The monoclonal PTX3 antibody (WHC-001) and isotype IgG1κ used for in vitro and in vivo experiments were previously developed by our laboratory [29]. Briefly, the design and production of the anti-PTX3 neutralizing antibody were performed in cooperation with Leadgene Biomedical, Inc. (Tainan, Taiwan). Recombinant human wild-type and N220A PTX3 proteins were purified from Expi293 cells according to the manufacturer's instructions. Briefly, wild-type or N220A PTX3 expression plasmids were transfected into Expi293 cells by using the ExpiFectamine™ 293 Transfection Kit (Life Technologies Corporation, 5781 Van Allen Way, PO Box 6482, Carlsbad, CA 92008), and the supernatants were harvested for His-tag purification. Other recombinant proteins used in this study included murine PTX3 (#2166-TS-025/CF, R&D Systems, Inc., a Bio-Techne Brand, 614 McKinley Place NE, Minneapolis, MN 55413, USA), murine IL-1β (#211-11B, PeproTech, Part of Thermo Fisher Scientific, 5 Cedarbrook Drive, Cranbury, NJ 08512), murine TNFα (#ab259411, Abcam Inc., 1 Kendall Square, Ste 341, Cambridge, MA 02139-1517, USA), murine M-CSF (#576404, BioLegend Inc., 8999 BioLegend Way, San Diego, CA 92121, USA), murine IL-4 (#214-14, PeproTech, Part of Thermo Fisher Scientific, 5 Cedarbrook Drive, Cranbury, NJ 08512), murine IL-13 (#210-13, PeproTech), human IL-1β (#200-01B, PeproTech), and human TNFα (#ab9642, Abcam Inc., 1 Kendall Square, Ste 341, Cambridge, MA 02139-1517, USA).

Total cell lysates were homogenized in TRIzol reagent (Invitrogen Life Technologies, Carlsbad, California, USA), and the samples were sent to Biotools Co., Ltd. (Taiwan) for RNA sequencing analysis. The DNA library was constructed by Illumina paired-end sequencing. Briefly, quality control was performed by using FastQC, MultiQC, and ReSeqTools, and the raw reads were trimmed with Trimmomatic (v.0.38) to remove low-quality reads and adaptors. The DNA library was mapped to a reference genome (GRCh38, Homo sapiens) by using HISAT2. The read numbers were analyzed by featureCounts (v1.6.3) and normalized to TPM (transcript per million) values.

Female C57BL/6, BALB/c and NOD-SCID mice were purchased from the Laboratory Animal Center of National Cheng Kung University (Tainan, Taiwan) or BioLASCO (Taiwan Co., Ltd). Female Ptx3 conditional knockout mice (C57BL/6J—Ptx3fl/fl conditional knockout × B6.Cg-Ndor1Tg(UBC−Cre/ERT2)1Ejb/2 J mice) were bred and maintained in the National Applied Research Laboratory (Taiwan). All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at NCKU (IACUC No. 109037). C57BL/6 and BALB/c mice were 8–9 weeks old, transgenic mice (Ptx3fl/fl and Ptx3fl/fl; Ubc-Cre) were 8–10 weeks old, and NOD-SCID mice were 6–7 weeks old when the experiments were performed.

For the mixed stroma-tumor (MEFs-MC38 cells) model, 8–9-week-old female C57BL/6 mice were anesthetized and 5 × 104 MC38 cells mixed with 1 × 105 MEFs were injected subcutaneously into the dorsal region (day 0); when the tumor volume reached 80–200 mm3, IgG1κ or WHC-001 (αPTX3 Ab) was intratumorally injected (100 μg in 100 μl) every 3 days for a total of 4 treatments by using 31G insulin syringes (BD Ultra-Fine 0.3 ml; Becton, Dickenson and Co.), and the mice were sacrificed on the 21 day post cell inoculation.

For the MC38 and CT26 cell subcutaneous tumor model, 8–9-week-old female C57BL/6 mice or BALB/c mice were anesthetized and 2 × 105 MC38 or CT26 cells were injected subcutaneously in the dorsal region (day 0) and on the 8th day post cell inoculation, IgG1κ or WHC-001 (αPTX3 Ab) was intraperitoneally injected (10 mg/kg) every 4 days for a total of 3 treatments, and the mice were sacrificed on the 22nd day (MC38) or 20th day (CT26); 6–7-week-old female NOD-SCID mice were anesthetized and 2 × 105 MC38 cells were injected subcutaneously in the dorsal region (day 0) and beginning on the 7th day post cell inoculation, IgG1κ or WHC-001 (αPTX3 Ab) was intratumorally injected (10 mg/kg) every 4 days for a total of 3 treatments, and the mice were sacrificed on the 21st day.

For the MC38 orthotopic tumor model, 8-week-old female C57BL/6 mice were anesthetized and 1 × 106 MC38 cells in 30 μl of PBS were injected into the cecal wall (day 0) by using a 30G insulin syringe (BD Ultra-Fine 0.5 ml; Becton, Dickenson and Co.); beginning on the 7th day post cell inoculation, IgG1κ or WHC-001 (αPTX3 Ab) was intraperitoneally injected (10 mg/kg) every 4 days for a total of 3 treatments, and the mice were sacrificed on the 22nd day.

For the MC38 subcutaneous tumor model in Ptx3 conditional knockout mice (C57BL/6 J—Ptx3fl/fl conditional knockout × B6.Cg-Ndor1Tg(UBC−Cre/ERT2)1Ejb/2 J mice), the genotypes of the wild-type and knockout mice were Ptx3fl/fl and Ptx3fl/fl;Ubc-Cre, respectively. For the latter knockout model, 8–10-week-old female or male mice were anesthetized and 2 × 105 MC38 cells were injected subcutaneously into the dorsal region (day 0); beginning on the 5th day post cell inoculation, tamoxifen (2 mg/mouse in 10% EtOH corn oil) was intraperitoneally injected daily for a total of 5 days, and the mice were sacrificed on the 19th day. Tumor volume (mm3) = (L × W2)/2.

Tumor tissues were cut into small pieces and incubated in 5 ml digestion medium (RPMI-1640 medium containing 5% FBS, 1% P/S, 150 μl 10,000 CDU/ml collagenase I #C0103) at 37 °C on a shaker at 100 rpm for 40 min. A GentleMACS dissociator was used to homogenize the tissues into a single-cell suspension according to the manufacturer’s instructions. The suspension was filtered through a cell strainer (70 μm mesh). Tumor-infiltrating immune cells were stained with fluorescence-conjugated antibodies, including CD45-BV510 (#563891, BD), CD8a-PE (#553033, BD), CD4-BV421(#562891, BD), NK1.1-APC (#550627, BD), CD3e-PECy7 (#552774, BD), CD11b-BV421 (#562605, BD), Ly6G-FITC (#551460, BD), Ly6C-PECy7 (#560593, BD), F4/80-PE (#565410, BD), CD206-Alexa647 (#565250, BD), CD86-BB700 (#742120, BD), CD4-FITC (#553046, BD), and IFNγ-BV421 (#563376, BD). The populations were analyzed by using a CytoFLEX™ Flow Cytometer (Beckman Coulter). The CD45-positive immune cells were gated for CD8+ T cells (CD8+), CD4+ T cells (CD4+), NK cells (NK1.1+CD3e−), PMN-MDSCs (Ly6G+Ly6CMid/CD11b+), M-MDSCs (Ly6C+Ly6Clow/CD11b+), M1-like macrophages (CD11b+F4/80+CD206− or CD11b+F4/80+CD86+CD206−), and M2-like macrophages (CD11b+F4/80+CD206+ or CD11b+F4/80+CD206+CD86−). For analysis of cytotoxic CD8+ T (IFNγ+CD8+/CD45+) and CD4+ T (IFNγ+CD4+/CD45+) cells, cells in suspension were further activated by incubation with stimulation medium (1.33 μM ionomycin, 20 ng/ml PMA, 10 μg/ml brefeldin A) at 37 °C for 4 h. Cells were harvested, and surface markers were stained with fluorophore-conjugated antibodies (CD45-BV510, CD8-PE, CD4-FITC). Stained cells were then fixed with 3.7% paraformaldehyde and stored at 4 °C overnight. The cells were permeabilized with Perm/Wash Buffer (#554723, BD BioSciences) and stained with the IFNγ-BV421 antibody for flow cytometric analysis.

For experiments with Ptx3 conditional knockout mice, 8–10-week-old female mice were intraperitoneally injected with tamoxifen (2 mg/mouse in 10% EtOH corn oil) daily for a total of 5 days (days − 5 to − 1), and the mice were sacrificed on the 8th day. The femurs and tibias were sterilized with 70% EtOH and washed with PBS. Bone marrow cells were flushed out with serum-free RPMI-1640 medium and collected. After centrifugation, red blood cells were removed by using Hybri-Max™ Red Blood Cell Lysing Buffer (R7757-100ML, Sigma) according to the manufacturer’s instructions. The cells were then filtered through a 70 μm mesh, washed with PBS and suspended in complete RPMI-1640 medium. For macrophage differentiation, a total of 4 × 106 cells were seeded in 6-well plates with 2 ml complete RPMI-1640 medium containing 25 ng/ml M-CSF and incubated for 5 days (the medium and M-CSF were refreshed after 3 days). To induce M2 macrophage polarization, cells were treated with 20 ng/ml IL-4 and IL-13 for another 2 days, and RNA was harvested for analysis. For experiments with wild-type C57BL/6 mice, 8–10-week-old female mice were sacrificed, and the bone marrow cells were harvested as mentioned above. A total of 4 × 106 cells were seeded in 6-well plates with 2 ml complete RPMI-1640 medium containing 25 ng/ml M-CSF and incubated for 7 days for macrophage differentiation (the medium and M-CSF were refreshed after 3 days). Macrophages were then cultured in medium without M-CSF for 24 h and treated with 100 ng/ml recombinant murine PTX3 protein for another 24 h.

For the experiment of BMDMs co-cultured with MEFs, 3 × 105 MEFs (shVoid, shPtx3#915, and shPtx3#916) were seeded in 6-well plates for 48 h. BMDMs (4 × 106 bone marrow cells incubated in 25 ng/ml M-CSF complete medium for 7 days of macrophage differentiation) were trypsinized and added to MEFs. After 48 h of co-culture, cells were harvested by using 5 mM EDTA and stained with fluorescence-conjugated antibodies including CD45, CD11b, F4/80, CD206, and CD86 for flow cytometric analysis.

For single-cell RNA-seq analysis, all data were analyzed and downloaded as png files on the Broad Institute Single Cell Portal (portals.broadinstitute.org/single_cell). The dataset named “Human Colon Cancer Atlas (c295)” was selected. For the scatter data and distribution data analysis, the settings were as follows: Gene name = PTX3, Clustering = All cells (tSNE), Annotation = ClusterTop, Subsampling = All Cells, Plot type = Violin plot, Data points = All.

The raw mRNA-seq data from the TCGA colon adenocarcinoma (COAD) dataset were obtained from the FireBrowse database (“http://firebrowse.org/”, [32]). In the mRNA-seq bar (Aliquot counts = 457), the illuminahiseq_rnaseqv2-RSEM_genes_normalized (MD5) and illuminaga_rnaseqv2-RSEM_isoforms_normalized (MD5) files were downloaded.

For gene set enrichment analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp, [33]), the raw mRNA-seq data from the COAD data file [illuminahiseq_rnaseqv2-RSEM_genes_normalized (MD5)] were transformed into log2(raw counts + 1) values and split into the PTX3 high/low groups by the median cutoff value. The resulting file was uploaded into GSEA v.4.1.0 software for analysis. For KEGG pathway analysis, the c2.cp.kegg.v2023.1.Hs.symbols.gmt [Curated] gene set was selected. The signaling pathways enriched in the PTX3high group are shown and were clustered into stromal activation-related signaling and inflammation-related signaling. The threshold criteria were NOM p value < 0.05 and FDR q value < 0.25. For inflammation response signature analysis, the HALLMARK_INFLAMMATORY_RESPONSE gene set was used.

For stromal score analysis, the stromal scores from the TCGA colorectal adenocarcinoma (RNA-seqV2) dataset were obtained from the ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data) database (“https://bioinformatics.mdanderson.org/estimate/”, [34]) with the following parameters: Disease type = Colorectal Cancer, and Platform type = RNA-Seq-V2. The raw data were downloaded and merged with PTX3 mRNA expression (illuminaga_rnaseqv2-RSEM_isoforms_normalized (MD5)) data obtained from the FireBrowse database mentioned above. The correlation plot was generated by using GraphPad Prism 8.0 software.

For analysis of fibroblast scores, the preanalyzed TCGA colon adenocarcinoma data were downloaded via the xCell web tool (“https://xcell.ucsf.edu/”, [35]). The score data were merged with the PTX3 mRNA expression [illuminahiseq_rnaseqv2-RSEM_genes_normalized (MD5)] data obtained from the FireBrowse database mentioned above and split into the PTX3 high/low or fibroblast high/low groups by the median cutoff value.

For analysis of survival prognosis, survival data from the TCGA colon adenocarcinoma dataset was obtained via the SurvExpress web tool (“http://bioinformatica.mty.itesm.mx/SurvExpress”, [36], downloaded in 2018). The survival prognosis data were merged with the PTX3 mRNA expression and fibroblast score data mentioned above, and analysis results were plotted by using GraphPad Prism 8.0 software. The samples were split into high/low fibroblast score groups by the best cutoff value (114/89), and these groups were further subdivided into PTX3 high/low groups by the median cutoff value, and analysis results were plotted by using GraphPad Prism 8.0 software.

For analysis of the immune scores in the TCGA colon adenocarcinoma dataset, the raw mRNAseq data of COAD (illuminahiseq_rnaseqv2-RSEM_genes_normalized (MD5)) were split into the PTX3 high/low groups by the median cutoff value and uploaded to the CIBERSORT web tool (“https://cibersort.stanford.edu/”, [37]). The LM22 signature (22 immune cell types) was selected. The results of immune score analysis were plotted by using GraphPad Prism 8.0 software.

For analysis of RNA-seq data to predict PTX3 downstream signaling, the Connectivity Map web tool ("https://clue.io/about", [38]) was used. The top 200 upregulated and downregulated genes were uploaded in the “Query app” in the Tools section. The query parameters were as follows: Gene expression (L1000), Touchstone, Individual query, and 1.0. For the forward analysis, the upregulated genes were input in the UP-regulated genes box, and the downregulated genes were input in the DOWN-regulated genes box; after analysis, the Gene Over-Expression was selected as the perturbagen type, and candidates with a score of > 0 were obtained. For the reverse analysis, the upregulated genes were input in the DOWN-regulated genes box, and the downregulated genes were input in UP-regulated genes box; after analysis, the Gene Knock-Down was selected as the perturbagen type, and the candidates with a score of > 0 were obtained. We selected the overlapping candidates from the forward and reverse analysis results.

Statistical analyses of all data in this study were performed by using GraphPad Prism 8.0, and data are presented as the means ± SEMs. Significant differences were identified by using two-tailed unpaired Student's t test, two-tailed paired Student's t test, one-way ANOVA or two-way ANOVA. GSEA data were analyzed by using GSEA_4.1.0. with the parameters set as p value of < 0.05 and FDR < 0.25. Statistical significance is represented as * p < 0.05, ** p < 0.01, *** p < 0.001. ns represents nonsignificant difference.