The CAFs-derived YAP1-periostin axis forms a fibrotic tumor microenvironment to promote metastatic lung tumor growth

Abstract

Background

The primary cause of poor prognosis in non-small cell lung cancer (NSCLC) patients is metastasis; however, little is known about the underlying mechanisms mediated by cancer-associated fibroblasts (CAFs) in the tumor microenvironment (TME).

Methods

We conducted comprehensive molecular and cellular analyses, including RNA sequencing, immunohistochemistry, coculture experiments, and genetic mouse models, to investigate the role of the CD248-periostin-integrin β1 (ITGB1) axis in NSCLC metastasis.

Results

We demonstrated that CD248⁺ CAFs are key regulators of NSCLC metastasis. RNA sequencing and clinical sample analysis revealed that CD248 expression in CAFs was positively correlated with increased secretion of periostin (encoded by POSTN). In CAFs, elevated periostin expression is strongly linked to lymph node metastasis (p = 0.002), advanced tumor stage (p = 0.0039), and poor overall survival (p < 0.05) in patients with NSCLC. By promoting the nuclear translocation of YAP1 and regulating POSTN promoter activity to increase periostin expression, CD248 mechanistically causes YAP1 activation in CAFs. According to functional tests, CAFs secreted periostin binds to ITGB1 in NSCLC cells, triggering the FAK/Src signaling pathway to cause epithelial-mesenchymal transition (EMT)-like phenotypic changes and promote NSCLC cell invasion and migration. Additionally, periostin promotes collagen I deposition and increases extracellular matrix (ECM) stiffness, which further amplifies YAP1-driven periostin secretion from CD248⁺ CAFs, resulting in the formation of a prometastatic positive feedback loop. Fibroblast-specific CD248 deletion or POSTN knockout mice exhibit dramatically reduced metastatic tumor growth in NSCLC in vivo. This is accompanied by decreased ECM stiffness in the TME, decreased EMT-like phenotypic changes, and decreased collagen I deposition.

Conclusions

Our results collectively suggested a CD248–YAP1-Periostin-ITGB1 axis that contributes to a prometastatic feedback loop in NSCLC, indicating that focusing on this axis may be a viable therapeutic approach to prevent NSCLC metastasis.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08664-3.

Introduction

The global burden of disease and mortality is profoundly affected by cancer, which continues to increase [1–3]. Lung cancer, in particular, accounts for the majority of cancer-related deaths, with non-small cell lung cancer (NSCLC) representing approximately 85% of lung cancer cases [4]. Despite significant advancements in therapeutic strategies, tumor metastasis remains a formidable bottleneck severely limiting the prognosis of NSCLC patients. A critical component of this metastatic process is the tumor microenvironment (TME), in which cancer-associated fibroblasts (CAFs) are recognized as central orchestrators. However, a comprehensive understanding of the precise molecular and cellular mechanisms through which CAFs facilitate metastasis is still lacking.

Chemotherapy has remained the cornerstone of systemic treatment for advanced NSCLC, with platinum-based doublet regimens constituting the standard first-line therapeutic approach for several decades [5, 6]. Conventional chemotherapeutic agents, including cisplatin, carboplatin, paclitaxel, and gemcitabine, exert antitumor effects predominantly through the induction of DNA damage, inhibition of microtubule dynamics, and disruption of nucleotide biosynthesis [7]. Although substantial progress has been made in advancing cytotoxic drug delivery systems—encompassing liposomal formulations and nanocarrier-based platforms specifically designed to enhance tumor targeting and attenuate systemic toxicity—the clinical efficacy of chemotherapy in NSCLC remains substantially compromised by the emergence of multidrug resistance, dose-limiting adverse effects, and the limited capacity to effectively prevent or eliminate metastatic dissemination [8, 9]. Consequently, the identification of appropriate therapeutic targets represents a critical imperative in NSCLC treatment strategies.

The TME has become a crucial regulator of the development of cancer, and stromal cells are essential for metastasis [10]. The most prevalent cell type among these stromal components is CAFs, which play a critical role in accelerating the spread of cancer through a variety of pathways, including immunological regulation, extracellular matrix (ECM) remodeling, and paracrine signaling [11–13]. However, the remarkable heterogeneity of CAFs and the lack of definitive surface biomarkers have limited our understanding of their specific contributions to NSCLC pathogenesis.

One important indicator of tumor-associated stromal cells is the type I transmembrane glycoprotein known as tumor endothelial marker 1 (TEM-1/endosialin/CD248) [14]. Research has indicated that CD248 is expressed mostly on stromal fibroblasts and tumor-associated pericytes and is expressed at negligible levels in healthy adult tissues [15–17]. According to our earlier research, CAFs originating from NSCLC have high levels of CD248 expression, and these levels are closely correlated with tumor grade, advanced clinical stage, and poor patient prognosis [15, 16, 18, 19]. The primary cause of cancer-related mortality is distant metastasis, and the development of the premetastatic niche (PMN) is an essential precondition for metastasis [20]. According to recent research, CD248⁺CAFs help NSCLC cells undergo epithelial-mesenchymal transition (EMT)-like phenotypic changes [21]. However, the precise molecular mechanisms through which CD248⁺ CAFs induce EMT-like phenotypic changes and facilitate NSCLC metastasis remain to be fully elucidated.

Periostin, also known as osteoblast-specific factor 2 (OSF-2), is an extracellular matrix protein encoded by the POSTN gene [22]. Periostin is significantly increased in a variety of cancers and has been linked to tumor growth and metastasis, despite being expressed at low levels in healthy adult tissues [22–24]. Notably, studies have demonstrated significantly elevated periostin expression in hepatic metastases compared with that in primary colorectal cancer lesions, underscoring its critical role in metastatic dissemination [25]. Emerging evidence indicates that CAFs represent the primary source of periostin secretion within the TME [22, 26, 27]. By reorganizing the extracellular matrix and immunomodulating the TME, CAFs-derived periostin mechanistically promotes tumor growth [26, 28]. The interaction between periostin and its cognate receptor integrin β1 (ITGB1), both of which are expressed on NSCLC cells, establishes a bidirectional signaling axis implicated in metastatic spread [27, 29]. Periostin, a pivotal regulatory molecule in the ECM, has pleiotropic effects on tumor metastasis [30, 31]. The specific function of periostin in NSCLC metastasis is still unclear, despite its acknowledged significance in tumor biology.

In this study, we investigated the hypothesis that CD248⁺ CAFs facilitate NSCLC metastasis through periostin secretion and subsequently bind to ITGB1 on NSCLC cells, thereby triggering the EMT-like phenotypic changes. Our findings demonstrate that CD248 activates YAP1 to induce periostin secretion by CAFs, which subsequently enhances NSCLC metastasis. Furthermore, we discovered that periostin promotes collagen I deposition, increases ECM stiffness and creates a positive feedback loop that further enhances periostin secretion. Mechanistically, the binding of periostin to ITGB1 in NSCLC cells triggers EMT-like phenotypic changes via FAK/Src pathway activation, ultimately promoting NSCLC metastasis. These findings offer fresh perspectives on possible treatment approaches for enhancing NSCLC patient survival and suggested a unique molecular axis connecting CD248⁺ CAFs to NSCLC metastasis via the Periostin-ITGB1 pathway.

Materials and methods

Study approval

Human NSCLC samples and corresponding normal tissues were obtained after the Clinical Research Ethics Committee of Guizhou Medical University provided clearance (protocol no. 2024LL-48), and the study was carried out in compliance with all applicable ethical guidelines. Every patient provided written informed consent. Every animal procedure was carried out in compliance with ARRIVE guidelines and in accordance with protocols authorized by Guizhou Medical University’s Institutional Animal Care and Use Committee (IACUC) (No. 2400105).

Cell lines and culture conditions

The American Type Culture Collection (ATCC, Manassas, VA, USA) provided human NSCLC cell lines (A549, NCI-H460, H23, H1299), normal lung epithelial BEAS-2B cells, and mouse Lewis lung cancer (LLC) cells. Using short tandem repeat (STR) analysis, all the cell lines were verified to be free of Mycoplasma contamination. The cells were maintained at 37 °C in a humidified 5% CO2 incubator with either RPMI 1640 media or DMEM/F12 (Gibco Life Technologies, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin.

As previously reported, fibroblasts were extracted and cultured [18, 32]. Fresh tumor tissue was minced and enzymatically digested with collagenase-3 at 37 °C for 1 hour under agitation. After centrifugation, the cells were filtered through 100 μm and 40 μm strainers. Cells were then cultured, and nonadherent cells were removed via differential adhesion for 2 hours. Following purification through serial passaging, the isolated cells were identified as fibroblasts on the basis of positive immunolabeling for vimentin and α-SMA. CAFs with stable YAP1 knockdown (CAFs-sh-YAP1) [33], stable CD248 knockdown (CAFs-sh-CD248), stable CD248 overexpression (CAFs-CD248OE), or stable Periosin knockdown (CAFs-sh-Periostin or CAFs-CD248OE-sh-Periostin) were produced via lentiviral transduction.

Through lentiviral integration, we produced A549-sh-ITGB1 and NCI-H460-sh-ITGB1 cells with persistent ITGB1 deficiency.

Coculture experiments

We used conditioned medium (CM) from normal fibroblasts (NFs), CAFs-sh-CON, CAFs-sh-CD248, CAFs-sh-CD248+rhPeriostin (R&D Systems, #3548-F2), CAFs-CD248OE, CAFs-CD248OE-sh-Periostin, PF-562271(FAK inhibitor) (MCE, #HY-10459) or Dasatinib (Src inhibitor) (MCE, #HY-10181) to plate A549 and NCI-H460 cells onto six-well plates for coculture assays, which was done before a 48-h exposure. After the cells reached 90% confluency, they were examined.

Mouse models

Suzhou Cyagen Co., Ltd., created mice with floxed cd248 or fsp-1-Cre to generate CD248 gene knockout mice specific to fibroblasts. The mice were crossed with floxed cd248 animals to generate cd248^fl/flfsp-1^+/+ (WT) and cd248^fl/flfsp-1^cre/+ (cKO) mice. POSTN^+/+ (WT) and POSTN^−/− (homozygous knockout, KO) mice were created by crossing POSTN^+/− heterozygous mice to create POSTN gene knockout mice. All the mice were maintained in pathogen-free facilities under standard conditions (12 hours light/dark cycle, 22 ± 1 °C, 55 ± 5% relative humidity) with ad libitum access to a standard laboratory diet. All the experimental animals were on the C57BL/6J background and were aged 6–12 weeks. Guizhou Medical University granted ethical approval for all the animal procedures.

Migration and invasion assays

To assess cell invasion and migration, Transwell chambers were used. As explained in Sect. 2.3, A549 and NCI-H460 cells were pretreated for 48 hours with different CAFs-conditioned media, and the tumor cells were placed in the upper chamber for 24 hours. The results of three separate experiments were averaged.

Immunofluorescence (IF) staining

The anti-IF antibodies used were anti-CD248 (CST, #47948), anti-α-SMA (Abcam, #ab5694), anti-collagen I (Servicebio, #GB11022), anti-paxilin (Absin, #abs154857), Alexa Fluor 594-phalloidin (Absin, #abs47048273), anti-periostin (Abcam, #ab14041), anti-ITGB1 (Servicebio, #GB115173), and anti-YAP1 (Servicebio, #GB113975). For multiplex IF, tyramide signal amplification (TSA) was employed. After the tissue sections were blocked with 5% BSA and 3% H₂O₂, they were sequentially incubated with primary antibodies overnight at 4 °C, HRP-conjugated secondary antibodies for 50 minutes at room temperature, and then TSA fluorophores for 10 minutes at room temperature. The antigen-antibody complexes were removed by boiling in EDTA buffer between cycles. For cellular IF, fixed and permeabilized cells were stained with appropriate antibodies, counterstained with DAPI, and imaged via confocal microscopy.

Extracellular matrix (ECM) assembly assay

After the six-well plates were coated with a 0.2% gelatin solution at 37 °C, the cells were washed with PBS. Crosslinking was initiated with a 1% glutaraldehyde solution. After additional PBS washes, 1 mol/L glycine solution was added to quench the crosslinking (20 minutes, RT). Following the final PBS wash, the wells were filled with PBS containing 1× penicillin-streptomycin. Various fibroblast populations were seeded and cultured in ECM assembly medium (DMEM supplemented with 0.2% FBS, 1× penicillin-streptomycin, and 50 µg/mL sodium ascorbate), which was replaced every other day. Six days later, the cells were lysed, and the ECM was preserved, labeled with anti-collagen I, and photographed via inverted fluorescence microscopy [34].

Immunohistochemistry (IHC)

The tissues were fixed in 4% paraformaldehyde (4 °C, overnight), paraffin-embedded, and sectioned (4–6 µm). Following deparaffinization and antigen retrieval (10 mM citrate buffer, pH 6.0), endogenous peroxidases were blocked with 3% H₂O₂. After the sections were incubated with either anti-ITGB1 (Santa Cruz, #sc365970) or anti-periostin (CST, #13730), HRP-conjugated secondary antibodies (Servicebio, #11022) were added. Hematoxylin counterstaining on a DAB substrate was used for visualization.

Tissue microarrays and patient cohorts

In this study, three human NSCLC tissue microarrays (TMAs) were used: IWLT-N-96SL91 (SLC-1901), IWLT-N-126AL91 (ALC-1901), and IWLT-C-102L41 (LAC-1401). A stringent inclusion criterion for all patients represented in these TMAs was their treatment-naïve status and having received no prior radiotherapy, chemotherapy, targeted therapy, or immunotherapy before surgical resection. The IWLT-N-96SL91 TMA included 48 pathologically confirmed LUSC cases. The IWLT-N-126AL91 TMA contained 63 pathologically confirmed LUAD cases. The IWLT-C-102L41 TMA included 102 distinct NSCLC cases, each represented by a single tissue core. This specific cohort included patients with diverse histological compositions: 43 with LUSC, 48 with LUAD, 2 with adenosquamous carcinomas, 4 with small cell lung cancers, 3 with neuroendocrine carcinomas, 1 with composite large cell carcinoma and 1 with sarcomatoid carcinoma. For all cases incorporated across these TMAs, comprehensive clinicopathological characteristics were meticulously collected. These data included patient age, sex, tumor location, histological type, differentiation grade, maximum tumor diameter, lymph node metastasis status, T/N/M classification and final TNM stage.

Western blot analysis

The primary antibodies used included anti-CD248 (CST, #47948), anti-vimentin (Servicebio, #GB111308), anti-YAP1 (Servicebio, #GB113975), anti-phospho-YAP1 (Ser127) (Servicebio, #GB114060), anti-periostin (Absin, #abs136266), anti-collagen I (Servicebio, #GB11068), anti-FAK (Absin, #abs155213), anti-phospho-FAK (Y397) (Servicebio, #GB115455), anti-Src (Servicebio, #GB112343), anti-phospho-Src (Tyr419) (Absin, #abs130790), anti-E-cadherin (Absin, #abs130068), anti-N-cadherin (Absin, #abs131133), anti-ZEB1 (Absin, #abs136926), anti-SNAI1 (Absin, #abs131634), anti-TWIST1 (Absin, #abs131127), anti-ITGB1 (Servicebio, #GB115173), and anti-GAPDH (Servicebio, #11002). Before being treated with primary and HRP-conjugated secondary antibodies (Servicebio, #GB23303; #GB23301), the proteins were separated via SDS-PAGE, transferred to PVDF membranes, and blocked with 5% BSA in TBST for 2 hours at room temperature. Three separate experiments were conducted. With improved chemiluminescence, the protein bands were visualized.

Biomechanical characterization

Atomic force microscopy (AFM; Bruker BioScope Catalyst) with pyramidal probes (cantilever spring constant: 0.02 N/m) was used to determine the Young’s modulus of different fibroblast populations. Data from 100 cells per group were averaged across three independent experiments. The data are expressed as the mean ± SD.

Quantitative PCR (qPCR)

PrimeScript RT Master Mix (Takara, #RR063A) was used to reverse transcribe the total RNA after it was obtained via the MiniBEST Universal RNA Isolation Kit (Takara, #9767). Using the primer sequences indicated in Supplementary Table S1, qPCR was carried out via the TB Green Premix Ex Taq II Kit (Takara, #RR820).

Enzyme-linked immunosorbent assay (ELISA)

For 48 hours, NFs, CAFs-sh-CON, CAFs-sh-CD248, and CAFs-CD248OE were cultivated in DMEM/F12. The periostin levels in the culture supernatants were measured via an ELISA kit (Novusbio, #NBP3-18052) in accordance with the guidelines provided by the manufacturer.

RNA sequencing and Bioinformatic Analysis

Total RNA was extracted from NFs, CAFs, A549 cells, and A549+rhPeriostin cells via TRIzol reagent. The RNA integrity was verified (RIN > 8.0, Agilent Bioanalyzer 2100). Libraries were prepared via poly(A) selection and sequenced (Illumina NovaSeq 6000; 150-bp paired-end) by Genechem Co. (Shanghai). Genes that were differentially expressed (|log₂FC| >1, FDR < 0.05) were examined via KEGG pathway mapping and GO enrichment. The FPKM technique was used to standardize the expression levels, and DESeq2 software (version 1.16.1) with Benjamini-Hochberg correction was used to perform differential expression analysis.

Luciferase reporter assay

HEK-293T cells were transfected with 50 ng of the Renilla luciferase plasmid and 200 ng of either the pGL3-POSTN, pcDNA3.1-YAP1, or control plasmid. Forty-eight hours after transfection, luciferase activity was assessed via the Dual-Luciferase® Reporter Assay System (Promega, #E2920).

Coimmunoprecipitation (Co-IP)

Co-IP was carried out using an Absin Co-IP Kit (#abs9649). The cells were cultured to 70–80% confluence and then incubated in serum-free medium for 4 hours. After being washed, the cells were treated with serum-free medium supplemented with 0.8 μg/mL recombinant periostin and incubated at 4 °C with gentle agitation for 2 hours. The cells were then lysed with prechilled nondenaturing membrane protein lysis buffer for 30 minutes. Following centrifugation at 12,000 rpm at 4 °C for 15 minutes, the supernatant was collected. Ten percent of the lysate was reserved as an input control. Equal amounts of protein samples were divided into two groups: the experimental group was incubated with anti-periostin antibody (Santa Cruz, catalogue no. sc398631), while the negative control group received rabbit IgG isotype control antibody (2 μg per 1 mg total protein). The samples were incubated overnight at 4 °C. Protein A/G agarose beads were then added, and the samples were incubated at 4 °C for 3 hours, followed by washing with prechilled lysis buffer 4–5 times. Bound proteins were eluted by boiling in 1× SDS loading buffer and subsequently analyzed by immunoblotting.

Parallel immunoprecipitation assays were performed to verify protein-protein interactions using anti-YAP1 antibody (Santa Cruz, #sc101199) and anti-CD248 antibody (Santa Cruz, #sc377221), following the same protocol described above.

In vivo tumor experiments

For the lung metastasis model, CD248 gene knockout mice (both WT and cKO mice) or POSTN knockout mice (both WT and KO mice) with fibroblasts of the same age and sex were given 5 × 10⁵ LLC-luciferase cells intravenously in 200 µl of medium. For the lung subcutaneous tumor model, age- and sex-matched fibroblast-specific CD248 gene knockout mice (both WT and cKO animals) or POSTN knockout mice (both WT and KO mice) received an injection of 1 × 10⁶ LLC-luciferase cells in 200 µL of media into the dorsal flank. After D-luciferin injection, bioluminescence imaging was used to track tumor progression. At the end of the experiment, the mice were allowed to sleep with isoflurane, and the tumors were removed, captured with a camera, and prepared for histological examination.

Statistical analysis

The data were analyzed via Excel 2019. Normality and homogeneity of variance were assessed before analysis. For two-way unpaired comparisons, Student’s t test was used, and for multigroup comparisons, either one-way or two-way ANOVA was used. All quantitative data were obtained from at least three independent biological replicates. The findings are displayed as the means ± SDs or means ± SEMs. The criteria for statistical significance were *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Results

Secretion of periostin by CD248-expressing CAFs is correlated with poor prognosis in NSCLC

We verified that CD248 is highly expressed in CAFs originating from NSCLC, in accordance with our earlier findings [15, 18, 33, 35] (Supplementary Fig. S1 A-D). CAFs with stable CD248 overexpression or knockdown were effectively produced (Supplementary Fig. S1 E-F).

We used RNA sequencing to examine the gene expression profiles of CAFs and NFs to clarify the molecular mechanisms underlying NSCLC metastasis. Differentially expressed genes were identified via volcano plot analysis; in CAFs, POSTN expression was significantly upregulated (Fig. 1A). POSTN expression and the proportion of CAFs were associated with TIMER 3.0 database analysis (Fig. 1B). IHC analysis revealed that 102 NSCLC tissues were positive for periostin expression (Fig. 1C–D). In both our patient cohort and the GEPIA database analysis, patients with higher POSTN levels had shorter survival periods than those with lower POSTN expression did (p < 0.05) (Fig. 1E). Periostin levels did not correlate with sex or age according to a statistical study of clinicopathological features. The periostin level was greater in patients with lymph node metastasis (p = 0.002), advanced tumor stage (p = 0.0039), moderately/poorly differentiated tumors (p = 0.0145), and higher TNM stage (p = 0.0237) (Tables 1 and 2). POSTN expression in CAFs was much greater than that in NFs (p < 0.0001), as determined by qPCR and Western blot analysis (Fig. 1F–G). POSTN and CD248 expression in LUAD and LUSC tissues was positively correlated according to the results of the GEPIA database analysis (Fig. 1H). Functional studies demonstrated that CD248 knockdown decreased periostin protein levels, whereas CD248 overexpression increased periostin expression (Fig. 1I–J). Periostin colocalized with CD248 and α-SMA in NSCLC tissues, as shown by IF staining (Fig. 1K–L). Reduced periostin secretion from CAFs-sh-CD248 was confirmed by ELISA (p < 0.001) (Fig. 1M). These findings suggest that periostin, which is secreted by CD248⁺CAFs, is linked to poor prognosis and clinicopathological characteristics in NSCLC, including tumor differentiation, tumor stage, lymph node metastasis, and TNM stage.

Fig. 1.

CD248-expressing CAFs secrete periostin. A. Using RNA sequencing data, a volcano diagram showing the genes that are differentially expressed (DEGs) between NFs and CAFs is shown. B. Periostin expression levels in CAFs from NSCLC patients were determined via TIMER 3.0 database analysis. C-D. representative images of periostin immunohistochemical staining in NSCLC tissues classified by the extent of staining: score 0 (negative), score 1 (6–25%), score 2 (26–50%), score 3 (51–75%), and score 4 (76–100%). Scale bars: 5000 μm and 50 μm. E. Kaplan-meier curves displaying overall survival in NSCLC patients categorized by high or low POSTN expression. *p < 0.05.F. qPCR analysis of POSTN mRNA expression in NFs and CAFs. The data are shown as the mean ± SD; n = 3 independent biological replicates. G. Periostin protein levels in NFs and CAFs were measured via western blotting. The data are shown as the mean ± SD; n = 3 independent biological replicates. H. LUAD and LUSC datasets: correlation between CD248 and POSTN expression. I.POSTN mRNA levels in CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE. The data are shown as the mean ± SD; n = 3 independent biological replicates. J. Periostin protein expression in CAFs, CAFs-sh-CD248 and CAFs-CD248OE, as determined by western blot analysis. The data are shown as the mean ± SD; n = 3 independent biological replicates. K. NAT and NSCLC tissues were subjected to immunofluorescence staining for α-SMA (pink), CD248 (red), periostin (green) and DAPI (blue). Scale bar: 50 μm. L. CD248 (red), periostin (green) and DAPI (blue) immunofluorescence staining of CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE. Profiles of fluorescence intensity and enlarged views are presented. Scale bars: 20 μm and 10 μm. M. Periostin levels in conditioned medium from NFs, CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE were quantified by ELISA. The data are shown as the mean ± SD; n = 3 independent biological replicates

Table 1.

Clinicopathological features of 102 patients with non-small cell lung cancer

Clinicopathological features	No. of patients (%)
Sex
Male	70(69)
Female	32(31)
Age, yrs
Mean	52
Range	40–81
Tumor differentiation
Well	26(25)
Moderately	39(38)
Poorly	37(37)
T stage
T1	41(40)
T2	39(38)
T3	15(15)
T4	7(7)
N stage
N0	69(68)
N1	16(16)
N2	17(16)
TNM stage
I	50(49)
II	28(27)
III	22(22)
IV	2(2)
TNM, tumor-node-metastasis.

Table 2.

Correlation between periostin expression and clinicopathological features in patients with non-small cell lung cancer (n = 102)

Clinicopathological features	N	Periostin expression Low High		χ2	Pvalue
Age				0.9865	0.3206
<60	47	21	26
≥60	55	30	25
Sex				0.21	0.6467
Male	70	23	47
Female	32	12	20
Tumor differentiation				5.973	0.0145*
Well	26	16	10
Moderately/Poorly	76	26	50
Tumor stage				8.345	0.0039*
T1—2	80	46	34
T3—4	22	5	17
Lymph node metastasis				9.560	0.002*
N0	69	37	32
N1-2	33	7	26
TNM stage				5.120	0.0237*
I—II	78	40	38
III—IV	24	6	18
TNM, tumor-node metastasis.

^*Statistically significant (p < 0.05)

Periostin induces EMT-like phenotypic changes and enhances NSCLC invasion and migration

According to earlier research, CD248-expressing CAFs stimulate NSCLC EMT-like phenotypic changes by controlling M2-polarized macrophages [21]. We examined EMT-related proteins and genes and revealed that CM from A549 and NCI-H460 cells increased the expression of mesenchymal markers (N-cadherin, vimentin, TWIST1, ZEB1 and SNAI1) but decreased the expression of epithelial E-cadherin (Fig. 2A). CD248-overexpressing CM from CAFs upregulated EMT-related genes (p < 0.05), whereas CD248-silenced CM from CAFs had no significant effect (Fig. 2B–C). CD248 knockdown increased E-cadherin expression and decreased mesenchymal marker expression, whereas the opposite effects were observed following CD248 overexpression (Fig. 2D). Rescue experiments using rhPeriostin or anti-periostin-neutralizing antibodies demonstrated that rhPeriostin treatment increased the expression of mesenchymal markers while decreasing E-cadherin expression. Conversely, the anti-periostin antibody antagonized CAFs-CM-mediated EMT-related markers upregulation (Fig. 2E).

Fig. 2.

Periostin induces EMT-like phenotypic changes and promotes invasion and migration in NSCLC cells. A. Western blot analysis revealed markers associated with EMT in A549 and NCI-H460 cells treated with CM from NFs or CAFs. These markers include E-cadherin, N-cadherin, vimentin, ZEB1, TWIST1 and SNAI1. The data are shown as the mean ± SD; n = 3 independent biological replicates. B-C. qPCR analysis of ZEB1, SNAI1, and TWIST1 mRNA expression in A549 and NCI-H460 cells treated with CM from CAFs-sh-CON, CAFs-sh-CD248 or CAFs-CD248OE. The data are shown as the mean ± SD; n = 3 independent biological replicates. D. Western blot analysis of EMT-associated proteins in CM from CAFs-sh-CON, CAFs-sh-CD248 or CAFs-CD248OE-treated A549 and NCI-H460 cells. The data are shown as the mean ± SD; n = 3 independent biological replicates. E. western blot analysis of EMT-related markers in A549 and NCI-H460 cells treated with CM from CAFs-sh-CON, CAFs-sh-CD248, CAFs-sh-CD248 supplemented with rhPeriostin or CAFs-sh-CON supplemented with anti-periostin antibody. The data are shown as the mean ± SD; n = 3 independent biological replicates. F.POSTN mRNA expression in CD248OE CAFs with or without periostin knockdown (sh-periostin #1-#3), as determined by qPCR. The data are shown as the mean ± SD; n = 3 independent biological replicates. G. Periostin protein levels in the CAFs-CD248OE and periostin-knockdown strains (sh-periostin #1-#3) were analyzed via western blotting. The data are shown as the mean ± SD; n = 3 independent biological replicates. H. CD248 (red), periostin (green), and DAPI (blue) immunofluorescence staining of CAFs-CD248OE and CAFs-CD248OE-sh-Periostin. The fluorescence intensity profiles and merged images are shown. Scale bars: 20 μm and 10 μm. I-M. Transwell assays for invasion and migration. CM from CAFs-sh-CON, CAFs-sh-CD248, CAFs-sh-CD248+rhPeriostin, CAFs-CD248OE or CAFs-CD248OE-sh-Periostin was added to A549 and NCI-H460 cells for 48 hours. I. Schematic of the experimental design. J. A549 cell results after 24 hours. K. A549 cell invasion and migration were quantified after 24 hours. L. Leads of NCI-H460 cells after 24 hours. M. Number of migratory and invasive NCI-H460 cells after 24 hours. The data are shown as the mean ± SD; n = 3 independent biological replicates. Scale bar: 200 μm

We successfully generated CAFs with periostin knockdown under CD248 overexpression conditions (Fig. 2F–H). Invasion and migration assays revealed significantly greater effects in NSCLC cells cocultured with rhPeriostin or CAFs-CD248OE than in control cells (p < 0.001) (Fig. 2I–M). These findings show that periostin stimulates EMT-like phenotypic changes and promotes the invasion and migration of NSCLC cells.

The periostin-binding receptor ITGB1 is highly expressed in NSCLC and is associated with poor prognosis

Previous studies have indicated that periostin promotes ovarian cancer metastasis through binding to ITGB1 [36] and that ITGB1 induces EMT-like phenotypic changes in NSCLC [37]. Tissue microarray analysis revealed ITGB1 expression patterns in LUAD and LUSC samples (Fig. 3A). When LUSC and LUAD tissues were compared to normal adjacent tissue (NAT), IHC staining revealed noticeably increased ITGB1 H-scores (p < 0.0001) (Fig. 3B–C). High ITGB1 expression was associated with a lower overall survival rate in both LUAD patients and LUSC patients according to Kaplan-Meier analysis (Fig. 3D). Clinicopathological analysis (Supplementary Tables S2–S3, Fig. 3E–F) revealed no correlation between ITGB1 levels and age or sex. However, patients with moderately/poorly differentiated tumors (p = 0.0457 in LUSC, p = 0.0187 in LUAD), advanced tumor stage (p = 0.0453 in LUAD, p = 0.0192 in LUSC), and lymph node metastasis (p = 0.0112 in LUAD, p = 0.031 in LUSC) presented elevated ITGB1 expression. Compared with BEAS-2B cells, A549 cells presented significant ITGB1 expression, whereas NCI-H460 cells presented moderate ITGB1 expression (Fig. 3G–H). ITGB1 colocalization in A549 and NCI-H460 cells was verified by IF experiments (Fig. 3I). Periostin and ITGB1 exhibited persistent binding, according to HDOCK analysis (docking score: −269.66; confidence score: 0.9163) (Fig. 3J–K). Co-IP analysis validated the interaction between periostin and ITGB1 in NSCLC cells (Fig. 3L–M). These findings demonstrate that ITGB1 is upregulated in NSCLC tissues and is associated with poor clinical outcomes and that it interacts with periostin in NSCLC cells, suggesting that the periostin-ITGB1 axis plays an important role in NSCLC.

Fig. 3.

ITGB1 is highly expressed in NSCLC. A. representative images of TMAs containing LUAD and LUSC samples. Scale bar: 5000 μm. B. ITGB1 expression in NAT and NSCLC tissues was detected via IHC. Scale bar: 50 μm. C. ITGB1 expression in NAT and NSCLC tissues (including 63 LUAD and 48 LUSC) was measured via the H score. The data are displayed as the mean ± SD; ****p < 0.0001. D. Kaplan-meier analysis of overall survival in LUAD and LUSC patients stratified by high or low ITGB1 expression in the TCGA database. E-F. ITGB1 expression and clinicopathological characteristics in LUSC (E) and LUAD (F) patients are correlated. G. qPCR analysis of ITGB1 mRNA expression in NSCLC cell lines (H1299, H23, NCI-H460 and A549) and normal lung epithelial cells (BEAS-2B). The data are shown as the mean ± SD; n = 3 independent biological replicates. H. western blot analysis of ITGB1 protein levels in BEAS-2B and NSCLC cell lines. The data are shown as the mean ± SD; n = 3 independent biological replicates. I. IF staining for ITGB1 (red) and DAPI (blue) in A549 and NCI-H460 cells. Scale bars: 20 μm and 10 μm. J-K. molecular docking analysis of periostin and ITGB1 via HDOCK. J. Superposition of optimal conformations. K. Detailed view of the predicted binding site. L-M. Co-IP was used to assess the periostin-ITGB1 interaction in the designated A549 and NCI-H460 groups

CD248⁺ CAFs secrete periostin to induce EMT-like phenotypic changes via the integrin-FAK-Src pathway

RNA-seq gene enrichment analysis revealed strong activation of the focal adhesion kinase pathway by periostin (Fig. 4A). The literature has confirmed that integrins primarily transduce downstream signals through the FAK/Src signaling pathway, thereby regulating a series of cellular processes, including proliferation, migration and survival [27, 38]. IF staining revealed that CD248⁺CAFs CM increased paxillin puncta formation, which was reduced by CD248 knockdown and restored by rhPeriostin supplementation (Fig. 4B). Cytoskeletal rearrangement and altered actin/paxillin distribution suggested enhanced cell motility. CD248⁺CAFs CM increased the expression levels of p-FAK and p-Src, which were counteracted by anti-periostin antibodies or CD248 knockdown and reversed by rhPeriostin, as shown by western blot analysis (Fig. 4C–D). rhPeriostin-induced EMT-related markers expression was inhibited by FAK and Src inhibitors (Fig. 4E).

Fig. 4.

CD248-positive CAFs promote EMT-like phenotypic changes through periostin-mediated activation of the integrin-FAK-Src pathway. A. To determine which signaling pathways are affected by rhPeriostin, RNA-seq data were subjected to GO enrichment analysis. B. A549 and NCI-H460 cells treated with CM from various CAFs groups, and then focal adhesions formation was examined by paxillin immunostaining (green). Nuclei were stained with DAPI (blue), and actin cytoskeleton was stained with phalloidin (red). Quantification of FA numbers. The data are shown as the mean ± SD; n = 3 independent biological replicates. Scale bar: 10 μm. C. Total and phosphorylated FAK and src levels were measured via western blotting in A549 and NCI-H460 cells treated with CM from different CAFs. The data are shown as the mean ± SD; n = 3 independent biological replicates. D. A549 and NCI-H460 cells treated with CM were subjected to western blot analyses of FAK, p-FAK (Y397), Src, p-Src (Tyr419) from CAFs-sh-CON, CAFs-sh-CD248, CAFs-sh-CD248 supplemented with rhPeriostin (0.8 μg/mL) and CAFs-CD248OE and CAFs-CD248OE with anti-periostin antibody (3 μg/mL). The data are shown as the mean ± SD; n = 3 independent biological replicates. E. Western blot analysis of FAK, p-FAK, Src, p-Src and EMT-related proteins (N-cadherin, E-cadherin, vimentin, ZEB1, SNAI1, and TWIST1) in A549 and NCI-H460 cells subjected to specific treatments. The data are shown as the mean ± SD; n = 3 independent biological replicates. F.ITGB1 and EMT-related genes expression correlation study in LUAD and LUSC tumor tissues. G. IF staining for ITGB1 (red) and DAPI (blue) in A549, A549-sh-ITGB1, NCI-H460 and NCI-H460-sh-ITGB1 cells. The combined images demonstrate the effectiveness of ITGB1 knockdown. Scale bars: 50 μm and 10 μm. H. ITGB1 expression in A549, A549-sh-ITGB1, NCI-H460, and NCI-H460-sh-ITGB1 cells was examined via western blotting. The data are shown as the mean ± SD; n = 3 independent biological replicates. I. A549-sh-ITGB1 and NCI-H460-sh-ITGB1 cells treated with rhPeriostin, and then focal adhesions formation was examined by paxillin immunostaining (green). Nuclei were stained with DAPI (blue), and actin cytoskeleton was stained with phalloidin (red). Quantification of FA numbers. The data are shown as the mean ± SD; n = 3 independent biological replicates. Scale bar: 10 μm. J-K. Western blot analysis of A549, A549-sh-ITGB1, NCI-H460 and NCI-H460-sh-ITGB1 cells treated with CM from CAFs-sh-CON or rhPeriostin to detect FAK, p-FAK, Src, p-Src and EMT-related markers. The data are shown as the mean ± SD; n = 3 independent biological replicates

GEPIA revealed a strong correlation between the expression of ITGB1 and that of EMT-related genes (SNAI1, TWIST1 and ZEB1) in LUAD and LUSC (Fig. 4F). ITGB1 knockdown in A549 and NCI-H460 cells (Fig. 4G–H) reduced paxillin puncta formation, which could not be restored by rhPeriostin (Fig. 4I). ITGB1 knockdown also antagonized the CAFs-sh-CON CM or rhPeriostin-induced upregulation of p-FAK, p-Src and EMT-related markers expression (Fig. 4J–K). These results suggested that ITGB1 was involved in the cellular response to Periostin and that the binding of Periostin with ITGB1 activated the FAK/Src pathway to induce NSCLC cell EMT-like phenotypic changes.

YAP1 activation induces periostin secretion by CD248⁺ CAFs

Our previous studies demonstrated that YAP1 is enriched in CAFs and that CD248-mediated collagen I secretion occurs through YAP1 activation [33]. Western blot analysis revealed that CD248 knockdown suppressed total and phosphorylated YAP1, whereas CD248 overexpression increased the expression of both forms (Fig. 5A). IF staining revealed increased nuclear localization of YAP1 in the CAFs-CD248OE cells (Fig. 5B). YAP1 silencing in CD248⁺CAFs reduced periostin expression (Fig. 5C). Co-IP confirmed the YAP1–CD248 interaction (Fig. 5D). qPCR revealed elevated POSTN mRNA expression in the CAFs-YAP1OE line (p < 0.05) (Fig. 5E). Western blot analysis revealed that YAP1 knockdown reduced periostin expression, which was restored by YAP1 overexpression (Fig. 5F). A luciferase assay revealed that YAP1 mediated POSTN promoter activation (p < 0.0001) (Fig. 5G). These results indicate that YAP1 regulates periostin secretion by CD248⁺CAFs.

Fig. 5.

Activation of YAP1 induces periostin secretion in CD248-positive CAFs. A. western blot analysis of YAP1 and p-YAP1 expression in CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE; the relative ratio of p-YAP1 (Ser127) to total YAP1 was calculated for quantitative analysis. The data are shown as the mean ± SD; n = 3 independent biological replicates. B. YAP1 nuclear localization in CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE was demonstrated by immunofluorescence staining. Scale bars: 50 μm and 10 μm. C. CD248 (red) and periostin (green) immunofluorescence staining of CAFs-sh-CON and CAFs-sh-YAP1. Scale bars: 50 μm and 10 μm. D. Co-IP was used to assess the CD248–YAP1 interaction in the designated CAFs groups. E.POSTN mRNA expression in CAFs-sh-CON, CAFs-sh-YAP1, CAFs-sh-CD248 and YAP1-overexpressing CAFs (CAFs-YAP1OE) was examined via qPCR. The data are shown as the mean ± SD; n = 3 independent biological replicates. F. YAP1, p-YAP1 and periostin protein levels in CAFs-sh-CON, CAFs-sh-YAP1, CAFs-sh-CD248–YAP1OE and CAFs-sh-CD248 were analyzed via western blotting; the relative ratio of p-YAP1 (Ser127) to total YAP1 was calculated for quantitative analysis. The data are shown as the mean ± SD; n = 3 independent biological replicates. G. A luciferase reporter assay was used to evaluate the YAP1-mediated regulation of POSTN promoter activity. The mean ± SD is displayed; there are four technical replicates; ****, p < 0.0001

Periostin induced collagen I deposition activates YAP1 to promote periostin secretion

By altering the tumor microenvironment, particularly its extracellular matrix, periostin accelerates tumor growth [22, 24, 39, 40]. An examination of the TIMER 3.0 database revealed a positive association between the proportions of CAFs and collagen I expression (Fig. 6A). The ECM is strongly related to CD248. Research has shown that collagen I is its ligand, and through controlling ECM-mediated signaling pathways, it can take part in activities such as cell adhesion, migration and tissue remodeling [41–43]. HDOCK analysis indicated stable binding between CD248 and collagen I (docking score: −304.44; confidence score: 0.9564) (Fig. 6B–C). Collagen I and periostin were colocalized in the CAFs-sh-CON and CAFs-CD248OE groups, whereas the expression of those genes was lower in the CAFs-sh-CD248 and CAFs-sh-Periostin groups, as determined by IF staining (Fig. 6D). According to the GEPIA database, NSCLC subtypes and collagen I expression were positively correlated (Fig. 6E). Co-IP analysis confirmed the interaction between CD248 and collagen I in CD248-overexpressing CAFs (Fig. 6F). In the CAFs-sh-CD248 and CAFs-sh-Periostin groups, western blot and IF analyses revealed decreased collagen I expression, which was restored by rhPeriostin supplementation (Fig. 6G–H). In the CAFs-CD248OE group, the results of the ECM assembly assays revealed strong collagen I fluorescence, in contrast to the weak signals observed in the other groups (Fig. 6I). Western blot analysis revealed that the levels of YAP1, p-YAP1, and collagen I were increased in the CAFs-sh-CD248–YAP1OE and CAFs-CD248OE groups but decreased in the CAFs-sh-CD248 and CAFs-CD248OE-sh-Periostin groups (Fig. 6J). When NSCLC tissues were compared with NAT, IF staining revealed greater colocalization of CD248, periostin, collagen I and YAP1 (Fig. 6K). These results suggest that CD248⁺CAFs stimulate the deposition of collagen I via the secretion of periostin. Collagen I then binds to CD248 and increases the production of periostin by activating YAP1.

Fig. 6.

Periostin induces collagen I deposition to bind CD248 activates YAP1 to promote periostin secretion. A. Collagen I expression levels in CAFs from NSCLC patients, as determined from data from the TIMER 3.0 database. B-C. CD248 and collagen I binding site analysis and superimposed optimum conformations produced by HDOCK molecular docking. D. Colocalization of collagen I, periostin and CD248 in CAFs-sh-CON, CAFs-sh-CD248 and CAFs-CD248OE was demonstrated by immunofluorescence labeling. Scale bars: 20 μm, 50 μm and 10 μm. E. COL1A1 and POSTN expression correlation study in the LUAD and LUSC datasets. F. Co-IP was used to assess the CD248-collagen I interaction in the designated CAFs-CD248OE groups. G. Collagen I expression in CAFs-sh-CON, CAFs-sh-CD248 supplemented with rhPeriostin, CAFs-sh-Periostin and CAFs-sh-CON was examined via western blot analysis. The data are shown as the mean ± SD; n = 3 independent biological replicates. H. Collagen I levels in CAFs-sh-CON, CAFs-sh-Periostin, CAFs-sh-CD248 and CAFs-sh-CD248+rhPeriostin were detected via immunofluorescence. Scale bars: 50 μm and 10 μm. I. Collagen I deposition in representative immunofluorescence images. Scale bar: 50 μm. J. YAP1, p-YAP1 and collagen I levels in CAFs-sh-CON, CAFs-sh-CD248, CAFs-sh-CD248–YAP1OE, CAFs-CD248OE and CAFs-CD248OE-sh-Periostin; the relative ratio of p-YAP1 (Ser127) to total YAP1 was calculated for quantitative analysis. The data are shown as the mean ± SD; n = 3 independent biological replicates. K. Colocalization of CD248, periostin and collagen I in LUSC, LUAD and NAT samples is shown in representative dual immunofluorescence images. Scale bars: 2000 μm, 50 μm, and 10 μm

Fibroblast-specific CD248 knockout inhibits tumor metastasis in vivo

To evaluate the role of CD248 in tumor metastasis, we subcutaneously injected LLC-L cells into WT and cKO mice (Fig. 7A). Bioluminescence imaging revealed substantially greater tumor growth in WT mice than in CD248-deficient mice (Fig. 7B–C). The tumor size and weight were significantly reduced in the cKO mice (p < 0.01) (Fig. 7D–E). In WT tumors, H&E staining revealed more necrotic lesions (Fig. 7F). Compared with control tumors, cKO tumors contained lower proportions of N-cadherin and more E-cadherin (p < 0.0001) (Fig. 7G–I), with CD248, collagen I, and periostin expression decreased (Fig. 7J); Moreover, compared with control tissues, cKO tissues presented a decreased Young’s modulus (Fig. 7K–L).

Fig. 7.

CD248 knockout specific to fibroblasts prevents tumor growth and metastasis in vivo. A-L. to create a subcutaneous tumor model, LLC-L cells (1 × 10⁶ in 200 μL) were injected subcutaneously into cd248^fl/flfsp-1^+/+ (WT) and cd248^fl/flfsp-1^cre/+ (cKO) mice (n = 3 per group). A. Schematic of the subcutaneous tumor model. B. Bioluminescence (BLI) imaging was used to track tumor growth. C. Average radiance intensity by group quantification. The data are displayed as the mean ± SEM; n = 3 mice per group. *p < 0.0001.D. At the endpoint, the mice were euthanized. Subcutaneous tumor photos that are representative of each category. E. Quantification of tumor weight. The data are displayed as the mean ± SEM; n = 3 mice per group. ****p < 0.05.F. H&E staining of tumor tissue. Scale bars: 2000 μm and 50 μm. G. western blot analysis of the expression of E-cadherin and N-cadherin in malignancies (n = 3 mice per group). H. E-cadherin (red) and N-cadherin (green) expression was detected via dual immunofluorescence in tumor tissues from WT and cKO mice. Scale bars: 2000 μm and 50 μm. I. N-cadherin-positive area quantification. The data are displayed as the mean ± SEM; n = 3 mice per group. *p < 0.0001.J. Colocalization of periostin (green), collagen I (red) and CD248 (yellow) in tumor tissues was demonstrated by immunofluorescence labeling. Scale bars: 50 μm, 20 μm and 10 μm. K-L. the Young’s modulus of tumor tissues from WT and cKO mice was measured, which indicates the stiffness of the extracellular matrix. The data are displayed as the mean ± SEM; n = 3 mice per group. ****p < 0.01. M-S. WT and cKO mice (n = 3 per group) were injected with LLC-L cells (5 × 10⁵) to establish a metastatic model. M. Schematic of the experimental metastatic model. N. BLI of ex vivo organs, such as the brain, lung, heart, spleen, liver, kidney and bowel. O. Fluorescence intensity in each organ was measured. The data are presented as the mean ± SEM; n = 3 mice per group. ****p < 0.0001. P. Typical pictures of lungs with white arrows indicating metastases. Q. Number of metastatic nodules in each lung. The data are shown as the mean ± SEM; n = 3 mice per group. *p < 0.05. R. Lung metastatic nodule H&E staining. Scale bar: 50 μm. S. Periostin (green), collagen I (red) and CD248 (yellow) immunofluorescence staining of lung metastases. Scale bars: 2000 μm, 50 μm and 10 μm

Intravenous LLC-L injection resulted in strong lung fluorescence in WT mice but faint signals in cKO mice (p < 0.0001) (Fig. 7M–O). H&E staining revealed larger, more extensive lung metastatic nodules in WT mice (p < 0.0001) (Fig. 7P–R), with increased CD248, collagen I and periostin expression (Fig. 7S). These findings suggest that fibroblasts expressing CD248 secrete periostin, which facilitates tumor spread.

POSTN knockout inhibits tumor metastasis in vivo

We injected LLC-L cells subcutaneously into WT and KO mice to evaluate the function of periostin in tumor metastasis (Fig. 8A). Compared with that in KO animals, bioluminescence imaging revealed much greater tumor growth in WT mice (Fig. 8B–C). KO mice presented substantial decreases in tumor size and weight (p < 0.01) (Fig. 8D–E). In WT tumors, H&E staining revealed more necrotic lesions (Fig. 8F). The proportion of N-cadherin-positive cells was reduced in the KO mice (Fig. 8G–H), and the expression of CD248, collagen I and periostin was decreased (Fig. 8I). The proportion of proliferating cells decreased in the KO mice (Fig. 8J–K), with a lower Young’s modulus (Fig. 8L–M).

Fig. 8.

POSTN knockout inhibits tumor growth and metastasis in vivo. A-M. to create a subcutaneous tumor-bearing model, LLC-L cells (1 × 10⁶ in 200 μL) were subcutaneously injected into POSTN^+/+ (WT) and POSTN^-/- (KO) mice (n = 3 per group). A. Schematic of the subcutaneous tumor model. B. BLI tracks the progression of the tumor. C. Average intensity of the radiance for each group. The data are displayed as the mean ± SEM. n = 3 mice per group. *p < 0.0001.D. At the endpoint, the mice were sacrificed. Typical subcutaneous tumor images. E. Quantification of tumor weight. The data are displayed as the mean ± SEM. n = 3 mice per group. ****p < 0.05.F. Tumor tissue stained with H&E. Scale bar: 50 μm. G. E-cadherin (red) and N-cadherin (green) expression was detected via dual immunofluorescence. Scale bars: 2000 μm and 50 μm. H. Quantification of the N-cadherin-positive area. The data are displayed as the mean ± SEM. n = 3 mice per group. *p < 0.0001.I. Immunofluorescence staining of periostin (green), collagen I (red), and CD248 (yellow). Scale bars: 50 μm, 20 μm and 10 μm. J. representative images of IHC staining for ki-67. Scale bars: 50 μm, 20 μm and 10 μm. K. Quantification of ki-67-positive cells. The data are displayed as the mean ± SEM. n = 3 mice per group. ****p < 0.05. L-M. Young’s modulus of tumor tissues from WT and KO mice. The data are displayed as the mean ± SEM. n = 3 mice per group. **p < 0.01. N-S. WT and KO mice (n = 3 per group) were injected with LLC-L cells (5 × 10⁵) for the metastatic model. N. Schematic of the experimental metastatic model. O. BLI tracking of tumor spread. P. Metastatic nodules in different organs stained with H&E. Scale bar: 100 μm. Q. Typical pictures of lungs with white arrows indicating metastases. R. Number of metastatic nodules in each lung. The data are displayed as the mean ± SEM. n = 3 mice per group. ****p < 0.01; data are presented as the mean ± SEM. S. Periostin (green), collagen I (red) and CD248 (yellow) immunofluorescence staining of lung metastases. Scale bars: 50 μm, 20 μm and 10 μm. T. Schematic illustration of the proposed mechanism

Compared with KO mice, WT mice presented significantly more metastases after receiving an intravenous LLC-L injection (Fig. 8N–O). No metastatic nodules were detected in any other tissues (Fig. 8P). The lung metastatic nodules in the WT mice were considerably larger and more widespread (p < 0.0001) (Fig. 8Q–R), and the expression of CD248, collagen I and periostin was greater (Fig. 8S). According to our research, CD248 activates YAP1 to induce the release of periostin by CAFs. ITGB1 on NSCLC cells is then bound by periostin, which causes EMT-like phenotypic changes and promotes metastasis. In parallel, CD248 is bound by periostin-induced collagen I deposition, which increases periostin production by CAFs via a positive feedback loop (Fig. 8T).

Discussion

The intricate mechanisms underlying NSCLC metastasis continue to be intensely investigated, with CAFs emerging as critical regulators of this process [44–46]. While CAFs are well-known contributors to NSCLC progression through ECM remodeling, immune modulation, and stromal-tumor crosstalk [47–49], significant gaps remain in our understanding of specific CAFs subpopulations and the molecular cascades that drive metastasis. The functional diversity of CAFs and the key effectors governing their paracrine signaling to tumor cells remain incompletely defined [12, 47, 50]. Published studies have suggested that CD248⁺ CAFs exhibit functional and phenotypic overlap with inflammatory CAFs (iCAFs) [51]. Our previous work demonstrated that CD248⁺ CAFs secrete IL-8 and promote chemoresistance in tumor cells, consistent with iCAF characteristics [15]. In the current study, we additionally observed that CD248⁺ CAFs substantially enhance collagen I deposition and ECM remodeling, features typically associated with myofibroblastic CAFs (myCAFs) [33]. These findings suggest that CD248⁺ CAFs in NSCLC may represent a hybrid CAF population exhibiting functional properties of both iCAFs and myCAFs, rather than a strictly defined single subtype. Our previous work revealed CD248 as a marker for a functionally distinct CAFs subpopulation, demonstrating that CD248⁺ CAFs mediate macrophage M2 polarization, confer chemoresistance and immune evasion, and induce ECM remodeling to promote NSCLC progression. However, the downstream molecular mediators of CD248⁺ CAFs and their precise role in orchestrating metastatic processes have not been systematically defined, representing a critical knowledge gap in NSCLC research. The current study addresses this gap by elucidating a novel mechanistic pathway linking CD248⁺ CAFs to NSCLC metastasis through periostin-mediated signaling.

Owing to its roles in ECM remodeling, cell adhesion, and stromal-tumor interactions, periostin has been established as a crucial factor in cancer metastasis across multiple tumor types [22, 26, 27, 40]. As a crucial downstream effector in the canonical Hippo pathway, YAP1 plays a vital role in the development and spread of tumors [37, 52, 53]. In breast cancer, periostin secretion is driven by the YAP1 pathway [40]; however, this relationship has remained largely unexplored in the context of NSCLC metastasis and CAFs-tumor crosstalk. While previous research in NSCLC has suggested that periostin is involved in cell migration and invasion, its specific source, regulatory mechanisms, and downstream signaling targets in the metastatic cascade remain unaddressed. By identifying CD248⁺ CAFs as the main source of periostin in the NSCLC TME, our study overcomes these limitations. We showed that by promoting EMT-like phenotypic changes in tumor cells, periostin, which is released by CD248⁺CAFs, plays a crucial role in mediating NSCLC metastasis. We clarify the previously unknown function of periostin in NSCLC metastasis by establishing a direct functional connection between periostin secretion from CD248⁺ CAFs and increased invasion and migration of NSCLC cells in both in vitro and in vivo models.

Periostin works mechanistically by binding to integrin family receptors and triggering intracellular signaling cascades that control TME remodeling and cell activity [54, 55]. Previous studies have shown that periostin activates FAK and Src family kinases upon integrin binding, thereby enhancing cell-ECM adhesion, promoting cell survival, and facilitating ECM remodeling [27, 40, 56]. While multiple integrins have been implicated in periostin signaling, the specific receptor that mediates its prometastatic effects in NSCLC remains unknown. Our work revealed that ITGB1 is the key periostin receptor in NSCLC cells, demonstrating that the interaction between periostin and ITGB1 triggers the FAK/Src pathway to induce EMT-like phenotypic changes. We confirmed the essential role of ITGB1 by showing that its abrogation eliminates periostin-mediated EMT-like phenotypic changes and metastatic capacity, clarifying the critical function of the ITGB1-FAK/Src axis in transducing prometastatic signals of periostin. These findings not only delineate a specific signaling pathway that drives NSCLC metastasis but also highlight ITGB1 as a potential therapeutic target for disrupting CAFs-tumor crosstalk.

In addition to defining the downstream signaling pathway of periostin, our study revealed the upstream regulatory mechanism governing its secretion by CD248⁺ CAFs. We indicated that CD248 regulates YAP1 and nuclear localization in CAFs and that YAP1 regulates periostin expression. These findings link CD248, a marker of prometastatic CAFs, to the YAP1-periostin axis, a pathway previously uncharacterized in NSCLC. We observed increased levels of both total YAP1 and Ser127-phosphorylated YAP1. Importantly, the relative ratio of Ser127-phosphorylated YAP1 to total YAP1 was not increased, suggesting that the inhibitory Ser127 phosphorylation level per se did not rise proportionally. Therefore, the overall increase in total YAP1 protein may contribute to increased YAP1 activity and nuclear accumulation in our experimental setting. It is also worth noting that Ser127 is a major inhibitory phosphorylation site, while YAP1 activity is regulated by additional phosphorylation events. Accordingly, an increase in Ser127 phosphorylation may not fully abrogate YAP1 function [57, 58]. This may, at least in part, explain the coexistence of increased Ser127 phosphorylation with apparent YAP1 activation in our results. In addition, the present analyses were performed at a single time point. Changes in total YAP1 abundance and its phosphorylation are dynamic processes, and transient co-increases of total YAP1 and Ser127-phosphorylated YAP1 at the measured time point could be plausible. Our results suggest that CD248 may regulate YAP1 activity through protein complex formation, while the upstream signaling cascade driving YAP1 nuclear translocation remains to be further explored. Furthermore, we discovered that periostin secreted by CD248⁺ CAFs induces collagen I deposition, increases ECM stiffness and further enhances NSCLC cell invasion and migration, which creates a prometastatic feedback loop: CD248⁺ CAFs secrete periostin, which remodels the ECM to create a stiffer, more invasive-friendly microenvironment while simultaneously activating EMT-like phenotypic changes in tumor cells to increase their metastatic potential. Together, these findings establish a novel pathway, CD248–YAP1-periostin-ITGB1, that drives NSCLC metastasis, addressing a critical gap in our understanding of CAFs-mediated metastatic progression.

Despite these advances, several limitations merit consideration. First, our study primarily employed established cell lines and murine xenograft models, which may not fully recapitulate the genetic and phenotypic complexity of human NSCLC. In addition, the FSP1-Cre transgenic mouse model is not fibroblast-specific and possesses inherent limitations. Future studies utilizing patient-derived organoids and xenografts could provide greater clinical relevance and better reflect tumor heterogeneity. Second, although we focused on ITGB1 as the key periostin receptor, other integrin subtypes (e.g., α_vβ₃ or α_vβ₅) may contribute to periostin-mediated effects in NSCLC, and their potential roles warrant systematic investigation. Additionally, while our work characterized the role of periostin in local invasion and migration, its involvement in later metastatic stages—including intravasation, extravasation, and organ-specific colonization—remains to be elucidated and represents an important direction for future research.

Conclusions

In summary, our study elucidates a previously uncharacterized regulatory axis whereby CD248⁺ CAFs orchestrate NSCLC metastasis. Specifically, we demonstrate that CD248 signaling activates YAP1 in CAFs, thereby increasing periostin secretion. Secreted periostin subsequently affects ITGB1 in NSCLC cells, triggering the FAK/Src signaling cascade and promoting EMT-like phenotypic changes. Concurrently, CAFs deposit increased quantities of collagen I, which augments ECM stiffness. This altered mechanical microenvironment establishes a feedforward loop in which collagen I binding to CD248 on CAFs further increases periostin secretion, thereby perpetuating a prometastatic TME. Our findings underscore that CD248 is a critical nodal regulator of this prometastatic signaling network. By identifying these key molecular nodes—CD248, YAP1, periostin and ITGB1—as potential therapeutic targets, our work provides a rationale for the development of targeted interventions that effectively suppress NSCLC metastasis and disrupt deleterious CAFs-tumor cell crosstalk.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

JHW: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, methodology and writing of the original draft. XYW: Investigation, data curation, project administration and writing of the original draft. ZYY: Data curation, formal analysis and methodology. YZ: Formal analysis, investigation and methodology. MQL: Resources, data curation and project administration. LW: Data curation, formal analysis and investigation. XZ: Data curation, investigation and funding acquisition. LL: Visualization, methodology and funding acquisition. JMR: Investigation and project administration. MLZ: Data curation and formal analysis. ZYT: Data curation and formal analysis. LXL: Data curation, formal analysis and funding acquisition. JX: Data curation and visualization. ZZ: Funding acquisition, conceptualization and investigation. AGY: Supervision and conceptualization. RZ: conceptualization, resources, data curation, supervision, investigation and methodology. All the authors have read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82460533, 12132006, 32371373), the Guizhou Provincial Basic Research Program (Natural Science) (qkhjc-zk[2025]ms549, qkhjc-zk[2025]ms470), the Guizhou Provincial Science and Technology Plan Project (qkhjcQN[2025]138, qkhjcQN[2025]140), the Science and Technology Fund Project of Guizhou Provincial Health Commission (gzwkj2025-623, gzwkj2026-143), Guizhou Province Young Science and Technology Talent Nurturing Project Funding, and the National Natural Science Foundation of China Cultivation Program of The Affiliated Hospital of Guizhou Medical University (gyfynsfc [2024]-17).

Data availability

The data are available upon request from the authors.

Declarations

Ethics approval and consent to participate

Animal studies were conducted in accordance with the Animal Use Protocol approved by Guizhou Medical University. The Affiliated Hospital of Guizhou Medical University provided the human NSCLC and matched healthy tissue biopsies for comparison. All participants provided written informed consent before the research began, and the study was authorized by the ethics committee of the same institution. The maximal tumor size/burden permitted by the Ethics Committee was 1.5 cm in diameter; the maximal tumor size/burden was not exceeded in the experiments conducted in this study.

Consent for publication

Not applicable.

Conflict of interest

All the authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.