Non-Oncogene Addiction of KRAS-Mutant Cancers to IL-1β via Versican and Mononuclear IKKβ

Simple Summary Kirsten rat sarcoma virus (KRAS)-mutant cancers are frequent, metastatic, lethal, and largely undruggable. The aim of this study was to investigate the pathways through which KRAS-mutant cancers foster their growth, thereby unravelling novel therapeutic targets. We show that KRAS-mutant tumors secrete the protein versican, which then drives the activation of NF-κB kinase (IKK) β in a type of host immune cells called macrophages. Following this activation, macrophages fuel the tumor with interleukin (IL)-1β, to close an inflammatory loop through which KRAS-mutant cancers attract host immune cells to the tumor site to accelerate tumor growth and aggressiveness. Importantly, we show that targeting IL-1β and/or versican can be an effective treatment for KRAS-mutant cancers, holding great promise for cancer patients. Abstract Kirsten rat sarcoma virus (KRAS)-mutant cancers are frequent, metastatic, lethal, and largely undruggable. While interleukin (IL)-1β and nuclear factor (NF)-κB inhibition hold promise against cancer, untargeted treatments are not effective. Here, we show that human KRAS-mutant cancers are addicted to IL-1β via inflammatory versican signaling to macrophage inhibitor of NF-κB kinase (IKK) β. Human pan-cancer and experimental NF-κB reporter, transcriptome, and proteome screens reveal that KRAS-mutant tumors trigger macrophage IKKβ activation and IL-1β release via secretory versican. Tumor-specific versican silencing and macrophage-restricted IKKβ deletion prevents myeloid NF-κB activation and metastasis. Versican and IKKβ are mutually addicted and/or overexpressed in human cancers and possess diagnostic and prognostic power. Non-oncogene KRAS/IL-1β addiction is abolished by IL-1β and TLR1/2 inhibition, indicating cardinal and actionable roles for versican and IKKβ in metastasis.


Mouse Tumor Models
For the generation of solid tumors, mice were injected subcutaneously (s.c.) in the shaven rear flank dermis with 5 × 10 5 tumor cells in 100 µL of phosphate-buffered saline (PBS), as described elsewhere [10,11,20,22]. Mice were weekly examined for tumor volume (V) by measuring three vertical tumor diameters (d1, d2, d3) using the formula V = π × d1 × d2 × d3 and were killed when the tumor volume reached 1 cm 3 (PANO2 cells) or 2 cm 3 (all other cell lines). For the induction of malignant pleural effusions (MPE), mice received intrapleural injections of 2 × 10 5 cancer cells suspended in 100 µL PBS and were sacrificed when showing signs of sickness or at the time-points indicated (14-28 days post-tumor cell delivery depending on the cell line used) [10,11,22]. In all models, both the mice and the inoculated cancer cells were always syngeneic to avoid inflammatory allograft rejection and artificial NF-κB activation.

Bioluminescence and Biofluorescence Imaging
Mice were imaged for NF-κB reporter bioluminescent signal daily starting at day 10 post-tumor cell injection until sacrifice. For this, mice were anesthetized by isoflurane inhalation and were imaged for bioluminescence on a Xenogen Lumina II (PerkinElmer, Waltham, MA, USA) 5-20 min after delivery of 1 mg D-Luciferin potassium salt diluted in 100 µL of sterile water into a retro-orbital vein. Pleural tumors isolated from NGL mice were also imaged ex vivo for green biofluorescence using 410-440 nm background control excitation, 445-490 nm experimental excitation, and 515-575 nm emission passbands on a Xenogen Lumina II. Cells were imaged for bioluminescence on a Xenogen Lumina II 0, 4, 8, 16, and 24 h after a single addition of 300 µg/mL (equivalent to 1 mM) Dluciferin to the culture media. Data were analyzed using Living Image v.4.2 (PerkinElmer, Waltham, MA, USA) as described previously [10,11,[20][21][22]34].

Sequencing
Genomic DNA was extracted from cell lines using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich). Kras exons 1-3 were amplified by PCR using Phusion Polymerase (New England Biolabs, Ipswich, MA, USA) and 60 • C annealing temperature. Primers are described in Table S4. PCR products were analyzed on 1% agarose gels, purified by QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany) and sequenced by Eurofins Genomics (Ebersberg, Germany).

Intrapleural Catheter
For in vivo MPE drainage, a 1.2 cm-long catheter bearing serial fenestrations at 1 mm intervals was used, according to the detailed model description reported previously [35]. Mice were anesthetized using isoflurane and the catheter insertion site was shaved and disinfected using 70% ethanol and 10% povidone iodide, and the catheter was then installed into the pleural space and sutured under the skin. Mice were imaged pre-and post-MPE drainage and were sacrificed thereafter.

Cytology
MPE fluid was treated with red blood cell lysis buffer (155 mM NH 4 Cl, 12 mM NaHCO 3 , 0.1 mM EDTA) and MPE cells were centrifuged and stained with May-Grünwald-Giemsa. Slides were then mounted with Entellan (Merck Millipore, Darmstadt, Germany) and microscopically analyzed for the differential counting of pleural cells. Pleural lavage was performed by injecting 1 mL of saline intrapleurally and recovering it after 30 s. Pleural cells were enumerated with a haemocytometer, were centrifuged, were stained with May-Grünwald-Giemsa or with anti-rabbit F4/80 antibody (ab111101; Abcam, London, UK; RRID:AB_10859466) and hematoxylin, and were microscopically analyzed for the differential counting of pleural cells.

Bone Marrow Transfer (BMT) and Liposomal Clodronate
For adoptive BMT experiments described in detail elsewhere [10,11,20], wild-type (WT) and NF-κB.eGFP.LUC (NGL) recipient mice on the C57BL/6 background received total body irradiation (1100 Rad) followed 12 h later by 10 7 intravenous (via retro-orbital injection) whole bone marrow cells obtained from WT and NGL donors. One irradiated mouse per group was not transplanted with BMT to control for effective elimination of endogenous bone marrow and died 5-15 days post-irradiation. After one month, allowing for complete bone marrow reconstitution by chimeric bone marrow cells, liposomal clodronate was prepared as described previously [25,38] and 500 µg were administered intrapleurally. After yet another month required for the replacement of pleural myeloid cells by transplanted bone marrow cells [38], the mice were injected with tumor cells.

Bone Marrow Derived Macrophages (BMDM)
For BMDM generation, 10 7 bone marrow cells were plated and cultured for 7 days in the presence of 100 ng/mL macrophage colony stimulating factor (M-CSF). Where appropriate, at day 6 of the culture, recombinant human versican (1 nM) was added to the culture medium or, alternatively, the culture medium was removed and BMDM were exposed to cancer cell-conditioned media for 4 h. Culture supernatants were then isolated for ELISA and cells were processed for western blot, flow cytometry, or qPCR.

Immunoblotting
Nuclear and cytoplasmic protein extracts were prepared using the NEPER Extraction Kit (Thermo Fisher Scientific, Waltham, MA, USA), separated by SDS-PAGE and electroblotted to PVDF membranes (Merck Millipore, Darmstadt, Germany). Membranes were probed with the following primary antibodies: anti-IKKα

qPCR and Microarrays
Triplicate cultures of 10 6 cells were subjected to RNA extraction using Trizol (Thermo Fisher Scientific, Waltham, MA, USA) followed by column purification and DNA removal (RNeasy Mini Kit, Qiagen, Hilden, Germany). Pooled RNA (5 µg) was quality tested on an ABI 2000 bioanalyzer (Agilent Technologies, Sta. Clara, CA, USA), labelled, and hybridized to GeneChip Mouse Gene 2.0 ST arrays (Affymetrix, Sta. Clara, CA, USA). All data were analyzed on the Affymetrix Expression and Transcriptome Analysis Consoles (RRID:SCR_018718). RNA was reverse transcribed with Superscript III (Thermo Fisher Scientific) and qPCR was performed using first-strand synthesis and SYBR FAST qPCR Kit (Kapa Biosystems, Wilmington, MA, USA) in a StepOne cycler (Applied Biosystems, Carlsbad, CA, USA). Primers for qPCR are listed in Table S4. Ct values from triplicate reactions were analyzed with the relative quantification method 2 −∆∆CT relative to mouse Gusb or human ACTB transcripts [39].

Shotgun Proteomics
Supernatants obtained from murine Kras MUT (LLC, MC38, AE17) and Kras WT (B16F10 and PANO2) cell cultures (pooled triplicate cultures for each cell line; 5 million cells/175 cm 2 culture flask/24 h in full DMEM followed by 24 h in FBS-free DMEM) were analyzed. For this, 600 µL of cell culture supernatant were enzymatically digested using a modified filter-aided sample preparation (FASP) protocol [40,41]. Peptides were stored at −20 • C until mass spectrometry (MS) measurements. MS data were acquired in data-dependent acquisition (DDA) mode on a Q Exactive (QE) high field (HF) mass spectrometer (Thermo Fisher Scientific). Approximately 0.5 µg per sample were automatically loaded to the online coupled RSLC (Ultimate 3000, Thermo Fisher Scientific) HPLC system. A nano trap column was used (300 µm inner diameter (ID) × 5 mm, packed with Acclaim PepMap100 C18, 5 µm, 100 Å (LC Packings, Sunnyvale, CA, USA) before separation by reversed phase chromatography (Acquity UPLC M-Class HSS T3 Column 75 µm ID × 250 mm, 1.8 µm; Waters, Eschborn, Germany) at 40 • C. Peptides were eluted from 3% to 40% over a 95 min gradient. The MS spectrum was acquired with a mass range from 300 to 1500 m/z at resolution 60,000 with AGC set to 3 × 10 6 and a maximum of 50 ms IT. From the MS pre-scan, the 10 most abundant peptide ions were selected for fragmentation (MSMS) if at least doubly charged, with a dynamic exclusion of 30 s. MSMS spectra were recorded at 15,000 resolution with AGC set to 1 × 10 5 and a maximum of 100 ms IT. CE was set to 28 and all spectra were recorded in profile type. Label-free quantification of DDA-MS data was performed with the Proteome discoverer (version 2.3; Thermo Fisher Scientific) using Sequest HT (as node in PD) and searching against the UniProtKB/Swiss-Prot Mouse database (release 2017_2, 16872 sequences). Searches were performed with a precursor mass tolerances of 10 ppm and fragment mass tolerances of 0.02 Da. Carbamidomethylation (C) was set as static modification, deamidation (N,Q), oxidation (M), and N-terminal Met-loss+Acetyl were selected as dynamic modifications and two missed cleavages were allowed. Percolator [42] was used for validating peptide spectrum matches and peptides, accepting only the top-scoring hit for each spectrum, and satisfying the cut-off values for FDR < 1%, and posterior error probability < 0.01. The final list of proteins complied with the strict parsimony principle. The quantification of proteins, after precursor recalibration, was based on abundance values (area under curve) for unique peptides. Abundance values were normalized in a retention time-dependent manner. The protein abundances were calculated summing the abundance values for admissible peptides. Comparisons between Kras MUT (LLC, MC38, AE17) and Kras WT (B16F10 and PANO2) cell lines were done using only the proteins detected in all five cell lines.

Mouse Treatments
The IL-1 receptor antagonist isunakinra [18] was given via daily intraperitoneal injections of 20-50 mg/kg drug diluted in 100 µL PBS. Therapy was initiated at 10-17 days post s.c. tumor cells or at 5 days post-intrapleural tumor cells, allowing for efficient tumor take and a therapeutic study design. Treatment with the TLR1/TLR2 antagonist Cu-CPT22 [19] was initiated 3 days after the intrapleural cancer cell injection and consisted of daily intraperitoneal injections of 100 µL corn oil containing 10% DMSO or 20 mg/kg Cu-CPT22 diluted in 100 µL corn oil containing 10% DMSO.

Transcription Factor Binding Site Analyses
We downloaded the RELA and RELB binding sequence motifs from the ENCODE portal [43] with the identifiers: ENCFF507YCV (CHIP-seq on HuH-7.5 cells) and ENCFF615HZF (CHIP-seq on 8988T cells), respectively, and queried the ChIPseq datasets from the ChIP-X Enrichment Analysis (CHEA) Transcription Factor Targets dataset [44][45][46].

Statistics
Sample size was calculated using power analysis on G*power [47], assuming α = 0.05, β = 0.05, and effect size d = 1.5. No data were excluded from analyses. Pooled data from repeated in vivo experiments are shown. All in vitro experiments were repeated independently at least three times and the stated n always reflects the biological and not technical sample size. Animals were allocated to treatments by randomization (when n ≥ 20) or alternation (when n < 20) and transgenic animals were enrolled case-controlwise. Data were collected by at least two blinded investigators from samples coded by non-blinded investigators. All data were tested for normality of distribution by the Kolmogorov-Smirnov test, are given as violin plots or mean ± SD, and sample size (n) always refers to the biological and not technical replicates. Differences in frequency were examined by Fischer's exact and χ 2 tests, in medians by Mann-Whitney or Kruskal-Wallis tests with Dunn's post-tests, and in means by t-test or one-way ANOVA with Bonferroni post-tests. Changes over time and the interaction between two variables were examined by two-way ANOVA with Bonferroni post-tests. Hypergeometric tests were done at the Graeber Lab website [48]. All probability (p) values are two-tailed and were considered significant when p < 0.05. All analyses and plots were done on Prism v8.0 (GraphPad, La Jolla, CA, USA; RRID:SCR_002798).

Non-Oncogene Addiction of KRAS-Mutant Human and Murine Cancers to IL-1β
Puzzled by the negative results of the CANOPY-2 trial, we focused on published mutation data from incident LUAD from the CANTOS trial [49] and cross-examined them with the cancer genome atlas (TCGA) LUAD dataset [50], hypothesizing that IL-1β neutralization with canakinumab would specifically prevent the development of incipient KRAS-mutant ( MUT ) LUAD. Indeed, KRAS, but not TP53, EGFR, and BRAF, mutations were statistically significantly under-enriched in CANTOS versus TCGA patients ( Figure 1A,B). We next analyzed TCGA pan-cancer transcriptome data to discover that IL1B mRNA levels were elevated in KRAS MUT and amplified cancers, and performed IL-1β immunohistochemistry in our own patients with resected LUAD [13] to find increased IL-1β protein expression in KRAS MUT LUAD compared with KRAS-wild-type ( WT ) LUAD and adjacent lung tissues ( Figure 1C,D). We next injected C57BL/6 mice competent (WT) and diploinsufficient for Il1b alleles (Il1b−/−) [32] with syngeneic cancer cell lines carrying Kras WT and Kras MUT alleles [10,11]. Both subcutaneous (s.c.) and pleural routes of tumor cell injection were employed, since we previously identified that malignant pleural effusions (MPE) in mice are exclusively elicited by Kras MUT tumor cells [10,11]. All cell lines were verified for Kras, Mycoplasma spp., and identity status multiple times during these investigations ( Figure S1A,B). These experiments showed that specifically Kras MUT tumors were dependent on host IL-1β ( Figure 1E). Taken together, these results show that IL-1β neutralization prevents the development of incipient KRAS MUT LUAD in humans, that KRAS MUT human cancers contain elevated IL-1β levels, and that mouse Kras MUT cancers are specifically dependent on host IL-1β signaling, supporting the hypothesis of a selective non-oncogene addiction of KRAS MUT cancers to IL-1β.

Tumor-Associated Macrophages as a Source of Tumorigenic IL-1β
We next investigated the source of increased IL-1β in KRAS MUT cancers, since both the host immune and tumor cells are capable of IL-1β production [20,53,54]. We were also, based on previous work, documenting that the IL-1β promoter lies under transcriptional control of NF-κB [55], a fact we validated in ChIPseq datasets from the ChIP-X Enrichment Analysis (CHEA) dataset [46] and the ENCyclopedia Of DNA Elements (ENCODE) portal [44] (Figure S1C). For this, we first searched TCGA pan-cancer transcriptomes (n = 10,071) for associations between mRNA levels of IL1B and established cancer and immune cellular lineage markers. IL1B mRNA levels were not correlated with mRNA levels of the neutrophil marker ELANE, the mast cell marker KIT, the fibroblast marker ACTA2, and the endothelial marker F8, were significantly associated with mRNA levels of KRAS per se, of the pan-lymphocyte marker CD3D, and the cancer cell marker KRT18, but showed the tightest correlation (coefficient = 0.4; p < 10 −300 ) with mRNA levels of the macrophage marker ADGRE1 (Figures S2 and S3). To further test this, we sought to identify the host cells that respond to KRAS MUT tumor cells with NF-κB activation, since the transcription factor controls IL-1β transcription [55] and is central to innate immune responses [56]. For this, we initiated in vivo screens of murine tumor cell lines with known Kras mutation status (Figures 2A and S1 [10]) by transplanting them into two strains of bioluminescent NF-κB reporter mice expressing ubiquitous HIV-LTR.Luciferase (HLL mice) [24] or NF-κB.GFP.Luciferase (NGL mice) [25] transgenes. Pleural injections were selected for tumor cell inoculation because they generate MPE with overt cancerinduced inflammation [10,11,20]. Serial imaging showed time-dependent NF-κB activation in host cells of recipient mice, conditional on the presence of Kras mutations in tumor cells ( Figures 2B-E and S4A,B). The NF-κB reporter signal was emitted from pleural tumors and fluid, both containing cancer and immune cells ( Figures 2F and S4C-F) [10,11,20]. Histologic and flow cytometric analysis and quantification localized the NF-κB reporter signal to tumor-infiltrating macrophages of mice with Kras MUT pleural tumors and effusions ( Figures 2G-I and S5-S7). Mast cells that foster MPE development [20] were not involved in the observed NF-κB response ( Figure S8A,B). Time-dependent NF-κB activation in host cells was stronger in pleural compared with s.c. tumor models, and required expression of mutant Kras by tumor cells (Figures S8C,D and S9A-C). Adoptive bone marrow transfer corroborated myeloid cells as the origin of tumor-induced NF-κB activation, and pharmacologic killing of pleural macrophages prevented host NF-κB activation and pleural carcinomatosis ( Figures S10 and S11A,B).

Tumor-Secreted Versican as a Key Macrophage Effector
We next compared Kras MUT with Kras WT cancer cells for secretory molecules triggering macrophage NF-κB activation. Microarrays identified 25 transcripts over-represented in Kras MUT tumor cells, and a proteomic screen of tumor cell-conditioned media detected 226 proteins secreted > 10-fold by Kras MUT over Kras WT cells, with the glycoprotein versican (VCAN; encoded by the human/murine VCAN/Vcan genes) emerging from both screens and withstanding validation ( Figures 4A-E and S12D,E, Table S2, and Data S1). Multiple NF-κB ligands were also screened using pNGL-expressing RAW264.7 macrophages, revealing that the toll-like receptor (TLR)2 ligand VCAN potently activates macrophage NF-κB-driven transcription to the same degree as the TLR4 ligand lipopolysaccharide (LPS) (Figure 4F,G). VCAN also induced IKKβ in primary murine BMDM, which were verified by microarray to overexpress > 10-fold over cancer cells TLR1, TLR2, TLR6-9, and TLR13 (Figures 4H, S12F,G and S13). Importantly, shRNA-mediated Vcan silencing in LLC cells diminished their ability to trigger NF-κB activation in NGL mice and to precipitate MPE ( Figures 4I-M and S12I-J). VCAN overexpression is not restricted to mouse Kras MUT cancers, since VCAN transcripts are also over-represented in human cancers with high KRAS MUT frequencies (derived from the catalogue of somatic mutations in cancer, COS-MIC), such as LUAD from smokers (GEO dataset GSE43458), and NSCLC and colorectal adenocarcinoma (COAD/READ; GEO dataset GSE103512) ( Figure S14A,B) [58][59][60]. High VCAN mRNA expression also portended poor survival in a number of human cancers from the KMplot pan-cancer RNAseq dataset (Figures S14C and S15) [61]. Analysis of samples from two of our own clinical studies [13,14] showed that VCAN protein expression was significantly increased in LUAD compared with adjacent lung tissues and that VCAN mRNA expression was significantly increased in human MPE compared with benign pleural effusions (BPE) (Figure 4N,O). To test whether the proposed inflammatory loop can serve as a diagnostic tool to distinguish MPE from BPE, which is an unmet clinical need [14], pNGL-expressing RAW264.7 macrophages were exposed to cell-free supernatants from human pleural effusions. After 4 h, a robust NF-κB reporter signal was triggered selectively by MPE supernatants ( Figure 4P). Taken together, these data indicate that VCAN secreted by cancer cells triggers IKKβ-mediated NF-κB activation in tumor-associated macrophages and promotes metastasis. Moreover, VCAN is overexpressed in human KRAS MUT cancers and can serve as a diagnostic and prognosis biomarker.

Myeloid IKKB as the VCAN Accessory
To identify the IKK responsible for NF-κB signaling in macrophages, we silenced the four main IKKs (encoded by the murine Chuk, Ikbkb, Ikbke, and Tbk1 genes) in RAW264.7 macrophages and identify IKKβ as the main mediator of NF-κB activation in these cells ( Figure 5A,B). To further define myeloid IKKβ functions, we obtained BMDM from intercrosses of Lyz2.Cre mice with mice carrying conditionally-deleted alleles of IKKα (Chuk f/f ) and IKKβ (Ikbkb f/f ), as well as with Cre-reporter mice switching from red to green fluorescence upon Cre-mediated recombination (mT/mG), all reported previously [21,22]. Treatment of bone marrow cells from mT/mG;Lyz2.Cre mice with macrophage-colony stimulating factor (M-CSF; 100 ng/mL) to drive them towards macrophage differentiation and lysozyme 2 (LYZ2) expression yielded efficient Cre-mediated recombination ( Figure 5C). Flow cytometric assessment of BMDM derived from these mice showed that intact IKKβ signaling in primary macrophages is essential for their differentiation and expression of critical pro-inflammatory genes including Lyz2, Il1b, and C3 ( Figure 5D-F and Table S3). Finally, two different syngeneic Kras MUT tumor cell lines featuring VCAN overexpression were inoculated into the pleural space of the above myeloid IKK-deleted mice, to reveal that intact IKKβ signaling in macrophages is required for MPE ( Figure 5G). Thus, VCANdriven IKKβ activation mediates NF-κB signaling, IL-1β expression, differentiation, and pro-tumor function of macrophages ( Figure 5H). To further query the proposed KRAS-VCAN-IKKβ connection, we interrogated mutations, copy number alterations, and fusions of the encoding genes in the TCGA pan-cancer dataset. Interestingly, VCAN and IKBKB alterations (mostly missense mutations) each occur in 5% of all cancer patients and are significantly mutually enriched (VCAN in KRAS and IKBKB in VCAN mutations) suggesting mutual addiction ( Figure S16A-C). In addition, KRAS, IKBKB, and VCAN alteration frequencies across 32 human cancer types are tightly correlated, and were highest in LUAD, COAD/READ, and uterine corpus endometrial carcinoma (UCEC), cancers that commonly cause MPE ( Figure S16D). In the latter tumor types featuring KRAS, IKBKB, and VCAN alteration frequencies, addiction of IKBKB and VCAN mutations persisted, and patients with VCAN and/or IKBKB-altered cancers displayed decreased body mass (cachexia), higher mutation burden, microsatellite instability, and hypoxia indices ( Figure S17). Collectively, these data support that tumor cell VCAN cooperates with myeloid IKKβ in mouse and human cancers.

Non-oncogene Addiction of KRAS-Mutant Tumors to IL-1β Is Actionable
To block the proposed inflammatory loop, we employed the novel IL-1 receptor antagonist isunakinra [18]. Systemic delivery of isunakinra to mice with already established tumors specifically inhibited s.c. growth of Kras MUT tumors ( Figure 6A). In addition, isunakinra limited NF-κB activation in Kras MUT cancer cells in vivo, a phenomenon we previously showed to be fueled by myeloid IL-1β, as well as their ability for lethal MPE induction ( Figure 6B-D). Since VCAN is a known TLR2 ligand [57], the pro-inflammatory loop proposed here was also targeted with the TLR1/2 inhibitor Cu-CPT22 [19]. The drug effectively inhibited VCAN-induced NF-κB activation and cellular survival in RAW264.7 macrophages at the low micromolar range and blocked tumor growth in vivo at clinically relevant concentrations ( Figure 6E-H). Hence, VCAN-IKKβ-mediated addiction of KRAS MUT cancers to host IL-1β can be used to indirectly target these tumors.

Non-oncogene Addiction of KRAS-mutant Tumors to IL-1β Is Actionable
To block the proposed inflammatory loop, we employed the novel IL-1 receptor antagonist isunakinra [18]. Systemic delivery of isunakinra to mice with already established tumors specifically inhibited s.c. growth of Kras MUT tumors ( Figure 6A). In addition, isunakinra limited NF-κΒ activation in Kras MUT cancer cells in vivo, a phenomenon we previously showed to be fueled by myeloid IL-1β, as well as their ability for lethal MPE induction ( Figure 6B-D). Since VCAN is a known TLR2 ligand [57], the pro-inflammatory loop proposed here was also targeted with the TLR1/2 inhibitor Cu-CPT22 [19]. The drug effectively inhibited VCAN-induced NF-κΒ activation and cellular survival in RAW264.7 macrophages at the low micromolar range and blocked tumor growth in vivo at clinically relevant concentrations ( Figure 6E-H). Hence, VCAN-IKKβ-mediated addiction of KRAS-MUT cancers to host IL-1β can be used to indirectly target these tumors.  and C57BL/6 (n = 45) mice received subcutaneous injections of 5 × 10 6 FULA1 (FVB mice) or LLC, MC38, B16F10, or PANO2 (C57BL/6 mice) cells that carry G12C, G13R, Q61R, or wild-type ( WT ) Kras alleles. Mice were allowed 10-23 days for tumor take (solid circles) and were treated with daily intraperitoneal PBS or 20 mg/Kg isunakinra until control tumor volume reached 1 cm 3 (PANO2 cells) or 2 cm 3 (all other cell lines). Shown are mouse numbers (n), tumor volume as mean (circles) and SD (bars), two-way ANOVA probability (P) for treatment effects, and average isunakinra effect at the last time-points (%). ** and ***: p < 0.01 and p < 0.001, respectively, Bonferroni post-tests. (B-D) C57BL/6 mice received intraperitoneal PBS or 20 mg/Kg isunakinra (n = 7/group) followed 1 h later by 10 6 intrapleural LLC cells stably expressing a κB.LUC reporter (NGL), and were imaged for bioluminescence 4 h later. Shown in (B) are representative chest (dotted lines) bioluminescent images with pseudo color scale. In addition, FVB mice (n = 48) received 2 × 10 5 intrapleural FULA1 cells, were allowed 5 days for tumor take and received daily intraperitoneal PBS or 20 mg/Kg isunakinra. Mice were sacrificed when morbid for survival analyses (n = 17/treatment) or at day 14 post-tumor cells for malignant pleural effusion (MPE) analyses (n = 7/treatment). Shown in (C) are Kaplan-Meier survival estimates (curves) with log-rank probability (P) and hazard ratio (HR), and in (D) data summary of chest bioluminescence and MPE volume, shown as raw data points (circles), medians (dashed lines), quartiles (dotted lines), kernel density distributions (violins), and probability (P), unpaired Student's t-test.

Discussion
Here, we show how KRAS-mutant tumors are dependent on IL-β provided by tumorassociated macrophages. Importantly, we show that tumor-secreted versican causes IKKβ activation in myeloid cells to foster this pro-inflammatory circuitry. Notwithstanding cancers with other mutations and other myeloid cells like neutrophils and mast cells that might also fuel tumors with IL-1β, we define here a non-oncogene addiction of KRAS and IL-1β, in tandem with their partners in crime VCAN and IKKβ. The findings stress the need for molecular stratification of current clinical trials of IL-1β inhibition against lung cancer. Unique experimental models for the study of tumor genome-host immunity interactions are provided, and novel diagnostic platforms and prognostic biomarkers are described for further validation.
Although sotorasib was recently approved in the U.S. against KRAS G12C -mutant NSCLC [62], KRAS-mutant cancers from multiple sites of origin remain notoriously aggressive and undruggable [63] and direct KRAS inhibition is associated with some toxicity that likely renders such treatments unsuitable for chemoprevention [64]. On the contrary, anti-IL-1β-directed therapies hold promise for chemoprevention, as shown by the CANTOS trial, (where tri-monthly administration of the IL-1β-neutralizing antibody canakinumab over 3.7 years of observation decreased overall and lung cancer mortality by 51% and 77%, respectively) based on their excellent safety profile [2]. The pro-inflammatory interplay between VCAN in tumor cells and IKKβ in macrophages described here is not only mecha-nistically intriguing, but also promising for innovations in cancer therapy and diagnosis. We identify cancer cell VCAN and myeloid IKKβ as the accomplices of KRAS that trigger secretion of IL-1β in the milieu of KRAS-mutant cancers. The results position these cancers as favorable candidates for anti-IL-1β therapy, and versican as a diagnostic and prognostic biomarker, as well as a therapeutic target in this tumor category that comprises 9% of all human cancers, alone or in combination with anti-IL-1β agents. In addition, since early diagnosis of metastasis is key to effective cancer therapy [57], VCAN can serve as a biomarker of metastasis. This might be achieved by monitoring local or systemic VCAN levels in patients at risk, or by using our NF-κB-reporter macrophages as a diagnostic platform. Indeed, our data indicate that the latter can accurately discriminate pleural metastasis from other pleural inflammatory processes, highlighting the clinical relevance of our findings.
NF-κB signaling in cancer and myeloid cells impacts modes of tumor progression and metastasis in various tumor types and is intimately addicted with oncogenic KRAS signaling [65,66]. However, the lessons learnt from clinical trials of proteasome (and hence also canonical NF-κB pathway) inhibitors against multiple myeloma dictate that therapeutic interventions into the NF-κB pathway are also associated with significant toxicity, since the pathway acts simultaneously in epithelial and immune cells in opposing fashions [67,68]. In addition to previous work elucidating the oncogenic functions of IKKβ in tumor cells [6,11,22,34,53,65,66], here we show how myeloid IKKβ functions to fuel tumor cell NF-κB signaling with IL-1β, further emphasizing the complex and multifaceted pro-tumor functions of NF-κB and the need for its therapeutic targeting against cancer.

Conclusions
In conclusion, KRAS-mutant cancers rely on host IL-1β, which they elicit from host macrophages via secretory versican that activates myeloid IKKβ. This inflammatory loop provides multiple opportunities for improved diagnosis, prognostication, and identification of therapeutic vulnerabilities of KRAS-mutant cancers.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/cancers15061866/s1, Figure S1: Seven murine cell lines with different Kras alleles and transcriptional control of interleukin (IL)-1β by nuclear factor (NF)-κB; Figure S2: Expression of IL1B in correlation with KRAS and the macrophage marker ADGRE1 in human tumors; Figure S3: Expression of IL1B in correlation with lineage-specific markers in human tumors; Figure S4: NF-κB activation in pleural metastases of NGL mice; Figure S5: NF-κB activation in pleural metastases of NGL mice; Figure S6: NF-κB activation in metastasis-associated macrophages; Figure S7: NF-κB activation in metastasis-associated macrophages; Figure S8: No impact of mast cells on the host NF-κB response to pleural metastasis and decreased intensity of the host NF-κB response to heterotopic tumor growth; Figure S9: Requirement for mutant Kras signaling for host NF-κB activation during pleural metastasis; Figure S10: Adoptive bone marrow transplants determine host NF-κB response to pleural metastasis; Figure S11: Pharmacologic and genetic macrophage ablation abolishes pleural metastasis; Figure S12: Uncropped immunoblots; Figure S13: Toll-like receptor (TLR) expression by murine bone marrow-derived macrophages (BMDM) by microarray; Figure S14: Versican as a potential diagnostic and prognostic biomarker of KRAS-mutant human cancers; Figure S15: Versican over-expression by KRAS-mutant human cancers predicts poor survival; Figure S16: KRAS, VCAN, and IKBKB alterations in human cancers; Figure S17: VCAN and IKBKB alterations in lung adenocarcinoma (LUAD), colon adenocarcinoma (COAD), rectal adenocarcinoma (READ), and uterine corpus endometrial carcinoma (UCEC); Table S1: Differential gene expression of BMDM-specific transcripts after incubation with tumor-conditioned media; Table S2: Differential gene expression of Kras-mutant cancer cells; Table S3: Differential gene expression of BMDMs lacking NF-κB signaling; Table S4: PCR primers used in this study; Table S5: Lentiviral shRNA pools used in this study; Research Data S1: *.xlsx file of proteomic analysis of secreted proteins of Kras-mutant and -wild-type cancer cells.