C-X-C Motif Chemokine Ligand 9 Correlates with Favorable Prognosis in Triple-Negative Breast Cancer by Promoting Immune Cell Infiltration

Triple-negative breast cancer (TNBC) is a breast cancer subtype that manifests as absent expression of estrogen receptor (ER) and progesterone receptor (PR) and no amplification of HER2 (1, 2). TNBC accounts for 10% to 20% of all breast cancers (1, 3, 4) and often occurs in young women. At diagnosis, patients present additional metastatic lymph nodes and the tumors tend to be large, at an advanced stage, of poor histologic grade, and harboring aggressive oncogenic characteristics (1, 3, 5). In addition, TNBC has earlier recurrence and metastasis and worse survival outcomes relative to other breast cancer subtypes (1, 3, 6). Due to that there are only a few options, systemic treatment with both adjuvant and neoadjuvant chemotherapy containing anthracycline and paclitaxel agents is still dominated (2), with recent studies also confirming the efficacy of platinum-based drugs as important chemotherapeutic options for TNBC (7).

For patients with early-stage TNBC harboring high-risk BRCA1 or BRCA2 pathogenic mutations, additional administration of olaparib (a PARP inhibitor) reduces the risk of recurrence and improves prognosis after local treatment and neoadjuvant or adjuvant chemotherapy (8). Nevertheless, the prognosis of patients with TNBC remains worse than that of patients with other breast cancer subtypes. Recent progress in immunotherapy includes FDA approval of inhibitors of programmed cell death ligand 1 (PD-L1) and programmed cell death (PD-1) used in combination with chemotherapy as a rescue treatment for metastatic TNBC (9, 10). In addition, the FDA approved combined use of a PD-1 inhibitor with chemotherapy as a neoadjuvant and enhanced-treatment strategy for high-risk early TNBC (11). However, the response rate of patients with TNBC to immunotherapy remains limited owing to the plasticity of tumor cells (TC) and the tumor microenvironment (TME).

Chemokines are small proteins (8–15 kDa) that can induce chemotaxis and tissue exosmosis and recruit leukocytes into tissues. Chemokines are subdivided into four groups according to structural motifs (C-X-C, C-C, C-X3-C, and C) that describe the number and spacing of two conserved N-terminal cysteine residues (12, 13). C-X-C chemokine ligand 9 (CXCL9) expression is highly induced by IFNγ, after which CXCL9 binds C-X-C chemokine motif receptor 3 (CXCR3) to regulate tumor angiogenesis, promote the directional migration of memory cells, and activate T cells in the TME (14–16). Recent evidence confirmed the important synergistic effect of the CXCL9/CXCR3 axis on anti–PD(L)-1 therapeutic responses (17–19).

Previous studies show that CXCL9 is highly expressed in the serum of patients with breast cancer (20, 21) and tumor tissues relative to levels identified in benign disease tissue or healthy controls (20, 22, 23). In addition, studies using public datasets demonstrate associations between CXCL9 mRNA levels and improved survival outcomes in patients with breast cancer (22), especially in ER-negative (23) and TNBC subgroups (24). Moreover, reports from an early breast cancer cohort indicated high CXCL9 expression as a favorable prognostic factor for both disease-free survival (DFS) and overall survival (OS) in a TNBC subgroup (25). Furthermore, analyses revealed that elevated CXCL9 levels were predictive of a better response to chemotherapy in TNBC (24). However, there are limited studies of CXCL9 expression in TNBC.

In this study, we investigated CXCL9 expression and its relationship with the clinicopathologic characteristics and prognosis of TNBC. Specifically, we analyzed correlations between the expression patterns of CXCL9 and PD-L1 and the levels and types of infiltrating immune cells (IC) to further evaluate the potential role of CXCL9 in antitumor immunity in TNBC.

Breast cancer tissues were collected from 268 untreated patients with TNBC (stages I–III) diagnosed at Peking Union Medical College Hospital (PUMCH) between July 2010 and June 2014. The inclusion criteria were untreated stage I–III TNBC without previous exposure to chemotherapy or radiotherapy. The exclusion criteria were de novo stage IV cancer, inflammatory breast cancer, history of other malignancies, and incomplete medical records. All included patients underwent radical surgery, followed by standard adjuvant therapy (including chemotherapy and radiotherapy) and regular follow-up. This study was approved by the Ethical Review Board of the PUMCH, and written informed consent was obtained from all patients participating in this study. A tissue microarray (TMA) was constructed by using a hollow needle to select areas (1-mm diameter) with sufficient tumor epithelial cells and stroma from formalin-fixed, paraffin-embedded breast cancer tissue after hematoxylin and eosin (H&E) staining. The tissue cores were then embedded in a paraffin block to construct a TMA.

TMA slides were deparaffinized and rehydrated. After blocking endogenous peroxidase activity with hydrogen peroxide, the slides were incubated with primary antibodies against CXCL9 (AB-9720; Abcam, Cambridge, UK), PD-L1 (SP28–8; Dako Agilent Technologies, Santa Clara, CA), CD3 (PA0553; Leica Biosystems, Wetzlar, Germany), CD4 (PA0427; Leica Biosystems), CD8 (PA0183; Leica Biosystems), and CD19 (ZA-0569; Leica Biosystems) at 4°C overnight. The slides were then incubated with secondary antibodies for 30 min at room temperature. Two independent pathologists evaluated the IHC staining results and scored each sample, with >90% of the evaluation results consistent between pathologists. In cases of inconsistent results, resolution was achieved through a joint reevaluation of specific tumor regions.

CXCL9 expression was evaluated in both TCs and ICs. In the TME, infiltrating ICs include lymphocytes, myeloid cells, dendritic cells (DC), and macrophages. We divided the PUMCH cohort according to the expression percentage of CXCL9 in ICs (cutoff: 1%) into IC-positive (≥1%) and IC-negative (<1%) groups. In addition, we divided the cohort into TC-positive (>0%) and TC-negative (0%) groups according to the expression percentage of CXCL9 in TCs. Similarly, evaluation of PD-L1 expression in both TCs and ICs resulted in groups divided according to expression percentage (cutoff: 5%): IC-positive (≥5%) and IC-negative (<5%). The percentages of CD3, CD4, CD8, and CD19 expressed in ICs were investigated as continuous values.

RNA sequencing data (level 3) of the Basal-like subtype of tissue samples from a breast-invasive cancer cohort (n = 140) and the corresponding clinical features (n = 138) were obtained from The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov) and integrated using R software (v4.0.3; https://www.r-project.org/). We used immuneeconv (https://github.com/omnideconv/immunedeconv), an R software package containing xCell algorithms (26), to assess the results of kinds of immune score evaluations for the TME in the Basal-like subtype of TCGA-BRCA cohort.

Categorical variables were compared using the χ² test, and Mann–Whitney U tests were used to analyze the significance of differences between groups. Survival analyses, including DFS and OS, were plotted using the Kaplan–Meier method, and comparisons were performed using log-rank tests. Univariate and multivariate analyses were performed using Cox regression models to determine independent prognostic factors in the cohorts. A two-sided P < 0.05 was considered statistically significant. Statistical analyses were performed using R software (v4.0.3), GraphPad Prism software (v9.0.2; GraphPad Software, San Diego, CA), and SPSS (v21.0; IBM Corp., Armonk, NY), with GraphPad also used for plotting.

The publicly available datasets analyzed in this study are available in TCGA database (http://portal.gdc.cancer.gov/). The original data presented in this study are available from the corresponding author upon reasonable request.

On the basis of the inclusion and exclusion criteria, 268 patients were enrolled and analyzed in the PUMCH cohort. The median age of the patients was 49 years (range: 23–77 years), with 124 (46.3%) aged >50 years. In addition, of the included patients, 138 (51.5%) presented with T2–4 tumors, 115 (42.9%) with metastatic lymph nodes, and 57 (21.3%) with at least four lymph node metastases. According to the American Joint Committee on Cancer, 91 (34%), 112 (41.8%), and 65 (24.2%) patients presented with tumors at stages I, II, and III, respectively. In terms of treatment, 188 (70.1%) patients received chemotherapy containing taxane, anthracycline, or capecitabine (combined or sequential), and 92 (34.3%) received radiotherapy. Overall, 192 (71.6%) and 20 (7.5%) patients demonstrated low differentiation and lymphovascular invasion, respectively, with a median mitotic index of 50% (Table 1).

CXCL9 was expressed in 80.6% (216) of the whole cohort and was mainly expressed in ICs rather than in TCs (Fig. 1A–D). In total, 7.5% (20) of the patients showed CXCL9 expression in both ICs and TCs, while 19.4% (52) showed no expression in both. Overall, 71.3% (196) of the patients showed CXCL9 expression only in ICs, whereas none of the patients showed expression only in TCs. Furthermore, 80.6% (216) of the patients had CXCL9 expression in ICs (> 0%), and 40.7% (109) had percentage of CXCL9 expressed in ICs greater than or equal to 1% (≥1%), with a median percent positivity of 1% and range from 0% to 25% (Supplementary Fig. S1A). We divided the PUMCH cohort into “IC-positive” (≥1%) and “IC-negative” (<1%) groups according to the expression level of CXCL9 in ICs. Only 20 of the patients had CXCL9 expression in TCs, with a median percent positivity of 0% and range from 0% to 10% (Supplementary Fig. S1B), and we divided the cohort into “TC-positive” (>0%) and “TC-negative” (=0%) groups based on the expression status of CXCL9 in TCs.

The CXCL9-IC–positive group strongly correlated with fewer than four lymph node metastases (P = 0.024). The CXCL9-IC–positive group did not correlate with age (P = 0.175), tumor stage (P = 0.168), tumor-node-metastasis (TNM) stage (P = 0.523), lymphovascular invasion (P = 0.313), chemotherapy (P = 0.612), or radiotherapy (P = 0.846). There was no correlation between the CXCL9 TC status and any clinicopathologic features (Table 2).

The CXCL9 expression in TCs (P = 0.0011; Fig. 2A) and PD-L1 expression in ICs (P < 0.0001; Fig. 2C) were significantly higher in the CXCL9-IC–positive group compared with the negative group, while with no difference of PD-L1 expression in TCs between the CXCL9-IC–positive and CXCL9-IC–negative groups (P = 0.710; Fig. 2B). PD-L1 was expressed in both ICs and TCs (Fig. 1E and F). The percentage of CD3, CD4, CD8 and CD19 expression in ICs were also detected (Fig. 1G–J). In addition, we identified significantly higher numbers of tumor-infiltrating lymphocytes (TIL; P < 0.0001), CD3⁺ TILs (P < 0.0001), CD4⁺ TILs (P < 0.0001), CD8⁺ TILs (P < 0.0001), and CD19⁺ TILs (P < 0.0001) in the CXCL9-IC–positive group than in the CXCL9-IC–negative group (Fig. 2D–H).

Spearman correlation analyses (Supplementary Table S1) showed that CXCL9 expression was strongly and positively correlated with PD-L1 expression in ICs (r = 0.530; P < 0.001) and the numbers of TILs (r = 0.562; P < 0.001), CD3⁺ TILs (r = 0.596; P < 0.001), CD4⁺ TILs (r = 0.551; P < 0.001), CD8⁺ TILs (r = 0.599; P < 0.001), and CD19⁺ TILs (r = 0.573; P < 0.001). Similarly, PD-L1 expression in ICs was positively correlated with the number of TILs (r = 0.478; P < 0.001), CD3⁺ TILs (r = 0.516; P < 0.001), CD4⁺ TILs (r = 0.620; P < 0.001), CD8⁺ TILs (r = 0.464; P < 0.001), and CD19⁺ TILs (r = 0.411; P < 0.001). Correlations among the numbers of CD3⁺, CD4⁺, CD8⁺, and CD19⁺ TILs were positive (r > 0.4; P < 0.001).

The CXCL9-IC–positive group demonstrated a significant survival advantage in both DFS (P = 0.038) and OS (P = 0.023) relative to the CXCL9-IC–negative group in the PUMCH cohort (Fig. 2I and J). On the basis of the strong correlations between CXCL9 and PD-L1 expression in ICs, we further divided the cohort into four subtypes: CXCL9(+)PD-L1(+) (CXCL9 in IC ≥1% and PD-L1 in IC ≥ 5%), CXCL9(+)PD-L1(−), CXCL9(−)PD-L1(+), and CXCL9(−)PD-L1(−). Survival analyses revealed significant differences in OS among the four subtypes, especially the CXCL9(+)PD-L1(+) group showed significantly advantage than the CXCL9(−)PD-L1(−) group (P = 0.016; Fig. 2L), whereas there was no significant difference in DFS (Fig. 2K).

Univariate Cox regression analysis showed that CXCL9-IC–positivity could predict better OS outcomes with a HR of 0.477 [95% confidence interval (CI), 0.248–0.916; P = 0.026]. However, the multivariate Cox regression analysis could not independently predict survival outcomes (HR = 0.518; 95% CI, 0.263–1.021; P = 0.057) as compared with age, TNM stage, mitotic index, histologic grade, and lymphovascular invasion. In addition, subtype with both CXCL9 and PD-L1 positive expression were independent predictive factors for better OS compared with other subtypes according to both univariate (HR = 0.688; 95% CI, 0.541–0.876; P = 0.002) and multivariate (HR = 0.700; 95% CI, 0.547–0.897; P = 0.005) Cox regression analyses (Table 3).

Using the median CXCL9 mRNA expression as the cutoff, we divided the Basal-like BRCA patient cohort from TCGA into high- and low-CXCL9 groups. The high-CXCL9 groups were strongly associated with fewer than four lymph node metastases (P = 0.016), which was similar to the result in the PUMCH cohort. In addition, CXCL9 expression correlated with early TNM stage (P = 0.041) but not with age (P = 0.856), race (P = 0.529), or tumor stage (P = 0.239), which was consistent with findings in the PUMCH cohort (Supplementary Table S2).

Survival analysis showed no significant difference between high- and low-CXCL9 groups in DFS (P = 0.269) or OS (P = 0.285; Supplementary Fig. S2). Both univariate (HR = 0.598; 95% CI, 0.231–1.549; P = 0.290) and multivariate (HR = 0.994; 95% CI, 0.347–2.843; P = 0.991) Cox regression analyses showed that CXCL9 mRNA expression could not predict OS in the TCGA-BRCA cohort (Supplementary Table S3).

Spearman's correlation analysis showed that CXCL9 mRNA expression strongly and positively correlated with the immune score (r = 0.771; P < 0.001) in the Basal-like TCGA-BRCA cohort (Fig. 3A), with immune scores significantly higher in the high-CXCL9 group as compared with those in the low-CXCL9 group (P < 0.001; Fig. 3B). Furthermore, we identified significantly higher numbers of CD4⁺ naïve, CD4⁺ memory, and CD4⁺ effector memory T cells; CD4⁺ T helper 2 cells; CD8⁺, CD8⁺ naïve, CD8⁺ central memory, and CD8⁺ effector memory T cells; naïve and memory B cells; M1 and M2 macrophages; monocytes; and natural killer (NK) cells in the high-CXCL9 group as compared with those in the low-CXCL9 group (all P < 0.001; Fig. 3C; Supplementary Table S4).

Analysis of immune checkpoint molecules between groups indicated significantly higher expression levels of PD-1, PD-L1, PD-L2, CTL-associated protein 4, lymphocyte-activation gene 3 protein, V-set immunoregulatory receptor, hepatitis A virus cellular receptor 2, T cell immunoreceptor with Ig and ITIM domains (all P < 0.001), and sialic acid-binding Ig-like lectin 15 (P = 0.0042) in the high-CXCL9 group relative to that in the low-CXCL9 group (Supplementary Fig. S3A and Supplementary Table S5). In addition, we identified upregulated expression of the CXCL family members CXCL10 (P < 0.0001) and CXCL11 (P < 0.0001) in the high-CXCL9 group relative to levels in the high-CXCL9 group (Supplementary Fig. S3B; Supplementary Table S6).

CXCL9 plays an important role in antitumor immunity by binding CXCR3 and regulating the activation, differentiation, and migration of ICs, particularly CTLs, NK cells, and macrophages (14–16, 27, 28). CXCL9 is secreted by ICs, such as T lymphocytes, NK cells, DCs, and macrophages, and non-ICs, such as TCs, endothelial cells, and fibroblasts (29). In the current study, we identified CXCL9 expression in TNBC TCs and/or infiltrating ICs, with ICs showing the most CXCL9 expressions. No more than one tenth of samples showed CXCL9 expression in both TCs and ICs.

Previous studies show that CXCL9 level is upregulated in breast cancer tumors relative to that in healthy controls (20), in TNBC as compared with luminal subtypes (22), and in TNBC as compared with non-TNBC tumors (23). In the current study, we found that more CXCL9-IC–positive patients had fewer than four metastatic lymph nodes, whereas more CXCL9-IC–negative patients had more than four metastatic lymph nodes. This was consistent with a previous report that identified upregulated CXCL9 levels in lymph node-negative breast cancer tumors (21). Evaluation of CXCL9 expression and tumor size revealed no statistically significant difference between groups, although we found that all T4 tumors were CXCL9-IC–negative, which agrees with a previous study that identified upregulated CXCL9 levels in small TNBC tumors as compared with those in tumors sized > 5 cm (24). Moreover, in the current study, we found that CXCL9-IC–positivity was significantly more common in tumors with a poor histologic grade, which agreed with a previous study indicating that CXCL9 was upregulated in high histologic grade early breast cancer tumors as compared with its expression in low-grade tumors (25). Furthermore, correlation analysis of CXCL9 expression with different prognostic factors of TNBC revealed inconsistent results, which might be due to the dual effects of CXCL9 on both tumor suppression and promotion (29).

We confirmed that CXCL9-IC expression was significantly and positively correlated with infiltrating ICs in the PUMCH cohort, including CD3⁺TILs, CD4⁺TILs, CD8⁺TILs, CD19⁺TILs, and total TILs. In addition, group demonstrating high levels of CXCL9 mRNA expression also showed significantly higher infiltration of effective and memory CD4⁺ T cells, CD8⁺ T cells, macrophages, and myeloid-derived suppressor cells as compared with the group showing low levels of CXCL9 mRNA expression in the TCGA-BRCA cohort. Previous studies report that CXCL9 is strongly and positively correlated with TIL infiltration in early breast cancer (25), as well as the infiltration of CD4⁺ T cells, CD8⁺ T cells, DCs, macrophages, and monocytes in all breast cancers, Basal-like subtypes (22), and ER-negative breast cancer (23). These studies suggest that CXCL9 is closely related to the infiltration of a variety of ICs in the TNBC TME and promotes tumor immune functions.

Furthermore, previous studies revealed that CXCL9 expression positively correlates with PD-L1 expression in ER-negative breast cancer (23), TNBC (24), gastric cancer (30), colorectal cancer (31), and stage T1 non–muscle-invasive bladder cancer (32). These findings are consistent with those in the current study. Specifically, studies on gastric (30) and bladder cancers (33) indicated that activation of the CXCL9/CXCR3 axis upregulates PD-L1 expression through the signal transducer and activator of transcription and PI3K–Akt pathways. However, reports concerning ovarian cancer identified inconsistencies in the influence of CXCL9 on the inflammatory immune microenvironment, leading to upregulated PD-L1 expression, although as a result of an indirect effect (34).

CXCL9 expression in ICs was found to be significantly associated with favorable OS in TNBC, which is consistent with previous reports related to breast cancer (22, 35), TNBC (24, 25), colorectal cancer (16, 29), early non—muscle-invasive bladder cancer (32), and ovarian cancer (36). We found that patients demonstrating both elevated CXCL9 and PD-L1 expression also showed the best OS relative to other groups. In the TME, IC-secreted CXCL9 stimulation promotes the infiltration of other ICs into the TME and polarization of activated CD8⁺T cells into CTLs, thereby promoting antitumor activity within the TME. This explains how elevated CXCL9 expression is associated with better prognosis in most solid tumors, including TNBC.

Immune checkpoint blockade (ICB) involving the PD-1/PD-L1 axis is currently a widely applied treatment strategy for solid tumors. ICB combined with chemotherapy constitutes a first-line treatment for advanced TNBC; however, ICB has limited effectiveness owing to its low therapeutic response and inclination toward therapeutic tolerance. A study on non–small cell lung cancer identified a combination of protein markers, including CXCL9, capable of accurately predicting the response of ICB targeting the PD-1/PD-L1 axis (37). CXCL9 is involved in M1 macrophage polarization, which enhances the efficacy of blocking the PD-1/PD-L1 axis in gastric cancer (38). During treatment of a mouse model of ovarian cancer, ICB resistance was effectively reversed following external administration of CXCL9, which promoted the efficacy of anti–PD-L1 treatment (34). Moreover, in a mouse model of CT26 or MC38 colorectal cancer, B16F10 melanoma, and 4T1 breast cancer, CXCL9⁺ tumor-associated macrophages predicted the therapeutic effect of avelumab (anti–PD-L1) therapy (39). Furthermore, recent studies demonstrated that CXCL9, which recruited and activated T cells, had co-localization with PD-1⁺ T cells. Thus, the simultaneous expression of CXCL9 and PD-1 could promote ICB treatment of solid tumors (40). In addition, CXCL9 could potentially be used as a biomarker in TNBC to predict the efficacy of ICB treatment, as patients with high CXCL9 and PD-L1 expression may benefit most from this treatment. For patients with high PD-L1 and low CXCL9 levels, it may be necessary to promote CXCL9 secretion in the TME to enhance the therapeutic effects of ICB therapy.

As reported by literature, CXCL10 and CXCL11, structurally and functionally related to CXCL9, were aimed for to a common primary receptor CXCR3 (12). Previous studies showed that serum CXCL10 level was elevated in patients with breast cancer compared with healthy controls (21, 41, 42), and the same pattern was observed in patients with ER-positive breast cancer compared that of ER-negative tumors (42). Meanwhile, CXCL10 expression were significantly associated with high tumor grade and the presence of peritumoral CD4⁺ and CD8⁺ lymphocytes in patients with breast cancer, a third of which showed BRCA1 or BRCA2 germline mutation (43). Previous researches demonstrated that high level of CXCL10 expression could strengthen the treatment effect of tamoxifen in patients with ER-positive breast cancer, resulting in lower local recurrence risk (44) and better survival in TNBC (42). However, other studies showed that high CXCL10 mRNA expression indicated less favorable DFS and OS in ER-positive breast cancer (45). Previous research has shown that long noncoding RNAs such as circ_0000514, circ_0001667, and hsa_circ_0000515 play a regulatory role in breast cancer progression, proliferation, and angiogenesis through the CXCL10-related axis (46–48). CXCL10 may be responsible for accelerating the transformation from ductal carcinoma in situ to invasive carcinoma of the breast, especially in hormone receptor-negative tumors (49). CXCL11 is another important chemokine that is upregulated in breast cancer tumors relative to healthy controls, especially in TNBC (21). RAMP2-AS1, a long noncoding RNA, can suppress the malignant phenotype of breast cancer by inhibiting CXCL11 (50). High levels of CXCL11 are positively correlated with immune activation, IC infiltration, and upregulation of immune checkpoint expression. CXCL11 plays an important role in antitumor immunity in breast cancer (51). Previous studies have shown that a risk signature based on the hub-genes CXCL9, CXCL10, and CXCL11 can predict survival outcome in breast cancer. This risk signature is also strongly associated with the infiltration of ICs such as CD4⁺ and CD8⁺ T cells (52). In the future, we will conduct analyses of CXCL9, CXCL10, and CXCL11 in both protein and mRNA levels, aiming to investigate the correlation between these combined indicators and survival outcomes, as well as their effects on immune infiltration in TNBC.

This study enrolled a PUMCH cohort and the TCGA-BRCA cohort for the analysis of CXCL9 expression in TNBC. The PUMCH cohort is homogeneous, which including untreated stage I–III patients who received curative surgery and adjuvant chemotherapy and radiotherapy if needed in a single center. There are some limitations in this research. First, this was a retrospective study, which may have resulted in selection bias. Second, this was a single-center study with a small sample size. Third, we used IHC staining to determine CXCL9 and PD-L1 expression and established cut-off values, suggesting the likelihood that different antibodies and cut-off values may affect the results. Therefore, future studies with a larger sample size are required to validate these findings.

In conclusion, this study revealed that CXCL9 expression in ICs represents a prognostic marker for OS in patients with TNBC with tumors at stages I–III. In addition, CXCL9 expression in the TME was strongly and positively correlated with the numbers of TILs and immune checkpoint molecules, particularly PD-L1. Furthermore, patients demonstrating both high CXCL9 and PD-L1 expression showed significantly better OS as compared with other patients. These results expanded the understanding of CXCL9 function in the TNBC TME and its impact on prognosis. In particular, these findings suggested its close relationship with PD-L1 and potential for use in developing effective ICB strategies for TNBC.

No disclosures were reported.

X. Cao: Conceptualization, formal analysis, writing–original draft, writing–review and editing. Y. Song: Data curation, formal analysis, writing–original draft, writing–review and editing. H. Wu: Resources, data curation, writing–review and editing. X. Ren: Conceptualization, formal analysis, writing–review and editing. Q. Sun: Conceptualization, investigation, methodology, writing–review and editing. Z. Liang: Conceptualization, supervision, methodology.

This study is supported by the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS; 2021–12M-1–053) and the funders are not involved in this study and without any interest connection.

We would like to thank Drs. Longyun Chen and Junyi Pang of Department of Pathology, Peking union medical college hospital, Beijing, China, for their support for this study.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.