Long Noncoding RNA MIR4435-2HG Suppresses Colorectal Cancer Initiation and Progression By Reprogramming Neutrophils

Long noncoding RNAs (lncRNA) are a class of over 200 nucleotide transcripts without protein-coding capacity (1). Various studies demonstrate the vital roles of lncRNAs in tumorigenesis, proliferation, metastasis, angiogenesis, and regulation of the immune microenvironment (2, 3). As the third leading cause of malignant cancer in the world (4), colorectal cancer is regulated at the initiation and progression stages by the aberrant expression and mutations of lncRNAs. Our previous studies demonstrate that LINC01133 can inhibit colorectal cancer epithelial–mesenchymal transition and metastasis by interacting with Serine And Arginine Rich Splicing Factor 6 (SRSF6; ref. 5), and high expression of the lncRNA CYTOR in tumor cells drives colorectal cancer progression (6). We also found that another novel lncRNA, MIR4435-2HG, which is highly homologous to CYTOR (7), is upregulated in colorectal cancer tumor tissue. An increasing number of publications have since then reported MIR4435-2HG as an oncogene in multiple tumor types, such as lung cancer (8, 9), hepatocellular carcinoma (10, 11), gastric cancer (12), breast cancer (13, 14), prostate cancer (15, 16), and colorectal cancer (17–19). These results were derived from analyses of clinical tissue samples or public databases and gain- or loss-of-function assays in in vitro tumor cell lines, but the roles of MIR4435-2HG in the tumor stroma or microenvironment, and its detailed regulatory mechanisms, remain unknown.

Myeloid-derived suppressor cells (MDSC) are a heterogeneous population of cells, including pathologically activated monocytes and relatively immature neutrophils, that are generated from many pathological conditions, such as infection, inflammation, and cancer. MDSCs consist of polymorphonuclear MDSCs (PMN-MDSC) and monocyte MDSCs (M-MDSCs), whose morphology and phenotype are similar to those of neutrophils and monocytes, respectively (20, 21). PMN-MDSCs are the prevalent population of MDSCs in most types of cancer, including lung, breast, colon, and pancreatic cancer (22, 23). The salient feature of MDSCs is their inhibition of T-cell proliferation and activation by inducing arginase-1 (Arg1), reactive oxygen species production, and nitric oxide production, which culminate in the promotion of tumor progression (24, 25). MDSCs have also been reported to produce prostaglandin E2, calcium-binding proteins S100A8/S100A9, matrix metalloproteinases, IL10, TGFβ, and other cytokines to promote angiogenesis, tumor proliferation, and metastasis (26). Clinical investigation has shown that the accumulation of MDSCs positively associates with advanced cancer stage and metastatic burden in patients with colorectal cancer (27, 28).

So far, some lncRNAs, such as PVT1 (29), lnc-C/EBPβ (30), and lnc-chop (31), have been identified to regulate the immunosuppressive activity of MDSCs. However, the roles of lncRNAs in MDSCs have still not been clarified in detail. In this study, our findings repositioned MIR4435-2HG as a tumor-suppressing lncRNA in colorectal cancer. Mir4435-2hg depletion increased tumor-infiltrating PMN-MDSCs and promoted lipid accumulation in MDSCs, thus enhancing the immunosuppressive function of MDSCs and promoting tumorigenesis and progression in both colitis-associated and Apc^min/+ spontaneous colorectal cancer mouse models. These findings imply that we need to reexamine the conclusions drawn from previous MIR4435-2HG expression data and experiments in vitro.

Colorectal cancer cell lines DLD1, HCT8, HT29, HCT116, RKO, SW480, SW620, and CT26 were purchased from the ATCC in 2011. The MC38 cell line was kindly provided by Cell Bank/Stem Cell Bank, Chinese Academy of Sciences (Shanghai, China) in 2015. DLD1, HCT8, HCT116, HT29, SW480, SW620, CT26 cells were cultured in RPMI1640 (Gibco) and RKO, MC38 cells were cultured in DMEM (Gibco) medium according to the manufacturer's instruction, with 10% FBS (Gibco) and 1% penicillin–streptomycin (Biosharp, catalog no. BL505A) in a humidified atmosphere at 37°C with 5% CO₂. All cells were frozen at passages 2 to 5 after purchase. Experiments were performed using passages 3 to 20 after removal from liquid nitrogen. Mycoplasma was test routinely using GMyc-PCR Mycoplasma Detector Kit (Yeasen). The most recent cell line authentication was in September 2018 by short tandem repeat (STR) analysis in Forensic Science Center, Zhejiang University, Hangzhou, China.

For overexpressing MIR4435-2HG in DLD1 and HCT116 or Mir4435-2hg in MC38 and CT26 cells, plasmids containing MIR4435-2HG (NR_015395.2) or Mir4435-2hg variant 1 (NR_028589.1), variant 2 (NR_028590.1), variant 3 (NR_028591.1), respectively, were constructed on the basis of pCDH empty backbone (WZ Biosciences Inc.) using ClonExpress II One Step Cloning Kit (Vazyme, catalog no. C112). pCDH empty vector was used as control. Plasmids were then transfected into DLD1, HCT116, MC38 or CT26 by lipo2000 (Invitrogen, catalog no. 11668019) according to the manufacturer's instructions. siRNAs targeting MIR4435-2HG (synthesized in Gemma, Supplementary Table S1) were transfected into SW480 and RKO cells using the Genmute siRNA transfection reagent (SignaGen, catalog no. SL100568) according to the manufacturer's instructions. The overexpressing or silencing effects were confirmed by RNA isolation and quantitative PCR, as detailed below.

To construct MIR4435-2HG-deficient SW480 cells and Mir4435-2hg–deficient MC38 cells, 2 μg pSpCas9-BB-2A-GFP plasmid containing two sgRNAs (purchased from Genscript; sgRNA sequences are listed in Supplementary Table S1) were transfected to cells at 50% confluence in a 6-well plate by Lipofectamine 2000 according to the manufacturer's instructions. After one week, single GFP⁺ cells were selected by Beckman moflo Astrios EQ into wells of a 96-well plate. These single-cell clones were cultured as detailed above, and knockout was confirmed by PCR and RT-qPCR at the genome and transcriptional level. The clones without editing by Cas9 were used as mock cells. sgRNA sequences and PCR primers are listed in Supplementary Table S1.

Mir4435-2hg^+/−, Mir445-2hg^flox/+, Villin-Cre, S100a8-Cre mice were constructed through CRISPR/Cas9 and pronuclear microinjection techniques by GemPharmatech Co., Ltd, as indicated in Supplementary Figures. Apc^min/+ mice were a gift from Dr. Rongpan Bai, Zhejiang University School of Medicine. Primer sequences for PCR genotyping are listed in Supplementary Table S1. All mice were C57BL/6J background which were bred and maintained under pathogen-free conditions at the laboratory animal center, Zhejiang University. All mice were monitored every day by the staffs of the laboratory animal center and twice one week by the authors. The protocol (ZJU20160023) for this study was approved by the Institutional Animal Care and Use Committee at Zhejiang University.

For the azoxymethane (AOM)/dextran sulfate sodium (DSS)-induced colorectal cancer model, 6-week male wild-type (WT), Mir4435-2hg^−/−, Mir445-2hg^flox/flox, Mir445-2hg^flox/flox Villin-Cre, Mir445-2hg^flox/flox S100A8-Cre mice were injected intraperitoneally with 10 mg/kg AOM (Sigma-Aldrich, catalog no. A5486), and after one week, they were fed on and off 1.5% DSS (MPbio, catalog no. 160110) in the drinking water. One week of 1.5% DSS in the drinking water, followed 2 weeks of normal water constituted one DSS cycle. As indicated figures, mice were euthanatized after two DSS cycles for the short-term cohort or after three DSS cycles for the classical term cohort. For the long-term cohort, mice were fed with normal water for another 15 weeks after three DSS cycles before euthanasia. In the cohort for survival analysis, mice were fed with normal water after three DSS cycles until death. The colon was harvested, cut open along the main axis, washed with saline, and tumor numbers were counted. Tumor size was defined as the tumor diameter measured by a caliper, and tumor burden was defined as the summation of total tumor volume. Histologic examination was performed on paraffin-embedded sections after hematoxylin and eosin (H&E) staining, as detailed below.

For spontaneous intestinal adenomatous polyposis, Apc^min/+ mice were sacrificed at 4 months old. Small intestines and colons were longitudinally cut open to count tumor numbers. Histologic examination was performed on paraffin-embedded sections after H&E staining, as described below.

For subcutaneous tumor models, male WT, Mir4435-2hg^−/−, Mir445-2hg^flox/flox, Mir445-2hg^flox/flox S100A8-Cre mice aged 6–8 weeks were subcutaneously injected in the right flank with 1×10⁵ WT MC38 cells. For Supplementary Fig. S5D, 1×10⁶, 1×10⁴, or 2×10³ WT MC38 cell were injected as indicated in the figure. In Fig. 2H, 1×10⁵ mock or KO MC38 cells were injected as indicted in the figure. Tumor length and width were measured with a Vernier caliper every three days. Tumor volume was calculated as length × width²/2.

For verifying tissue-specific depletion of Mir4435-2hg, neutrophils and intestinal epithelial cells were collected from naïve Mir445-2hg^flox/flox, Mir445-2hg^flox/flox Villin-Cre and Mir445-2hg^flox/flox S100A8-Cre mice. Neutrophils were collected from bone marrow (as described below) by flow cell sorter (Beckman moflo Astrios EQ) as labeled CD11b⁺Ly6G⁺ cells (antibodies list in Supplementary Table S2). Intestinal epithelial cells were collected by digesting colons using striping buffer containing 5 mmol/L EDTA (Invitrogen, catalog no. AM9260G), 1 mmol/L DTT (Beyotime, catalog no. ST041), 5% FBS in HBSS (Beyotime, catalog no. C0219), in 37°C 150 rpm for 40 minutes in the MQW-63R shaker. RNA isolation and quantitative PCR were then carried out as detailed below to verify the Mir4435-2hg depletion.

Brain, heart, kidney, liver, lung, spleen, stomach, small intestine, colon tissues from naïve 8-week male WT and Mir4435-2hg^−/− mice, colon tissues from AOM/DSS-induced colorectal cancer models, and small intestine and colon tissues from 4-month Apc^min/+ mice were fixed with 10% neutral buffered formalin and paraffin embedded. After being sectioned, deparaffinized, and rehydrated, 4-μm-thick sections were stained with hematoxylin (Sigma-Aldrich, catalog no. H9627) for 10 minutes and eosin (Sigma-Aldrich, catalog no. 230251) for 2 seconds at room temperature. The slides were scanned using the NanoZoomer digital slice scanner (Hamamatsu). Tissue alterations were determined via tissue morphology, structure and cell morphology by a pathologist.

Four-micron-thick paraffin-embedded colon sections from the AOM/DSS-induced colorectal cancer models were deparaffinized in xylene (Sinopharm) and rehydrated through a graded series of ethanol solutions (Sinopharm). After blocking endogenous peroxidase activity in 3% hydrogen peroxide (ZSGB-BIO) at room temperature for 15 minutes, the tissue sections were treated with 0.01 mol/L citrate buffer (pH 6.0; OriGene) under high pressure in a pressure cooker for 3 minutes to complete antigen retrieval. Blocking was performed with 10% bovine serum (Gibco) for 30 minutes at room temperature, and then the sections were incubated with primary antibodies to CD8a (1:500) and MRP8 (1:1,000) overnight at 4°C (Supplementary Table S2), followed by incubation with 100 μL goat anti-rabbit secondary antibody (ZSGB-BIO, catalog no. PV6001) at room temperature for 30 minutes. Staining was visualized with 3,3′-diaminobenzidine (DAB) (ZSGB-BIO, catalog no. PV8000), and sections were counterstained with hematoxylin, dehydrated, and covered with a coverslip. The slides were scanned using the NanoZoomer digital slice scanner (Hamamatsu). Image analysis and quantitation were performed with ImageJ (NIH, Bethesda, MD).

WT MC38 cells in logarithmic phase were trypsinized, washed twice to remove residual FBS, and then stained with 5 μmol/L CFSE (eBioscience, Cat# 65–0850) at room temperature for 10 minutes. Labeling was stopped by adding 4 to 5 volumes of cold 1640 medium containing 10% FBS, and then cells were incubated on ice for 5 minutes. After washing three times with complete medium, 2×10⁵ CFSE-labeled MC38 cells were plated in 12-well plates with splenocytes isolated WT or Mir4435-2hg^−/− mice after erythrocytes were lysed using RBC Lysis Buffer (Santa Cruz Biotechnology, catalog no. sc-296258). After incubating for 40 hours, both splenocytes and tumor cells were collected for flow cytometry analysis, as described below.

Tibias and femurs from WT and Mir4435-2hg^−/− mice were removed, and bone marrow (BM) cells were flushed from the bones in PBS containing 5% FBS. Erythrocytes were removed using RBC Lysis Buffer. A total of 1 × 10⁶/mL BM cells were plated in 6-well plates in RPMI1640 complete medium containing 10% FBS, 1% penicillin–streptomycin, and 20 ng/mL GM-CSF (BioLegend, catalog no. 576304). 25% (v/v) MC38/CT26 tumor cell culture medium was added as required, which was prefiltered using 0.22-μm filters (Millipore). After 3 days, BM-derived MDSCs were collected for flow cytometry analysis, RNA isolation, and suppressive function assays, as detailed below.

Cells used for flow cytometry analysis were isolated from the indicated tissues. Peripheral blood was collected from the retro-orbital plexus or submandibular vein plexus. Splenocytes were collected by dissociating the spleens with sterile forces and passing through a 40-μm cell strainer (Biosharp). BM cells were collected from tibias and femurs as described above. Erythrocytes in samples were lysed using RBC Lysis Buffer. Tumor tissues from mice were cut into small pieces and digested in RPMI1640 medium supplemented with 5% FBS, 1 mg/mL collagenase IV (Sigma-Aldrich, catalog no. C5138), 20 μg/mL hyaluronidase (Solarbio, catalog no. H8030) for 2 hours at 37°C in a shaker. Single-cell suspensions were obtained by passing through a 40-μm nylon cell strainer. Single-cell suspensions were first stained with Fc block (Biolegend, catalog no. 101320) and then treated with the indicated antibodies for 30 minutes in 4°C in the dark. All antibodies used are listed in Supplementary Table S2. 7-AAD (BD Biosciences, catalog no. 559925) was used to exclude dead cells. For intracellular staining, the Fix & Perm Kit (Multi Science, catalog no. GAS005) was used following the manufacturer's instructions. For lipid staining after surface staining, cells were resuspended in 500 μL of BODIPY 493/503 (Invitrogen, catalog no. D3922) at 1 μg/mL for 20 minutes at room temperature in the dark, and then washed three times with PBS. For reactive oxygen species (ROS) analysis, surface-stained cells were incubated with 10 μmol/L DCFH-DA probe (Beyotime, catalog no. S0033S) for 20 minutes at 37°C, and then washed three times with PBS before detection. Cells were run on Beckman CytoFLEX LX, and the data were analyzed by FlowJo 10.4 (Tristar).

Total RNA from cells or mouse colon tissues were isolated using TRIzol reagent (Invitrogen, catalog no. 15596018). Mouse colon tissues were homogenized by grinding in liquid nitrogen. For nuclear and cytoplasmic RNA fractionation, the PARIS Kit (Ambion, catalog no. AM1921) was used according to the manufacturer's instruction. cDNA was synthesized by HiScript II Reverse Transcriptase (Vazyme, catalog no. R223–01) and quantitative PCR analysis was performed by SYBR qPCR Master Mix (Vazyme, catalog no. Q711–02) according to the manufacturer's instruction in LightCycler 480 (Roche). GAPDH was detected in each experimental sample for normalization. Fold-change was calculated using the 2^−ΔΔC_t method. Three independent experiments were performed in triplicate. Genes and primers are listed in Supplementary Table S1.

Cell proliferation was measured by CCK8 reagent (Boster, catalog no.AR1160) according to the manufacturer's instructions. Briefly, 2,000 cells were seeded in 96-well plates and incubated with 100 μL culture medium. An aliquot of 10 μL CCK8 was added and incubated for 2 hours. The absorbance at 450 nm was measured to calculate the numbers of viable cells. Migration and invasion assays were performed with transwell and Matrigel chamber plates (24-well format; 8-μm pore size; Corning Costar, catalog no. 3422) as described previously (32). ImageJ software was used to measure the cell area in the bottom chamber to quantify the cells that migrated across the filter. Each measurement was performed in triplicate, and the experiments were repeated three times.

Splenocytes were isolated from spleens of wild-type C57BL/6 mice, and erythrocytes were lysed by RBC Lysis Buffer. Splenocytes were stained with 5 μmol/L CFSE (eBioscience, catalog no. 65–0850), and then 3×10⁵ CFSE-labeled splenocytes were cocultured with 3×10⁴ WT or Mir4435-2hg^−/− MDSCs in a round-bottom 96-well plate pre-coated with 5 μg/mL anti-CD3 (eBioscience, catalog no. 14–0031) and 5 μg/mL anti-CD28 (eBioscience, catalog no. 14–0281) at 4°C overnight. After coculturing for 72 hours, T-cell proliferation was assessed by CFSE dilution using Beckman CytoFLEX LX, and data were analyzed in FlowJo 10.4.

The Gene Expression Omnibus (GEO) databases were systematically and comprehensively searched (up to October 16, 2017). The search details were ‘(colorectal [All Fields] AND (adenoma [MeSH Terms] OR adenoma [All Fields]) AND 9[n_samples]:1000[n_samples]) AND “Homo sapiens”’. The inclusion criteria were as follows: (i) datasets had paired clinical sample tissues; (ii) experiment type was expression profiling. The exclusion criteria were as follows: (i) spotted cDNA (two-channel ratio data) array; (ii) patient had antitumor treatments. All available data were included in processing and analysis. The selected gene expression data were downloaded from the GEO database in SOFT format. The SOFT files were reformatted to a table separate expression matrix using a custom C script (https://github.com/wenjie1991/GEO_parser). The expression data were annotated by the gene information from NCBI (ftp://ftp.ncbi.nlm.nih.gov/gene/DATA/GENE_INFO/Mammalia/Homo_sapiens.gene_info.gz). The Cancer Genome Atlas (TCGA) colorectal cancer dataset was downloaded from UCSC Xena (https://xenabrowser.net/). ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data; ref. 33) R package was used to calculate the immune score in R software.

Five colorectal cancer patient samples were used in the RNAscope assay, which were obtained from Sir Run Run Shaw Hospital, Zhejiang University in 2019 after informed consent by patients. The samples were formalin-fixed immediately after removal from patients and then dehydrated and paraffin-embedded. The collection and use of patient samples were approved by the Ethics Committee of department of Medicine, Zhejiang University (2016007).

The probe for MIR4435-2HG was synthesized by Advanced Cell Diagnostics (ACD lnc.). Detection of MIR4435-2HG in formalin-fixed paraffin-embedded (FFPE) human colorectal cancer tissues was performed using the RNAscope 2.0 HD Reagent Kit-Brown (ACD, catalog no. 322310) following the manufacturer's instructions. Briefly, 4-μm FFPE sections were deparaffinized, boiled with target retrieval reagent for 15 minutes, and then digested by protease at 40°C for 30 minutes, followed by hybridization for 2 hours at 40°C with Probe-MIR4435-2HG. After six steps of amplification, the probe was visualized with DAB, and cell nuclei were counterstained with hematoxylin. The slides were evaluated according to manufacturer's instructions by a pathologist.

RNA of colorectal tumor tissues from WT and Mir4435-2hg^−/− long-term AOM/DSS mice, 4-month Apc^Min/+/WT, Apc^Min/+/ Mir4435-2hg^−/− mice, and neutrophils from bone marrow of 8-week WT and Mir4435-2hg^−/− mice were extracted, sequenced, and analyzed by Bioacme. The tumor tissues were stored in RNA Keeper (Vazyme, catalog no.R501) and neutrophils were stored in TRIzol in −80°C and then shipped under the protection of dry ice. Three biological replicates were used. Gene differential expression analysis was performed using the Cuffdiff program in Cufflinks package (https://github.com/cole-trapnell-lab/cufflinks) (34). The Benjamini–Hochberg false discovery rate method was applied to correct for multiple hypothesis testing. The genes with adjust P < 0.05, fold change > 1.5 or < 0.67 were defined as differentially expressed genes and were candidates for further analysis. A gene expression heatmap was generated with Heatmapper (http://www.heatmapper.ca/). Gene ontology enrichment analysis was performed using the online tool DAVID (https://david.ncifcrf.gov/). The results were visualized using the R package ggplot2 in R software. Gene set enrichment analysis (GSEA) was applied to the H (hallmark) gene sets. The gene set collection database MSigDBv7.2 (http://www.gsea-msigdb.org/gsea/downloads_archive.jsp) was used and sets with size of 15 to 500 were selected. Permutation of genes was used to generate null distribution, and all other parameters were kept as default.

A two-tailed Student t test was used to test for significant differences between two groups in GraphPad v6.0 (GraphPad Software) or SPSS v23 (SPSS Inc.). Kaplan–Meier survival analysis was performed using the software IBM SPSS Statistics v23 with the Log-rank (Mantel–Cox) test. Data in this article are presented as mean ± SD. P-value <0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

All data that support the findings of this study are available from the corresponding authors upon reasonable request. The raw RNA-sequencing (RNA-seq) data have been submitted to the National Center for Biotechnology Information (accession number: PRJNA777810, PRJNA779546, PRJNA779548).

To explore the function of MIR4435-2HG in colorectal cancer, we analyzed 10 public datasets from The Cancer Genome Atlas (TCGA) and eligible Gene Expression Omnibus (GEO) databases (including more than 10 paired normal–tumor colorectal cancer samples), which showed that MIR4435-2HG was upregulated in tumor tissues compared with paired adjacent normal tissues in 7 datasets but not in the other 3 datasets (Supplementary Fig. S1A). Therefore, we performed a meta-analysis to confirm that MIR4435-2HG was overexpressed in colorectal cancer (Fig. 1A). Survival meta-analysis conducted on 8 datasets revealed no significant relationship between the expression of MIR4435-2HG and overall patient survival when we set the cut-off value as the Youden index or median (P50; Supplementary Fig. S1B–S1E).

Next, we detected the expression and subcellular localization of MIR4435-2HG in colorectal cancer cell lines (Supplementary Fig. S2A). MIR4435-2HG had relatively high expression in both the nucleus and cytoplasm of SW480 and RKO cells (Supplementary Fig. S2B). We knocked-out MIR4435-2HG through CRISPR/Cas9 in SW480 cells (Supplementary Fig. S2C) and knocked-down MIR4435-2HG by siRNA in SW480 and RKO cells (Supplementary Fig. S2D). Neither knockout nor knockdown of MIR4435-2HG in colorectal cancer cells had an effect on proliferation (Supplementary Fig. S2E and S2F), migration, or invasion (Supplementary Fig. S2G and S2H). Consistent with these findings, ectopic expression of MIR4435-2HG in DLD1 and HCT116 cells (Supplementary Fig. S2I) did not change their proliferation (Supplementary Fig. S2J), migration, or invasion (Supplementary Fig. S2K). RNA scope showed high expression of MIR4435-2HG in the tumor stroma but very low expression in colorectal cancer cells and adjacent normal epithelial cells (Fig. 1B), which might clarify why MIR4435-2HG had high expression in tumor tissues but had no biological effect within colorectal cancer cells.

It remains unknown whether MIR4435-2HG in the tumor stroma plays biological roles in colorectal cancer development. We generated Mir4435-2hg–deficient mice by deleting the Mir4435-2hg locus through CRISPR/Cas9 (Supplementary Fig. S3A–3C). Consist with a former report (35), loss of Mir4435-2hg led to a decline of neutrophils, monocytes, and eosinophils but not lymphoid cells (Supplementary Fig. S3D–S3F). No other obvious alterations in major organ development or dyslipidemia were observed after deleting Mir4435-2hg (Supplementary Fig. S4A and S4B). We then divided the mouse colitis-associated (AOM/DSS) colorectal cancer model into three cohorts, including short-term, classical term, and long-term tumor models (Fig. 1C), which resemble human colorectal cancer progression at the molecular level, including Apc mutations and β-catenin translocation (36, 37). Although the adenoma in the colorectum from the short-term tumor model was not large enough for gross observation, H&E staining showed that Mir4435-2hg^−/− mice carried more colorectal adenomas and lymphoid follicles than wild-type (WT) mice (Fig. 1D–F). Through gross observation and H&E staining of the classical term model, we found more tumors, larger tumor sizes, and higher tumor burdens in Mir4435-2hg^−/− mice than WT mice (Fig. 1G–K). However, neither gross observation nor H&E staining of the long-term model presented a significant difference in the tumor number or tumor size between Mir4435-2hg^−/− mice and WT mice (Fig. 1L–O). Further histologic analysis (Fig. 1M) showed that all tumors from Mir4435-2hg^−/− mice, but only 40% from WT mice, had penetrated through the muscularis mucosae into the submucosa (Fig. 1P). Another cohort of the AOM/DSS model was subjected to survival analysis (at the 30th week), and the Mir4435-2hg^−/− mice exhibited worse survival (Fig. 1Q).

Next, Apc^Min/+ mice, an animal model of spontaneous intestinal adenomatous polyposis (38), were crossed with Mir4435-2hg^−/− mice to mimic spontaneous tumors. Higher Mir4435-2hg expression was detected in intestinal tumors than in normal tissues from Apc^Min/+ mice (Supplementary Fig. S4C). Nevertheless, more intestinal tumors were observed in Apc^Min/+/Mir4435-2hg^−/− mice (Fig. 1R; Supplementary Fig. S4D and S4E). We also found that Apc^Min/+/Mir4435-2hg^−/− mice had worse prognosis than in Apc^Min/+/WT mice (Fig. 1S). Taken together, these data suggest that Mir4435-2hg depletion contributes to sporadic colorectal tumorigenesis and progression and should be considered a tumor suppressor lncRNA.

To demonstrate how Mir4435-2hg regulated colorectal cancer progression, we performed transcriptome analysis on tumor tissues from WT and Mir4435-2hg^−/− AOM/DSS mice by RNA-seq. Mir4435-2hg deletion caused 92 genes to be significantly upregulated and 138 genes to be significantly downregulated (Fig. 2A), which were further validated for their representative genes by qRT-PCR (Supplementary Fig. S5A). These differentially expressed genes were mainly enriched in immune response biological processes, including inflammatory response, neutrophil chemotaxis, immune system process, antigen processing and presentation, according to Gene Ontology (GO) enrichment analysis (Fig. 2B and C).

To exclude the side effects caused by AOM/DSS-induced colitis, we also collected tumor tissues from Apc^Min/+/Mir4435-2hg^−/− and Apc^Min/+/WT mice for RNA-seq analysis. Consistent with the AOM/DSS model, GO enrichment analysis of the upregulated genes showed that most genes associated with tumor immune-related biological processes, such as lymphocyte chemotaxis, cellular response to interferon-gamma, inflammatory response, and neutrophil chemotaxis (Supplementary Fig. S5C). These data further support that Mir4435-2hg might function as a tumor suppressor by regulating the immune microenvironment of colorectal cancer.

To test this hypothesis, the syngeneic murine colorectal cancer cell line MC38 was subcutaneously inoculated into Mir4435-2hg^−/− and WT mice. We found that tumors grew more rapidly in Mir4435-2hg^−/− mice (Fig. 2D). The volume and weight of tumors in Mir4435-2hg^−/− mice were also significantly larger than those in WT mice (Fig. 2E and F), and the tumor formation rate in Mir4435-2hg^−/− mice was significantly higher than that in WT mice when the inoculated cells were diluted at different concentrations (Supplementary Fig. S5D).

To further demonstrate that Mir4435-2hg in the tumor stroma, but not in tumor cells, regulated colorectal cancer development, we knocked-out Mir4435-2hg in MC38 cells with CRISPR/Cas9 (Supplementary Fig. S5E). Mir4435-2hg knockout did not change their proliferation, migration, or invasion (Supplementary Fig. S5F and S5G). Overexpression of three Mir4435-2hg variants in MC38 and CT26 cells also had little effect on proliferation and migration (Supplementary Fig. S5H–S5J). Mir4435-2hg knockout MC38 and mock knockout cells were synchronously injected into Mir4435-2hg^−/− and WT mice. Knocking out Mir4435-2hg in MC38 cells did not affect tumor proliferation in either Mir4435-2hg^−/− or WT mice; however, the microenvironment of Mir4435-2hg^−/− mice significantly promoted tumor growth, not only of mock but also of Mir4435-2hg knockout MC38 cells (Fig. 2G). A similar phenotype was also observed for tumor volume and weight (Fig. 2H and I). Next, we separated splenocytes from Mir4435-2hg^−/− and WT mice to coculture with MC38 cells. As expected, the apoptosis of MC38 cells was greater when they were cocultured with splenocytes, but the MC38 cell apoptosis rate was decreased by the splenocytes from Mir4435-2hg^−/− mice compared with WT mice (Fig. 2J and K).

We also reanalyzed survival data from TCGA and GEO databases. The ESTIMATE algorithm was used to calculate the immune score for each sample, and samples with an immune score higher than 1,000 were selected for survival analysis. In these high immune infiltrate samples, higher expression of MIR4435-2HG tended to associate with a better prognosis, although not statistically significant (Supplementary Fig. S6). Overall, these data indicated that MIR4435-2HG might play an anticancer role by remodeling the colorectal cancer immune microenvironment.

Mir4435-2hg can control the lifespan of eosinophils, neutrophils, and monocytes through allele-specific suppression of Bim expression (35). However, it remains unknown which type of cell in the colorectal cancer stroma expresses Mir4435-2hg. Therefore, we performed an RNAscope assay in human colorectal cancer tissues, which showed that MIR4435-2HG was mainly located in neutrophils (Fig. 3A). Next, we monitored the number variations of main leukocytes in peripheral blood during the growth of subcutaneous tumors. Consistent with previous reports (39), neutrophils, monocytes, and eosinophils were increased and lymphocytes were decreased in tumor-bearing mice. Monocytes and eosinophils in Mir4435-2hg^−/− mice remained lower than those in WT mice, but the number of neutrophils in Mir4435-2hg^−/− mice gradually reached the same frequency as that in WT mice as tumors grew (Fig. 3B). A similar phenotype was observed in splenocytes (Supplementary Fig. S7A).

Neutrophils in tumor-bearing mice are heterogenetic, including both classical neutrophils and PMN-MDSCs (40, 41). From our data above, we assumed that Mir4435-2hg could not control the lifespan of PMN-MDSCs and that the elevated neutrophils in tumor-bearing Mir4435-2hg^−/− mice were mainly PMN-MDSCs. We, thus, detected the apoptosis of neutrophils in tumor-free and tumor-bearing WT and Mir4435-2hg^−/− mice. Tumor-free Mir4435-2hg^−/− mice had significantly more neutrophil apoptosis than the WT mice. However, in tumor-bearing mice, neutrophil apoptosis was reduced in both Mir4435-2hg^−/− and WT mice, which may result from the GM-CSF produced by MC38 cells (42), and in tumor-bearing Mir4435-2hg^−/− mice, neutrophil apoptosis rate was reduced to the same level as that in tumor-bearing WT mice (Fig. 3C and D). In BM-derived PMN-MDSCs, we also observed a decreased apoptosis rate after treatment with MC38 tumor cell culture medium, especially in Mir4435-2hg^−/− cells (Fig. 3E).

Because Mir4435-2hg regulates neutrophil apoptosis by suppressing Bim expression, we detected the expression of BIM protein in neutrophils. BIM was upregulated in both tumor-free and tumor-bearing Mir4435-2hg^−/− mice, and bearing tumors did not alter the expression of BIM (Fig. 3F and G; Supplementary Fig. S7B). The Bcl2 family regulates apoptosis by forming homodimers or heterodimers. Therefore, we detected alterations in all anti-apoptotic Bcl2 family members. Among them, only Bcl2l1 in BM-derived PMN-MDSCs from either WT or Mir4435-2hg^−/− mice was upregulated by stimulation with MC38 tumor cell culture medium (Fig. 3H), whereas Bcl2 and Bcl2a1a were downregulated, and Bcl2l2 and Mcl1 were unaltered (Supplementary Fig. S7C). Flow cytometry also showed elevated expression of BCL2L1 in neutrophils from both WT and Mir4435-2hg^−/− tumor-bearing mice (Fig. 3I and J). These results indicated that the elevated expression of BCL2L1 in PMN-MDSCs might antagonize the function of the increased BIM in Mir4435-2hg^−/− mice, helping the neutrophils in tumor-bearing Mir4435-2hg^−/− mice recover to the same level as in WT mice and leading to a higher proportion of PMN-MDSCs.

Single-cell RNA-seq analysis (43) has identified CD14 as a marker of PMN-MDSCs for distinguishing classical neutrophils in tumor-bearing mice. In our study, a significantly higher proportion of CD14⁺ cells was observed in splenic neutrophils from tumor-bearing Mir4435-2hg^−/− mice than in those from WT mice. Even in tumor-free mice, Mir4435-2hg^−/− mice also carried a higher proportion of CD14⁺ neutrophils than WT mice (Fig. 3K and L). We also investigated several types of tumor-infiltrating immune cells within subcutaneous tumors by flow cytometry. A higher proportion of neutrophils was detected in Mir4435-2hg^−/− mice than in WT mice (Fig. 4A). Within the neutrophil compartment, the percentage of CD14⁺ cells reached average 85.7% in WT mice and 93.0% in Mir4435-2hg^−/− mice (Fig. 4B), which demonstrated that tumor-infiltrating neutrophils were mostly PMN-MDSCs that were increased within the tumors of Mir4435-2hg^−/− mice. In contrast, CD3⁺ T cells and CD8⁺ T cells were decreased (Fig. 4C and D). Monocytes/M-MDSCs were also significantly decreased in Mir4435-2hg^−/− mice, whereas total CD45⁺ cells, macrophages, and M2 macrophages were unchanged (Fig. 4E–H). Similar results were observed in the AOM/DSS colorectal cancer model; tumor-infiltrating PMN-MDSCs were significantly increased, and CD8⁺ T cells were decreased (Fig. 4I) in Mir4435-2hg^−/− mice compared with WT mice by IHC. In the splenocytes/MC38 coculture system, we also observed that PMN-MDSCs were increased in the Mir4435-2hg^−/− group (Supplementary Fig. S7D). These data demonstrate that loss of Mir4435-2hg could cause an increase in PMN-MDSCs, which contribute to an immunosuppressive environment.

To investigate the biological mechanism by which Mir4435-2hg regulated PMN-MDSCs, we performed RNA-seq of neutrophils from BM, which showed 429 downregulated genes and 336 upregulated genes (Fig. 5A). GO enrichment analysis showed that the downregulated genes were enriched in mitosis-associated biological processes, including cell division, mitotic nuclear division, and regulation of mitotic centrosome separation (Supplementary Fig. S8A). Because MDSCs are considered relatively immature myeloid cells, these results might indicate a disordered process of neutrophil maturation in Mir4435-2hg^−/− mice, which provides another perspective from which to understand the decline in neutrophils and the elevation of PMN-MDSCs in Mir4435-2hg^−/− mice.

Most upregulated genes associated with the cholesterol biosynthetic process (Fig. 5B), and GSEA also showed that these genes were enriched in cholesterol homeostasis and fatty acid metabolism (Fig. 5C and D). Because accumulation of lipids is a critical regulator of the immunosuppressive function of PMN-MDSCs (44, 45), we presumed that higher lipid accumulation caused a stronger immunosuppressive effect of PMN-MDSCs in Mir4435-2hg^−/− mice. Through BODIPY lipid staining, we observed an increasing trend of lipid accumulation in neutrophils from tumor-free Mir4435-2hg^−/− mice, although the difference was not significant. Nevertheless, the PMN-MDSCs from the spleens of tumor-bearing Mir4435-2hg^−/− mice had more lipids than those of WT mice (Fig. 5E and F). Consistent with this, Mir4435-2hg^−/− mouse BM-derived PMN-MDSCs induced by CT26 and MC38 in vitro had more lipids than those from WT mice (Fig. 5G). Mir4435-2hg^−/− mouse-derived PMN-MDSCs expressed higher Arg1, Nos2, Cox2, and ROS (Fig. 5H–K), which resulted in a stronger inhibitory effect on the proliferation of T cells (Fig. 5L). Altogether, these data illustrate that loss of Mir4435-2hg enhances the immunosuppressive ability of PMN-MDSCs.

To further confirm that Mir4435-2hg depletion in neutrophils, but not in intestinal epithelial cells, caused the phenotypes observed in the mouse models, we generated conditional knockout Mir4435-2hg^flox/flox mice and crossed them with S100a8-Cre mice or Villin-Cre mice to construct neutrophil- or intestine-specific Mir4435-2hg deletion mice (Fig. 6A; Supplementary Fig. S8B and S8C). Flow cytometry analysis of leukocytes in peripheral blood showed a decrease in neutrophils in Mir4435-2hg^flox/flox S100a8-Cre mice compared to Mir4435-2hg^flox/flox and Mir4435-2hg^flox/flox Villin-Cre mice, and the number of eosinophils was also significantly decreased in Mir4435-2hg^flox/flox S100a8-Cre mice (Supplementary Fig. S8D). The subcutaneous tumor model showed that tumors grew significantly faster in Mir4435-2hg^flox/flox S100a8-Cre mice than in control mice (Fig. 6B), and tumor volume and weight (Fig. 6C and D) were significantly increased in neutrophil-deficient Mir4435-2hg mice. Consistent with the Mir4435-2hg^−/− mice, we found a higher lipid accumulation in the PMN-MDSCs from tumor-bearing Mir4435-2hg^flox/flox S100a8-Cre mice (Fig. 6E), as well as in BM-derived PMN-MDSCs induced by MC38 cultured medium in vitro (Fig. 6F).

In the AOM/DSS model, Mir4435-2hg^flox/flox S100a8-Cre mice developed more and larger tumors than Mir4435-2hg^flox/flox Villin-Cre and Mir4435-2hg^flox/flox mice (Fig. 6G; Supplementary Fig. S8E). Although the tumor counts were not significantly higher in Mir4435-2hg^flox/flox S100a8-Cre mice (Fig. 6H), the average tumor size and tumor burden (Fig. 6I and J) were significantly increased in Mir4435-2hg^flox/flox S100a8-Cre mice. No obvious changes in tumor count, tumor size, or tumor burden were observed between Mir4435-2hg^flox/flox Villin-Cre mice and Mir4435-2hg^flox/flox mice. We also observed more tumor-infiltrating PMN-MDSCs and fewer CD8⁺ T cells in Mir4435-2hg^flox/flox S100a8-Cre mice than in Mir4435-2hg^flox/flox Villin-Cre mice and Mir4435-2hg^flox/flox mice (Fig. 6K). Overall, deletion of Mir4435-2hg in neutrophils/PMN-MDSCs, but not in the intestinal epithelium, promoted colorectal cancer progression.

Here, genetically engineered mouse models demonstrated that MIR4435-2HG suppressed colorectal cancer initiation and progression by regulating the immunosuppressive activity of PMN-MDSCs. These findings are a significant departure from the literature, especially the colorectal cancer literature. Ouyang and colleagues (17) report MIR4435-2HG as a poor prognostic marker in their analysis of TCGA and two selected GEO databases. However, their analysis did not consider the high homology of MIR4435-2HG with CYTOR, which could have resulted in a false expression value for MIR4435-2HG that included CYTOR expression (TCGA anlaysis). In the GEO database, probe 232918_at is designed as a specific probe targeting MIR4435-2HG, whereas probe 225799_at targets both MIR4435-2HG and CYTOR. The previous study (17) used probe 225799_at to evaluate the expression value of MIR4435-2HG. In this study, we used the signal value of 232918_at to determine expression of MIR4435-2HG. We also downloaded all eligible GEO databases as of October 2017 to perform a meta-analysis, and the results revealed no significant relationship between MIR4435-2HG expression and overall survival. Shen and colleagues (18) investigated MIR4435-2HG expression in clinical colorectal cancer samples, but the primers utilized to detect MIR4435-2HG do not match MIR4435-2HG sequence. Another study reports that MIR4435-2HG promotes colorectal cancer proliferation and metastasis through the miR-206/YAP1 axis (19), but both the primers and the shRNA targeting MIR4435-2HG used also target CYTOR. Therefore, we believe the conclusions from these studies may not reflect the true expression of MIR4435-2HG and that the phenotypes observed may have been confounded by alterations in CYTOR expression. Other studies of MIR4435-2HG in other types of cancer may also need to be reexamined to verify whether off-target effects occurred. In the meanwhile, MIR4435-2HG might have different distribution and diverse functions in different tumor types.

In our study, we used specific CRISPR-Cas9/sgRNAs, siRNAs, and an overexpression vector to alter the expression of MIR4435-2HG in colorectal cancer cells, which did not affect CYTOR expression (Supplementary Fig. S9). Alteration of MIR4435-2HG did not change the proliferation, migration, or invasion of human colorectal cancer cells. The same phenotypes were also observed in colorectal cancer cell lines from mice. In mice, CYTOR is deficient (no homologous sequence found). Because of the high level of MIR4435-2HG in the tumor stroma and the loss of CYTOR in mice, we generated Mir4435-2hg–deficient mice. Mir4435-2hg-specific depletion promoted colorectal tumorigenesis and progression in a colitis-associated colorectal cancer model, spontaneous intestinal adenomatous model, and subcutaneous tumor experiments. Although no significant change in the tumor number or tumor size was observed in the long-term colitis-associated colorectal cancer model using Mir4435-2hg^−/− mice, H&E staining showed that tumors in Mir4435-2hg^−/− mice exhibited higher invasion potency. The reason might be that tumor formation and proliferation reached a maximum for long-term induction, but Mir4435-2hg depletion still promoted colorectal cancer progression in later stages. In Mir4435-2hg^−/− mice, BIM upregulation caused neutrophil apoptosis, but PMN-MDSCs with high BCL2L1 expression resisted BIM-induced apoptosis. The increased lipid accumulation in Mir4435-2hg^−/− PMN-MDSCs enhanced their immunosuppressive activities. Therefore, we observed higher infiltration of PMN-MDSCs and fewer T cells in tumors from Mir4435-2hg^−/− mice. In splenocyte/MC38 co-culture assays, we also observed increased PMN-MDSCs in the Mir4435-2hg^−/− group, but possibly due to the low cell population and insufficient incubation time, no difference in T-cell number was observed. Nevertheless, the increased number and enhanced immunosuppressive activity of PMN-MDSCs in the Mir4435-2hg^−/− group might also suppress T-cell activity, which needs to be further confirmed.

Hypercholesterolemia and dyslipidemia have previously been associated with increased baseline inflammatory responses and higher colorectal cancer incidence (46, 47). Mir4435-2hg depletion only led to high lipid accumulation in MDSCs, with no obvious alteration of plasma cholesterol and triglycerides (Supplementary Fig. S4B) in our mouse models. Conditional knockout of Mir4435-2hg in mice further confirmed that Mir4435-2hg depletion in neutrophils and PMN-MDSCs, but not in intestinal epithelial cells, promoted colorectal cancer development. All together, these data showed that Mir4435-2hg depletion regulated not only the ratio of neutrophils to PMN-MDSCs, but also the immunosuppressive activity of PMN-MDSCs, which contributed to colorectal cancer initiation and progression. Thus, we call for a rigorous reassessment of the biological characterization and clinical significance of lncRNA MIR4435-2HG.

Myelopoiesis is a structured process in which the common myeloid precursors in the bone marrow differentiate into mature circulating leukocytes. However, a variety of pathological conditions, such as chronic inflammation, autoimmune diseases, and cancer, can induce aberrant myelopoiesis, causing the accumulation of immature myeloid cells that diverge from the standard pathway of differentiation (48). Neutrophil differentiation involves several stages, from hematopoietic progenitor cells, common myeloid progenitors, and granulocyte-macrophage progenitors to myeloblasts, myelocytes, metamyelocytes and band forms, and then mature neutrophils (49). Under pathologic conditions, some immature myeloid cells might expand and convert to PMN-MDSCs (50). Consistent with a previous report (35), we observed a decrease in neutrophils in Mir4435-2hg^−/− mice, and many differentially expressed genes regulated by Mir4435-2hg associated with mitosis. It needs to be further investigated whether the decrease in neutrophils in Mir4435-2hg^−/− mice could increase conversion to PMN-MDSCs. Classical neutrophils, monocytes, and pathologically activated MDSCs coexist in humans and mice within tumors, and MDSCs accumulate with tumor progression (50). Mouse PMN-MDSCs are defined by the surface markers CD11b⁺Ly6G⁺Ly6C^lo, similar to classical neutrophils, which makes it difficult to distinguish PMN-MDSCs from neutrophils. Therefore, all neutrophils from tumor-bearing mice have been considered PMN-MDSCs for comparison with neutrophils from tumor-free mice in most studies. CD80, CD115, CD124, CD224, and PD-L1 are believed to be PMN-MDSC markers due to their absence on neutrophils in some experimental models (21, 51). However, these markers are not widely accepted. Veglia and colleagues propose CD14 as a specific marker of PMN-MDSCs to distinguish them from classical neutrophils using scRNA-seq in a pure population of neutrophils (43). In our study, we also observed a higher proportion of CD14⁺ cells among splenic and tumor-infiltrating neutrophils in Mir4435-2hg^−/− mice, which represents a higher infiltration of PMN-MDSCs. However, whether CD14 could become the gold standard to discriminate between PMN-MDSCs and classical neutrophils needs further verification. Tumor-associated neutrophils (TAN) are another, related definition of PMN-MDSCs. These comprise a heterogeneous population that includes N1 TANs with antitumor properties and N2 TANs with suppressive functions (52). N2 TANs have been considered PMN-MDSCs (25, 53), and N1 TANs, as normally activated neutrophils, exert antitumor effects by activating unconventional T cells and exerting direct cytotoxic activity against tumors and antimicrobial activity (54). In this study, the decrease in neutrophils in Mir4435-2hg^−/− and Mir4435-2hg^flox/flox S100a8-Cre mice might also contributed to cancer progression, especially at early-stage disease.

Henao-Mejia and colleagues first reported that murine Mir4435-2hg (termed myeloid RNA regulator of Bim-induced death) could regulate Bim and the short myeloid cell lifespan (35). They also determined that Mir4435-2hg plays a vital role in viral infection (55), leukemia (56–58), and stress-induced abnormalities and clonal hematopoiesis (59). However, most studies on human MIR4435-2HG have focused on cancer biology. Reports analyzing TCGA data for colorectal cancer (60), hepatocellular carcinoma (61), and pan-cancer (62) have revealed that MIR4435-2HG associates with glycolysis and immune infiltration. Together, these data and our data suggest that these correlations should be attributed to the high expression of MIR4435-2HG in immune cells. High immune infiltration leads to a high expression of MIR4435-2HG in tumor tissues. MIR4435-2HG could activate oxidative phosphorylation in myeloid dendritic cells (DCs) from HIV-1 elite controllers by regulating epigenetic modifications and facilitating the transcription of Regulatory-Associated Protein of MTOR Complex 1 (RPTOR) (63). This emerging evidence implies that MIR4435-2HG could also control other immune cells besides neutrophils and PMN-MDSCs, such as DCs and eosinophils, to regulate tumor progression. According to our data, Mir4435-2hg depletion almost resulted in the absence of eosinophils. Higher eosinophil infiltration is reported to associate with better prognosis in colorectal cancer (64). Therefore, we did not exclude the tumor-suppressive roles of eosinophils, which are worth further exploring.

In conclusion, our study demonstrates the unexpected role of MIR4435-2HG using multiple mouse models. It functioned as a tumor suppressor in the tumor stroma by remodeling the immune microenvironment rather than as an oncogene in tumor cells. The deficiency of MIR4435-2HG increased tumor-infiltrating PMN-MDSCs and enhanced their immunosuppressive potential to promote colorectal cancer development, further illustrating mechanisms behind the pathogenesis of colorectal cancer and providing a potential antitumor immunotherapy target.

No disclosures were reported.

H. Yu: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Chen: Data curation, software, formal analysis, visualization, methodology, writing–review and editing. F. Han: Data curation, software, formal analysis, methodology. J. Tang: Formal analysis, visualization, methodology. M. Deng: Investigation. Y. Niu: Investigation, writing–original draft. M. Lai: Conceptualization, resources, data curation, supervision, funding acquisition, project administration. H. Zhang: Conceptualization, resources, data curation, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work is supported by the National Natural Science Foundation of China (81871937, 81672730, 91859204, 82072629, 82173223), the Natural Science Foundation of Zhejiang Province (LZ21H160001) and CAMS Innovation Fund for Medical Sciences (CIFMS, 2019-I2M-5-044). We are grateful to Rongpan Bai, Zhejiang University School of Medicine, for providing APC^min/+ mice. We thank Ms. Xueping Zhou in the Laboratory Animal Center of Zhejiang University for technical assistance on breeding and management of mice and Ms. Qiong Huang for the technical support by the Core Facilities, Zhejiang University School of Medicine.

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.