggested to cleave MHC class I but the C1S silencing did not modify the protein integrity of MHC class I, as detected by Western blot analysis on Caki-1 lysates (Supplementary Fig. S14A). Exposure of Caki-1 cells to IFNγ resulted in upregulation of MHC class I, as detected by flow cytometry (Fig. 7B and C; Supplementary Fig. S14B and S14C). The siC1s Caki-1 showed a trend toward upregulation of MHC class I at basal level (Fig. 7B). This increase is also observed upon IFNγ stimulation, but a clear alteration of the expression pattern was detected, with robust appearance of two populations with different granularity (Fig. 7C; Supplementary Fig. S14C), which requires further study.

Figure 7.

C1s production by tumor cells impacts CD8 T cells. A, C1s secretion in the supernatant after treatment of Caki-1 cells with IFNγ (0, 100, and 1,000 ng/mL), detected by ELISA; error bars represent mean with SD. Expression of MHC class I A/B/C, as assessed by flow cytometry, on resting (B) and IFNγ-activated (C) Caki-1 cells, untreated, siC, or siC1s. D, Kaplan–Meier curve of PFS, with patients in cohort 1 stratified by CD8+ T-cell density. E, Correlation between the expression of C1S (log2TPM), separated in quartiles, and CD8 (log2TPM) from TCGA cohort; the T-cell signature was identified by the MCP-counter tool (22). F, Density of CD8+ cells in each group of C1s classification determined by IHC, cohort 1. G, Density of PD-1+ cells in each group of C1s classification determined by IHC, cohort 1. H–M, Coculture of activated sorted CD4+ (HJ) and CD8+ (KM) T cells with untreated, siC or siC1s Caki 1 cells, error bars represent mean with SD. H and K, % of CD38+ cells among CD4+ (H) and CD8+ (K) T cells cultured for 24 hours alone or in the presence of siC or siC1s Caki-1 cells at a ratio of five tumor cells to one T cell (5:1). I and L, Percentage of CD38+ T cells after 24 hours of coculture, with varying tumor cell–to–T cell ratios. J and M, Percentage of CD38+ T cells [CD4 (J) or CD8+ (M) T cells] after 24 hours of coculture, with addition of purified C1s enzyme + proenzyme in the supernatant of the coculture of the siC1s (siC1s+C1s condition) cells. Albumin (Alb) was used as an irrelevant protein control. N, Proposed mechanism linking the complement-rich phenotype of the ccRCC cells to cancer progression. At least three nonexclusive processes are suggested: (i) activation of the canonical complement cascade by tumor cell–derived C1r, C1s, C3, C4, and C5 and macrophage-produced C1q. C3a could reduce tumor cell migration, whereas C5a attracts and activates macrophages. C4d deposits at the tumor cell surface and plasma C4d, as well as macrophage infiltration and C5aR1 expression, confer poor prognosis. (ii) C1s acts in a noncanonical manner inside the tumor cell, modifying its transcription program and phenotype. It promotes proliferation and survival. (iii) C1s-producing tumor cells have a complex interplay with infiltrating T cells. T cell–derived IFNγ stimulates the production of C1s. The C1s-expressing tumor cells fail to activate CD4+ and CD8+ T cells. In contrast, C1s depletion in tumor cells results in aberrant expression of MHC class I and alleviates the inability of tumor cells to activate CD8+ T cells. The figure was generated with BioRender.com.

Figure 7.

C1s production by tumor cells impacts CD8 T cells. A, C1s secretion in the supernatant after treatment of Caki-1 cells with IFNγ (0, 100, and 1,000 ng/mL), detected by ELISA; error bars represent mean with SD. Expression of MHC class I A/B/C, as assessed by flow cytometry, on resting (B) and IFNγ-activated (C) Caki-1 cells, untreated, siC, or siC1s. D, Kaplan–Meier curve of PFS, with patients in cohort 1 stratified by CD8+ T-cell density. E, Correlation between the expression of C1S (log2TPM), separated in quartiles, and CD8 (log2TPM) from TCGA cohort; the T-cell signature was identified by the MCP-counter tool (22). F, Density of CD8+ cells in each group of C1s classification determined by IHC, cohort 1. G, Density of PD-1+ cells in each group of C1s classification determined by IHC, cohort 1. H–M, Coculture of activated sorted CD4+ (HJ) and CD8+ (KM) T cells with untreated, siC or siC1s Caki 1 cells, error bars represent mean with SD. H and K, % of CD38+ cells among CD4+ (H) and CD8+ (K) T cells cultured for 24 hours alone or in the presence of siC or siC1s Caki-1 cells at a ratio of five tumor cells to one T cell (5:1). I and L, Percentage of CD38+ T cells after 24 hours of coculture, with varying tumor cell–to–T cell ratios. J and M, Percentage of CD38+ T cells [CD4 (J) or CD8+ (M) T cells] after 24 hours of coculture, with addition of purified C1s enzyme + proenzyme in the supernatant of the coculture of the siC1s (siC1s+C1s condition) cells. Albumin (Alb) was used as an irrelevant protein control. N, Proposed mechanism linking the complement-rich phenotype of the ccRCC cells to cancer progression. At least three nonexclusive processes are suggested: (i) activation of the canonical complement cascade by tumor cell–derived C1r, C1s, C3, C4, and C5 and macrophage-produced C1q. C3a could reduce tumor cell migration, whereas C5a attracts and activates macrophages. C4d deposits at the tumor cell surface and plasma C4d, as well as macrophage infiltration and C5aR1 expression, confer poor prognosis. (ii) C1s acts in a noncanonical manner inside the tumor cell, modifying its transcription program and phenotype. It promotes proliferation and survival. (iii) C1s-producing tumor cells have a complex interplay with infiltrating T cells. T cell–derived IFNγ stimulates the production of C1s. The C1s-expressing tumor cells fail to activate CD4+ and CD8+ T cells. In contrast, C1s depletion in tumor cells results in aberrant expression of MHC class I and alleviates the inability of tumor cells to activate CD8+ T cells. The figure was generated with BioRender.com.

Close modal

This difference in surface expression of MHC class I was not shared by other surface markers, because C3aR and C5aR1, detected in triple staining on the same cells, did not show any alteration (Fig. 4G). No significant differences in the patterns of MHC class I staining on C1s-high or C1s-low ccRCC tumors was observed (Supplementary Fig. S14D).

T cells are a major source of IFNγ in the TME (29, 30). In ccRCC, the high CD8+ T-cell density is associated with poor prognosis (17, 31), as confirmed in cohort 1 (Fig. 7D). A previously defined CD8+ T-cell signature (22) correlated of C1S with CD8+ T cells (R = 0.42, P = < 2.2e-16; Fig. 7E). The expression of C1S correlated with CD4 (R = 0.35, P = 1.5 × 10−16). Furthermore, the number of C1s+ cells within the tumor correlated with increased CD8+ T-cell infiltration (Fig. 7F) and PD1 expression (Fig. 7G) as detected by IHC in cohort 1. In this cohort, the high density of CD8+ T cells was associated with poor prognosis (PFS: P = 0.0363; OS: P = 0.0192). For 34 patients from cohorts 2 and 3 for whom RNA-seq data were available, C1S also correlated with IFNG (R = 0.366, P = 0.0281) and showed a trend for CD8B (R = 0.31, P = 0.0694).

To find out whether tumor cell–derived C1s could impact the T cells, a coculture was set up between untreated, siC or siC1s Caki-1 cells and resting or activated sorted CD4+ and CD8+ T cells. Incubation with siC1s cells induced further activation of prestimulated T cells, measured by increase in the percentage of CD38+ CD4+ and CD8+ T cells (Fig. 7H and K), without modification of the percentage of CD69+ or PD-1+ (Supplementary Fig. S15). These effects were dose dependent and increased with the proportion of tumor cells in the coculture (0:1; 1:1, and 5:1 tumor cell–to–T cell ratio; Fig. 7I and L). Purified C1s enzyme + proenzyme had no effect on the activation status of the T cells, compared with albumin and could not revert the phenotype induced by the silencing of C1s (Fig. 7J and M). Prestimulated T cells exposed to purified C1s, C3a, or C5a did not show alteration of the expression of CD38, CD69, or PD1. Since the experiments were not performed with antigen-specific T cells, the observed increase in T-cell activation may be due to a different expression of costimulating molecules or soluble mediators by the tumor cells after C1s silencing. These data suggest that the effect is mediated by intrinsic (intracellular) C1s.

No correlation between immune markers and C4d deposits (P = 0.22 for CD8, P = 0.2 for PD1 and P = 0.48 for CD163), plasma C4d (see Supplementary Table S3), or C3d deposits (P = 0.9 for CD8, P = 0.35 for PD-1 and P = 0.56 for CD163) was found. This also suggests another, potentially complement-independent role for C1s in the modulation of the TME.

Here, we describe a complement-rich phenotype for ccRCC in which overexpression of C1s confers a poor prognosis in early-stage disease. The associated protumor impact is driven in part by canonical complement system–related inflammation, as reflected by an elevation in plasma C4d, which has the potential to be a noninvasive, negative prognostic biomarker for ccRCC. Furthermore, we discovered noncanonical, cascade-unrelated, intracellular functions for C1s that affect cell transcriptional activity, proliferation, viability, and cross-talk with T cells. The joint action of these mechanisms might explain the negative prognostic impact of C1s in ccRCC (Fig. 7N), which appears to be a prototypical complement-dependent cancer.

ccRCC is a cancer type in which concomitant overexpression of complement genes confers poor prognosis (2). Although C1q originates from macrophages, the remaining proteins are produced by tumor cells. The production of complement proteins within the tumor is heterogeneous, but we found that in most cases, the same tumor cell produces C1s, C1r, C2, C4, and C3; we termed this a complement-rich phenotype. These data support the notion that in a subset of ccRCC tumors, the classical pathway components are activated locally (32, 33).

In searching for reliable, noninvasive complement-related biomarkers that could predict patient outcome, we found that plasma C4d was significantly increased in the plasma of patients with ccRCC and conferred a poor prognosis. This result, together with similar findings in lung cancer (10) and malignant pleural mesothelioma (11), make plasma C4d a novel promising parameter for evaluating the prognosis of patients with cancer. Such a rapid screening test has already been used in some patients to predict recurrence of lupus nephritis (34). Upon further validation, it might be possible for expert complement laboratories to introduce such a test into the routine screening of patients with ccRCC and lung cancer. The origin of plasma C4d requires further investigation. No hypocomplementemia was present in any of the patients evaluated, suggesting no consumption of C4 in the circulation. C4d can be generated via the classical or the lectin pathway, but both require a multivalent ligand on a surface for activation (4). Therefore, plasma C4d likely derives from the tumor. Consistent with this, we detected a trend toward an association between C4d deposits in the tumor and in the plasma. Thus, plasma C4d might arise due to unsuccessful binding of C4b to tumor cells or proteolytic shedding of the C4 bound target and subsequent inactivation by factor I or due to release as a part of membrane-bound C4d through microvesicles. Such a process of C4d elimination via microvesicles is described in renal transplantation (35) but remains to be examined in cancer.

Intratumor complement activation allows the release of C5a, generating cancer-promoting signals through C5aR1 (2, 8, 32, 35–37). At optimal cutoff, C5AR1 gene expression was associated with poor prognosis, but it is difficult at this stage to conclude which cells this negative prognostic impact is due to, that is, immune cells and/or tumor cells. This is due to doubts on the specificity of the S5/1 and P12/1 C5aR1-specific antibody clones (38), which were used to detect C5aR1 in renal cancer (33, 36, 39). As in ref. 38, we used another clone, 8D6, and found few tumor cells positive for C5aR1 by IHC in agreement with the scRNA-seq. The ccRCC tumor cell lines tested here did not generate detectable C5a in the supernatant, arguing against an autocrine effect in the absence of canonical complement activation. Although Caki-1 cells expressed low level of C5aR1 on the surface, there were no changes in morphology, proliferation, or migration in response to C5a. The most likely effect of C5a is, therefore, on the immune infiltrate, attracting and activating tumor-associated macrophages and neutrophils.

C3AR gene expression at optimal cutoff was associated with favorable outcome in ccRCC. We detected C3aR on the tumor cell lines, and Caki-1 cells had reduced migration capacity in response to C3a. Expression of the C3 gene and the intracellular presence of C3 protein conferred poor prognosis in ccRCC, but deposits of C3d did not (8). The intracellular factor H staining partially colocalized with C3 and also conferred poor prognosis in ccRCC (40). Together, these results hint to a peculiar, dual action for C3 in ccRCC, a protective one via C3a and a deleterious one, potentially via progression of the cascade to C5a (without formation of C5b-9) and likely by a novel, intracellular mode of action, similar to what is already described in T cells (41) and pancreatic islets (42, 43), which may be in part linked to factor H.

In this study, we describe an intracellular, noncanonical mode of action for C1s in tumor cells. Multivariate Cox analysis revealed the negative prognostic value of C1s to be independent of other complement activation markers such as C1q, and C4d and C3d deposits, highlighting a cascade-independent role for C1s. C1S silencing altered the transcriptional program of ccRCC cells and reduced cell proliferation and viability. This phenotype was independent of VHL mutation, as it was observed in Caki-1 cells, which are VHL wild-type and A498 cells, which are VHL mutated. Noncanonical functions for C1s have also been reported in cSCC, although the genes differentially expressed after C1S silencing were different (15). Moreover, reduced proliferation and viability were observed in the lung adenocarcinoma cell line A549, suggesting the general relevance of C1s as a proliferation- and cell survival–promoting factor.

Some studies suggest that C1s can cleave substrates other than C4 (44–51), forming a C1s substratome. Therefore, we hypothesized that upon secretion, C1s can cleave a specific protein(s) from the extracellular space or at the cell surface that will allow an increase in proliferation or survival. C1s was reported to degrade components of the extracellular matrix such as decorin, type 1 gelatine, and type 2 collagens (50), MHC class I (52) and insulin-like growth factor binding protein 5 (IGFBP5; ref. 47). We did not detect MHC class I cleavage, collagen degradation, or IGFR signaling. Moreover, addition of purified C1s proenzyme or enzyme in the culture medium failed to prevent the effect of the C1S gene silencing on proliferation, survival, and T-cell activation. Secretion of recombinant C1s from HCT116 cells also had no measurable effect on their phenotype. Therefore, the extracellular activity of C1s does not seem to play a detectable role in our experimental conditions and cannot explain the observed phenotype.

The subcellular localization of C1s in the cytoplasm hints at potential mechanisms by which it can affect fundamental cellular functions. C1s interacts and cleaves HMGB-1 (44, 53), which shuttles between the cytoplasm and the nucleus (54). In the nucleus, HMGB1 interacts with nucleosomes, transcription factors, and histones, regulating transcription. Therefore, C1s may affect transcription via its interaction/cleavage of HMGB-1. This may explain the link between C1s expression and the enrichment in genes related to extracellular matrix organization, regulation of IGF transport and uptake by IGF binding proteins or vesicular transport, as evidenced by scRNA-seq of a patient with ccRCC and RNA-seq of siC1s cells.

Intracellular cleavage of C3 has a prosurvival role in T cells (41). C1s is coexpressed with other complement proteins, such as C1r, C2, C3, and C4 in the context of the complement-rich phenotype of the tumor cells. Therefore, intracellular C1s may get activated and exert enzymatic activity toward its canonical substrates but in a noncanonical location. Therefore, the activity of C1s may result in functions, similar to the ones described for C3 in T cells (55).

A fraction of the C1s-high tumors is rich in T cells, as seen here and in a recent unsupervised transcriptomic analysis 823 patients with RCC (56), suggesting an interplay between these two factors. Proinflammatory cytokines such as IFNγ produced by T cells stimulate expression of C1S and other complement genes in many cells (28, 57, 58), including the ccRCC cell lines used in our study. CD8+ T cells and IFNγ confer poor prognosis in ccRCC (17, 31). The increased CD4+ and CD8+ T-cell activation occurred without concomitant PD1 increase when in contact with C1s-silenced tumor cells. These data suggest a bidirectional cross-talk in C1s-high/T cell–high tumors whereby infiltrating T cells stimulate C1s via IFNγ and C1s-high tumor cells prevent T-cell activation and likely promote exhaustion, as the amount of PD1+ cells also correlated with C1s in situ. This cross-talk may be related to the phenotype of the tumor cells (expressing C1s or not) and not to the C1s in the surrounding milieu, again pointing toward an intracellular, noncanonical mode of action for C1s.

The action of C1s is context dependent. C1s is associated with deleterious clinicopathologic parameters in urothelial carcinoma (12) and cSCC (15), similar to our findings for ccRCC. Nevertheless, C1s overexpression in sarcomas confers a favorable outcome (2). Therefore, the mechanisms of action of complement in cancer are complex and context dependent and must be experimentally explored in every particular context.

In conclusion, local synthesis of complement proteins allows activation of the classical complement pathway, that correlates with decreased survival. Moreover, C1s promotes tumor growth both by canonical classical pathway activation and a noncanonical, intracellular mechanism resulting in enhancement of tumor cell proliferation and modulation of the TME. These mechanisms likely synergize to confer a poor prognosis. Overexpression of C1s by tumor cells could be a new escape mechanism to promote tumor progression.

A. Grunenwald reports grants from Fondation pour la Recherche Médicale (FRM) during the conduct of the study. Y.A. Vano reports personal fees from Bristol Myers Squibb, Pfizer, Merck, Ipsen, Roche, Novartis, MSD, Janssen, Sanofi, and Astellas outside the submitted work. L.T. Roumenina reports grants from ARC Pour La Recherche Sur Le Cancer, the Ligue Regionale Contre le Cancer, The Labex Immuno-Oncology Excellence Program, and INSERM during the conduct of the study, as well as grants from Roche, CSL Behring, and Novartis outside the submitted work. No disclosures were reported by the other authors.

M.V. Daugan: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Revel: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. J. Russick: Investigation, methodology, writing–review and editing. M.-A. Dragon-Durey: Validation, investigation, methodology, writing–review and editing. C. Gaboriaud: Resources, formal analysis, validation, writing–review and editing. T. Robe-Rybkine: Investigation, methodology. V. Poillerat: Investigation, methodology, writing–review and editing. A. Grunenwald: Investigation, methodology. G. Lacroix: Investigation. A. Bougouin: Investigation. M. Meylan: Software, investigation, methodology. V. Verkarre: Resources. S.M. Oudard: Resources. A. Mejean: Resources. Y.A. Vano: Resources, formal analysis, validation, writing–review and editing. G. Perkins: Resources. P. Validire: Resources. X. Cathelineau: Resources. R. Sanchez-Salas: Resources. D. Damotte: Resources, formal analysis, validation, methodology. V. Fremeaux-Bacchi: Investigation. I. Cremer: Resources, validation, methodology, writing–review and editing. C. Sautès-Fridman: Conceptualization, resources, validation, methodology, writing–review and editing. W.H. Fridman: Conceptualization, resources, validation, writing–review and editing. L.T. Roumenina: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, project administration, writing–review and editing.

The authors are grateful to Veronique Rossi and Isabelle Bally (IBS, Grenoble) for providing the pcDNA3.1 C1s-expressing vector, described in ref. 53.

The authors are grateful to the team of the platform Centre d'Histologie, d'Imagerie et de Cytometrie (CHIC), at Centre de Recherche des Cordeliers UMRS 1138 (Paris, France), for the technical assistance. CHIC is a member of the Sorbonne University Flow Cytometry Network (RECYF). They thank Isabelle Sauret and Simon Lefranc from Centre de Ressources Biologiques and the Pathology Department at the IMM for their help with sample storage and collection.

This work was supported by grants from the Ligue Regionale Contre le Cancer and Fondation ARC Pour La Recherche Sur Le Cancer to L.T. Roumenina. This work was also supported by grants from CARPEM (ExhauCRF program), INCa (HTE program), Canceropole Ile de France (R17054DD), Association pour la recherche en thérapeutiques innovantes en cancérologie (ARTIC) to C. Sautès-Fridman, and a French National Research Agency grant (ANR-16-CE91-0004-01) to C. Gaboriaud. The Labex Immuno-Oncology Excellence Program, INSERM, University of Paris, and Sorbonne University also supported this work. M.V. Daugan received a PhD fellowship from La Fondation ARC Pour La Recherche Sur Le Cancer. A. Grunenwald received a PhD fellowship from Fondation pour la Recherche Médicale (FRM).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Hajishengallis
G
,
Reis
ES
,
Mastellos
DC
,
Ricklin
D
,
Lambris
JD
. 
Novel mechanisms and functions of complement
.
Nat Immunol
2017
;
18
:
1288
98
.
2.
Roumenina
L
,
Daugan
MV
,
Petitprez
F
,
Sautès-Fridman
C
,
Fridman
WH
. 
Context-dependent roles of complement in cancer
.
Nat Rev Cancer
2019
;
19
:
698
715
.
3.
Gaboriaud
C
,
Ling
WL
,
Thielens
NM
,
Bally
I
,
Rossi
V
. 
Deciphering the fine details of c1 assembly and activation mechanisms: “mission impossible”?
Front Immunol
2014
;
5
:
565
.
4.
Merle
NS
,
Church
SE
,
Fremeaux-Bacchi
V
,
Roumenina
LT
. 
Complement system part I - molecular mechanisms of activation and regulation
.
Front Immunol
2015
;
6
;
262
.
5.
Bally
I
,
Inforzato
A
,
Dalonneau
F
,
Stravalaci
M
,
Bottazzi
B
,
Gaboriaud
C
, et al
Interaction of C1q with pentraxin 3 and IgM revisited: mutational studies with recombinant C1q variants
.
Front Immunol
2019
;
10
:
461
.
6.
Sharp
TH
,
Boyle
AL
,
Diebolder
CA
,
Kros
A
,
Koster
AJ
,
Gros
P
. 
Insights into IgM-mediated complement activation based on in situ structures of IgM-C1-C4b
.
Proc Natl Acad Sci U S A
2019
;
116
:
11900
5
.
7.
Ugurlar
D
,
Howes
SC
,
de Kreuk
BJ
,
Koning
RI
,
de Jong
RN
,
Beurskens
FJ
, et al
Structures of C1-IgG1 provide insights into how danger pattern recognition activates complement
.
Science
2018
;
359
:
794
7
.
8.
Roumenina
LT
,
Daugan
MV
,
Noé
R
,
Petitprez
F
,
Vano
YA
,
Sanchez-Salas
R
, et al
Tumor cells hijack macrophage-produced complement C1q to promote tumor growth
.
Cancer Immunol Res
2019
;
7
:
1091
105
.
9.
Ajona
D
,
Pajares
MJ
,
Corrales
L
,
Perez-Gracia
JL
,
Agorreta
J
,
Lozano
MD
, et al
Investigation of complement activation product C4d as a diagnostic and prognostic biomarker for lung cancer
.
J Natl Cancer Inst
2013
;
105
:
1385
93
.
10.
Ajona
D
,
Okrój
M
,
Pajares
MJ
,
Agorreta
J
,
Lozano
MD
,
Zulueta
JJ
, et al
Complement C4d-specific antibodies for the diagnosis of lung cancer
.
Oncotarget
2017
;
9
:
6346
55
.
11.
Klikovits
T
,
Stockhammer
P
,
Laszlo
V
,
Dong
Y
,
Hoda
MA
,
Ghanim
B
, et al
Circulating complement component 4d (C4d) correlates with tumor volume, chemotherapeutic response and survival in patients with malignant pleural mesothelioma
.
Sci Rep
2017
;
7
:
16456
.
12.
Chang
IW
,
Lin
VCH
,
Wu
WJ
,
Liang
PI
,
Li
WM
,
Yeh
BW
, et al
Complement component 1, s subcomponent overexpression is an independent poor prognostic indicator in patients with urothelial carcinomas of the upper urinary tract and urinary bladder
.
J Cancer
2016
;
7
:
1396
405
.
13.
Sakai
N
,
Kusunoki
M
,
Nishida
M
,
Toyoguchi
T
,
Fukutomi
H
,
Sakiyama
H
. 
Tumorigenicity of BALB3T3 A31 cells transfected with hamster-complement-C1s cDNA
.
Int J Cancer
1994
;
58
:
309
13
.
14.
Sakiyama
H
,
Nishida
M
,
Sakai
N
,
Nagino
K
,
Miyatake
S
,
Saito
T
, et al
Site-directed mutagenesis of hamster complement C1S: characterization with an active form-specific antibody and possible involvement of C1S in tumorigenicity
.
Int J Cancer
1996
;
66
:
768
71
.
15.
Riihilä
P
,
Viiklepp
K
,
Nissinen
L
,
Farshchian
M
,
Kallajoki
M
,
Kivisaari
A
, et al
Tumor cell-derived complement components C1r and C1s promote growth of cutaneous squamous cell carcinoma
.
Br J Dermatol
2020
;
182
:
658
70
.
16.
Yu
K
,
Yang
H
,
Lv
QL
,
Wang
LC
,
Tan
ZL
,
Zhang
Z
, et al
Construction of a competitive endogenous RNA network and analysis of potential regulatory axis targets in glioblastoma
.
Cancer Cell Int
2021
;
21
:
102
.
17.
Giraldo
NA
,
Becht
E
,
Pagès
F
,
Skliris
G
,
Verkarre
V
,
Vano
Y
, et al
Orchestration and prognostic significance of immune checkpoints in the microenvironment of primary and metastatic renal cell cancer
.
Clin Cancer Res
2015
;
21
:
3031
40
.
18.
Daugan
MV
,
Revel
M
,
Lacroix
L
,
Sautès-Fridman
C
,
Fridman
WH
,
Roumenina
LT
. 
Complement detection in human tumors by immunohistochemistry and immunofluorescence
.
Methods Mol Biol
2021
;
2227
:
191
203
.
19.
Love
MI
,
Huber
W
,
Anders
S
. 
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol
2014
;
15
:
550
.
20.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio Cancer Genomics Portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
21.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
22.
Becht
E
,
Giraldo
NA
,
Lacroix
L
,
Buttard
B
,
Elarouci
N
,
Petitprez
F
, et al
Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression
.
Genome Biol
2016
;
17
:
218
.
23.
Young
MD
,
Mitchell
TJ
,
Vieira Braga
FA
,
Tran
MGB
,
Stewart
BJ
,
Ferdinand
JR
, et al
Single-cell transcriptomes from human kidneys reveal the cellular identity of renal tumors
.
Science
2018
;
361
:
594
9
.
24.
Stuart
T
,
Butler
A
,
Hoffman
P
,
Hafemeister
C
,
Papalexi
E
,
Mauck
WM
, et al
Comprehensive integration of single-cell data
.
Cell
2019
;
177
:
1888
902
.
25.
Hu
J
,
Chen
Z
,
Bao
L
,
Zhou
L
,
Hou
Y
,
Liu
L
, et al
Single-cell transcriptome analysis reveals intratumoral heterogeneity in ccRCC, which results in different clinical outcomes
.
Mol Ther
2020
;
28
:
1658
72
.
26.
Motzer
RJ
,
Robbins
PB
,
Powles
T
,
Albiges
L
,
Haanen
JB
,
Larkin
J
, et al
Avelumab plus axitinib versus sunitinib in advanced renal cell carcinoma: biomarker analysis of the phase 3 JAVELIN Renal 101 trial
.
Nat Med
2020
;
26
:
1733
41
.
27.
Hoadley
KA
,
Yau
C
,
Hinoue
T
,
Wolf
DM
,
Lazar
AJ
,
Drill
E
, et al
Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer
.
Cell
2018
;
173
:
291
304
.
28.
Katz
Y
,
Strunk
RC
. 
Synthesis and regulation of C1 inhibitor in human skin fibroblasts
.
J Immunol
1989
;
14
:
2041
5
.
29.
Takeda
K
,
Nakayama
M
,
Hayakawa
Y
,
Kojima
Y
,
Ikeda
H
,
Imai
N
, et al
IFN-γ is required for cytotoxic T cell-dependent cancer genome immunoediting
.
Nat Commun
2017
;
8
:
14607
.
30.
Schoenborn
JR
,
Wilson
CB
. 
Regulation of interferon-gamma during innate and adaptive immune responses
.
Adv Immunol
2007
;
96
:
41
101
.
31.
Fridman
WH
,
Zitvogel
L
,
Sautès-Fridman
C
,
Kroemer
G
. 
The immune contexture in cancer prognosis and treatment
.
Nat Rev Clin Oncol
2017
;
14
:
717
34
.
32.
Reese
B
,
Silwal
A
,
Daugherity
E
,
Daugherity
M
,
Arabi
M
,
Daly
P
, et al
Complement as prognostic biomarker and potential therapeutic target in renal cell carcinoma
.
J Immunol
2020
;
205
:
3218
29
.
33.
Netti
GS
,
Lucarelli
G
,
Spadaccino
F
,
Castellano
G
,
Gigante
M
,
Divella
C
, et al
PTX3 modulates the immunoflogosis in tumor microenvironment and is a prognostic factor for patients with clear cell renal cell carcinoma
.
Aging
2020
;
12
:
7585
602
.
34.
Martin
M
,
Smoląg
KI
,
Björk
A
,
Gullstrand
B
,
Okrój
M
,
Leffler
J
, et al
Plasma C4d as marker for lupus nephritis in systemic lupus erythematosus
.
Arthritis Res Ther
2017
;
19
:
266
.
35.
Tower
CM
,
Reyes
M
,
Nelson
K
,
Leca
N
,
Kieran
N
,
Muczynski
K
, et al
Plasma C4d+ endothelial microvesicles increase in acute antibody-mediated rejection
.
Transplantation
2017
;
101
:
2235
43
.
36.
Xi
W
,
Liu
L
,
Wang
J
,
Xia
Y
,
Bai
Q
,
Xiong
Y
, et al
Enrichment of C5a-C5aR axis predicts poor postoperative prognosis of patients with clear cell renal cell carcinoma
.
Oncotarget
2016
;
7
:
80925
34
.
37.
Xi
W
,
Liu
L
,
Wang
J
,
Xia
Y
,
Bai
Q
,
Long
Q
, et al
High level of anaphylatoxin C5a predicts poor clinical outcome in patients with clear cell renal cell carcinoma
.
Sci Rep
2016
;
6
:
29177
.
38.
Kiafard
Z
,
Tschernig
T
,
Schweyer
S
,
Bley
A
,
Neumann
D
,
Zwirner
J
. 
Use of monoclonal antibodies to assess expression of anaphylatoxin receptors in tubular epithelial cells of human, murine and rat kidneys
.
Immunobiology
2007
;
212
:
129
39
.
39.
Maeda
Y
,
Kawano
Y
,
Wada
Y
,
Yatsuda
J
,
Motoshima
T
,
Murakami
Y
, et al
C5aR is frequently expressed in metastatic renal cell carcinoma and plays a crucial role in cell invasion via the ERK and PI3 kinase pathways
.
Oncol Rep
2015
;
33
:
1844
50
.
40.
Daugan
MV
,
Revel
M
,
Thouenon
R
,
Dragon-Durey
MA
,
Robe-Rybkine
T
,
Torset
C
, et al
Intracellular factor H drives tumor progression independently of the complement cascade
.
Cancer Immunol Res
2021
;
9
:
909
25
.
41.
Liszewski
MK
,
Kolev
M
,
Le Friec
G
,
Leung
M
,
Bertram
PG
,
Fara
AF
, et al
Intracellular complement activation sustains T cell homeostasis and mediates effector differentiation
.
Immunity
2013
;
39
:
1143
57
.
42.
King
BC
,
Kulak
K
,
Krus
U
,
Rosberg
R
,
Golec
E
,
Wozniak
K
, et al
Complement component C3 is highly expressed in human pancreatic islets and prevents β cell death via ATG16L1 interaction and autophagy regulation
.
Cell Metab
2019
;
29
:
202
10
.
43.
King
BC
,
Renström
E
,
Blom
AM
. 
Intracellular cytosolic complement component C3 regulates cytoprotective autophagy in pancreatic beta cells by interaction with ATG16L1
.
Autophagy
2019
;
15
:
919
21
.
44.
Yeo
JG
,
Leong
J
,
Arkachaisri
T
,
Cai
Y
,
Teo
BHD
,
Tan
JHT
, et al
Proteolytic inactivation of nuclear alarmin high-mobility group box 1 by complement protease C1s during apoptosis
.
Cell Death Discov
2016
;
2
:
16069
.
45.
Naito
AT
,
Sumida
T
,
Nomura
S
,
Liu
ML
,
Higo
T
,
Nakagawa
A
, et al
Complement C1q activates canonical Wnt signaling and promotes aging-related phenotypes
.
Cell
2012
;
149
:
61298
1313
.
46.
Sakiyama
H
,
Kaji
K
,
Nakagawa
K
,
Nagino
K
. 
Inhibition of bFGF activity by complement C1s: covalent binding of C1s with bFGF
.
Cell Biochem Funct
1998
;
16
:
159
63
.
47.
Busby
WH
,
Nam
TJ
,
Moralez
A
,
Smith
C
,
Jennings
M
,
Clemmons
DR
. 
The complement component C1s is the protease that accounts for cleavage of insulin-like growth factor-binding protein-5 in fibroblast medium
.
J Biol Chem
2000
;
275
:
37638
44
.
48.
Li
SW
,
Yu
B
,
Byrne
G
,
Wright
M
,
O'Rourke
S
,
Mesa
K
, et al
Identification and CRISPR/Cas9 inactivation of the C1s protease responsible for proteolysis of recombinant proteins produced in CHO cells
.
Biotechnol Bioeng
2019
;
116
:
2130
45
.
49.
Eriksson
H
,
Nissen
MH
. 
Proteolysis of the heavy chain of major histocompatibility complex class I antigens by complement component C1s
.
Biochim Biophys Acta
1990
;
1037
:
209
15
.
50.
Sakiyama
H
,
Nakagawa
K
,
Kuriiwa
K
,
Imai
K
,
Okada
Y
,
Tsuchida
T
, et al
Complement Cls, a classical enzyme with novel functions at the endochondral ossification center: immunohistochemical staining of activated Cls with a neoantigen-specific antibody
.
Cell Tissue Res
1997
;
288
:
557
65
.
51.
Cai
Y
,
Teo
BHD
,
Yeo
JG
,
Lu
J
. 
C1q protein binds to the apoptotic nucleolus and causes C1 protease degradation of nucleolar proteins
.
J Biol Chem
2015
;
290
:
22570
80
.
52.
Lu
J
,
Kishore
U
. 
C1 complex: an adaptable proteolytic module for complement and non-complement functions
.
Front Immunol
2017
;
8
:
592
.
53.
Bally
I
,
Dalonneau
F
,
Chouquet
A
,
Gröbner
R
,
Amberger
A
,
Kapferer-Seebacher
I
, et al
Two different missense C1S mutations, associated to periodontal Ehlers-Danlos syndrome, lead to identical molecular outcomes
.
Front Immunol
2019
;
10
:
2962
.
54.
Rapoport
BL
,
Steel
HC
,
Theron
AJ
,
Heyman
L
,
Smit
T
,
Ramdas
Y
, et al
High mobility group box 1 in human cancer
.
Cells
2020
;
9
:
1664
.
55.
Merle
NS
,
Singh
P
,
Rahman
J
,
Kemper
C
. 
Integrins meet complement: the evolutionary tip of an iceberg orchestrating metabolism and immunity
.
Br J Pharmacol
2020
Jun 19 [Epub ahead of print]
.
56.
Motzer
RJ
,
Banchereau
R
,
Hamidi
H
,
Powles
T
,
McDermott
D
,
Atkins
MB
, et al
Molecular subsets in renal cancer determine outcome to checkpoint and angiogenesis blockade
.
Cancer Cell
2020
;
38
:
803
17
.
57.
Luo
C
,
Chen
M
,
Madden
A
,
Xu
H
. 
Expression of complement components and regulators by different subtypes of bone marrow-derived macrophages
.
Inflammation
2012
;
35
:
1448
61
.
58.
Gulati
P
,
Lemercier
C
,
Guc
D
,
Lappin
D
,
Whaley
K
. 
Regulation of the synthesis of C1 subcomponents and C1-inhibitor
.
Behring Inst Mitt
1993
;(
93
):
196
203
.