Expression of CXCL10 and CXCL16 in Triple-Negative Breast Cancer and Their Association with Immunotherapy Efficacy and Prognosis

Abstract

Background

Triple-negative breast cancer (TNBC) represents the most aggressive breast cancer subtype with limited therapeutic options. This study investigated the expression patterns of chemokines CXCL10 and CXCL16 in TNBC and their association with immunotherapy response and clinical outcomes.

Methods

We retrospectively analyzed 128 TNBC patients treated at our institution between 2019–2023. CXCL10 and CXCL16 expressions were assessed by immunohistochemistry. Sixty-eight patients received immune checkpoint inhibitor (ICI)-based therapy. The primary endpoints were objective response rate (ORR), progression-free survival (PFS), and overall survival (OS).

Results

High CXCL10 expression (≥50% positive cells) was observed in 72 patients (56.25%), while high CXCL16 expression was found in 65 patients (50.78%). In the immunotherapy cohort, patients with high CXCL10 expression demonstrated significantly higher ORR (44.44% vs 18.75%, p=0.024) and longer median PFS (8.62 vs 4.35 months, HR=0.467, 95% CI: 0.268–0.814, p=0.007). Combined high expression of both chemokines showed the most favorable outcomes, with ORR of 52.63% and median PFS of 10.84 months. Multivariate Cox regression confirmed CXCL10 (HR=0.512, p=0.008) and CXCL16 (HR=0.623, p=0.032) as independent prognostic factors for PFS.

Conclusion

CXCL10 and CXCL16 expression levels are associated with improved immunotherapy response and favorable prognosis in TNBC and may serve as candidate predictive biomarkers. Combined assessment of both chemokines may optimize patient selection for ICI-based therapy. Prospective validation in larger cohorts is warranted.

Introduction

Triple-negative breast cancer (TNBC), characterized by the absence of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression, accounts for approximately 15–20% of all breast cancers and represents the most aggressive subtype with the poorest prognosis.^1,2 The lack of targetable receptors has historically limited treatment options to conventional chemotherapy, necessitating the urgent identification of novel therapeutic strategies and predictive biomarkers.³

The emergence of immune checkpoint inhibitors (ICIs) has revolutionized the treatment landscape for TNBC.⁴ The KEYNOTE-355 trial established pembrolizumab combined with chemotherapy as a standard first-line treatment for PD-L1-positive advanced TNBC, demonstrating improved progression-free survival compared to chemotherapy alone.^5,6 However, only a subset of patients derives durable benefit from immunotherapy, highlighting the critical need for refined biomarker-driven patient selection strategies beyond PD-L1 expression alone.⁷

Chemokines, small chemotactic cytokines that orchestrate immune cell trafficking and activation, have emerged as key regulators of the tumor immune microenvironment.⁸ Among these, C-X-C motif chemokine ligand 10 (CXCL10), also known as interferon-γ-induced protein 10 (IP-10), plays a pivotal role in recruiting CXCR3+ effector T cells to tumor sites.^9,10 Recent evidence suggests that CXCL10 expression correlates with enhanced responses to checkpoint blockade therapy, particularly in homologous recombination-deficient tumors.¹¹ Similarly, CXCL16, a transmembrane chemokine that binds to CXCR6, facilitates the recruitment of natural killer (NK) cells and cytotoxic T lymphocytes, creating an immunologically favorable microenvironment.^12,13

Despite accumulating preclinical evidence supporting the immunomodulatory roles of CXCL10 and CXCL16, their clinical significance as predictive biomarkers for immunotherapy response in TNBC remains inadequately characterized. Previous studies have been limited by small sample sizes, heterogeneous treatment regimens, and lack of standardized assessment methods.^14,15 Furthermore, the potential synergistic predictive value of combined CXCL10 and CXCL16 expression has not been systematically evaluated in the clinical setting.

This study aimed to comprehensively investigate the expression patterns of CXCL10 and CXCL16 in a well-characterized cohort of TNBC patients and evaluate their association with immunotherapy efficacy and long-term clinical outcomes. We hypothesized that high expression of these chemokines would correlate with improved responses to ICI-based therapy and favorable prognosis, potentially serving as complementary biomarkers to optimize patient selection for immunotherapy.

Materials and Methods

Study Design and Patient Population

This retrospective cohort study was conducted at Xuzhou Cancer Hospital with approval from the institutional review board (IRB No. XCH-2023-087). We identified 128 patients with histologically confirmed TNBC diagnosed between January 2019 and December 2023. Inclusion criteria were: (1) triple-negative phenotype confirmed by immunohistochemistry (ER <1%, PR <1%, HER2 0–1+ or 2+ with negative FISH); (2) available tumor tissue for chemokine expression analysis; (3) complete clinical follow-up data; and (4) no prior malignancies within 5 years. Exclusion criteria included: (1) bilateral breast cancer; (2) inflammatory breast cancer; and (3) insufficient tissue sample quality.

Immunohistochemistry Analysis

Formalin-fixed paraffin-embedded (FFPE) tumor samples were sectioned at 4 μm thickness. CXCL10 and CXCL16 expression was assessed using validated antibodies (anti-CXCL10, clone 6D4, dilution 1:200, Abcam; anti-CXCL16, polyclonal antibody, dilution 1:150, R&D Systems). Staining was performed on an automated immunostainer (Ventana BenchMark ULTRA) following standard protocols. PD-L1 expression was assessed using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies, Santa Clara, CA, USA) on an automated staining platform according to the manufacturer’s instructions. PD-L1 status was quantified using the combined positive score (CPS), defined as the number of PD-L1-staining cells (tumor cells, lymphocytes, and macrophages) divided by the total number of viable tumor cells, multiplied by 100; CPS ≥10 was considered PD-L1 positive.

Expression levels were evaluated by two independent pathologists blinded to clinical outcomes. The percentage of positive tumor cells and staining intensity (0=negative, 1=weak, 2=moderate, 3=strong) were recorded. An H-score was calculated as: Σ(percentage of cells × intensity score). High expression was defined as ≥50% positive cells with moderate-to-strong intensity (H-score ≥100), based on receiver operating characteristic (ROC) curve analysis for optimal cutoff determination.

Assessment of Tumor-Infiltrating Immune Cells

Stromal tumor-infiltrating lymphocytes (TILs) were evaluated on hematoxylin and eosin-stained sections according to the recommendations of the International Immuno-Oncology Biomarker Working Group and were reported as the percentage of tumor stromal area occupied by mononuclear inflammatory cells. CD8+ T cells and natural killer (NK) cells were detected by IHC using anti-CD8 (clone C8/144B, dilution 1:100, Dako, Agilent Technologies) and anti-CD56 (clone 123C3, dilution 1:50, Dako, Agilent Technologies) antibodies. Positively stained cells were counted in five representative high-power fields (×400) within tumor-rich stromal areas, and the mean density was recorded as cells/mm². All assessments were performed independently by two experienced pathologists blinded to clinical outcomes, with discrepancies resolved by consensus review.

Treatment Regimens

Among the 128 patients, 68 (53.13%) received ICI-based therapy, including: (1) pembrolizumab plus chemotherapy (n=42, 61.76%); (2) atezolizumab plus nab-paclitaxel (n=18, 26.47%); and (3) durvalumab plus chemotherapy (n=8, 11.77%). The remaining 60 patients received conventional chemotherapy alone. Treatment allocation was based on clinical guidelines, PD-L1 expression status, and physician discretion. ICI-based regimens were administered as first-line or later-line therapy for unresectable locally advanced or metastatic disease, or as (neo)adjuvant therapy for early-stage disease, according to contemporary guidelines. Tumor tissue used for IHC analysis was obtained from core-needle biopsy or surgical specimens collected before initiation of the systemic therapy under evaluation. Tumor response was assessed according to RECIST v1.1 criteria in patients with at least one measurable lesion at baseline; for patients with early-stage disease, the primary tumor and/or regional nodal lesions served as measurable target lesions before (neo)adjuvant treatment. Radiologic assessment by computed tomography or magnetic resonance imaging was performed every 6–8 weeks during treatment.

Statistical Analysis

Categorical variables were compared using chi-square or Fisher’s exact tests. Continuous variables were analyzed using Student’s t-test or Mann–Whitney U-test as appropriate. The primary efficacy endpoints were objective response rate (ORR) by RECIST v1.1 criteria and progression-free survival (PFS). Secondary endpoints included overall survival (OS) and disease control rate (DCR). Survival analyses were performed using the Kaplan-Meier method with Log rank tests. Hazard ratios (HR) with 95% confidence intervals (CI) were estimated using Cox proportional hazards models. Multivariate analysis included variables with p<0.10 in univariate analysis. The proportional hazards assumption of Cox models was verified using Schoenfeld residuals, and internal validation of the multivariate model was performed using bootstrap resampling with 1000 iterations. The synergistic predictive value of combined CXCL10/CXCL16 expression was assessed using interaction terms in Cox models. To distinguish predictive from prognostic effects, biomarker × treatment interaction terms were tested in Cox models fitted in the entire cohort, and inverse probability of treatment weighting (IPTW) based on propensity scores (estimated from age, stage, grade, Ki-67 index, and PD-L1 CPS) was applied in sensitivity analyses. All statistical tests were two-sided with significance set at p<0.05. Analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY) and R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria),¹⁶ together with the survival¹⁷ and survminer¹⁸ packages.

Results

Baseline Clinical and Pathological Characteristics of the 128 TNBC Patients

The study cohort comprised 128 TNBC patients with a median age of 52.45 years (range: 28–75). Clinical characteristics are summarized in Table 1. The majority presented with stage II (n=48, 37.50%) or stage III (n=56, 43.75%) disease. Histologically, invasive ductal carcinoma was predominant (n=108, 84.38%), with high-grade tumors (grade 3) in 92 patients (71.88%).

Table 1.

Baseline Patient Characteristics (N=128)

Characteristic	Overall (n=128)	High CXCL10 (n=72)	Low CXCL10 (n=56)	p
Age	52.45 (28–75)	51.82 (28–72)	53.26 (31–75)	0.428
Stage, n (%)				0.236
I	12 (9.38)	8 (11.11)	4 (7.14)
II	48 (37.50)	30 (41.67)	18 (32.14)
III	56 (43.75)	28 (38.89)	28 (50.00)
IV	12 (9.38)	6 (8.33)	6 (10.71)
Grade, n (%)				0.042
2	36 (28.13)	16 (22.22)	20 (35.71)
3	92 (71.88)	56 (77.78)	36 (64.29)
Ki-67 ≥30%, n (%)	85 (66.41)	52 (72.22)	33 (58.93)	0.115
PD-L1 CPS ≥10, n (%)	42 (32.81)	31 (43.06)	11 (19.64)	0.005

Expression Patterns of CXCL10 and CXCL16 and Their Associations with Clinicopathological Features

High CXCL10 expression (H-score ≥100) was observed in 72 patients (56.25%), with a median H-score of 125.68 (range: 0–280). CXCL10 expression showed significant associations with tumor grade (p=0.042) and PD-L1 combined positive score (CPS) ≥10 (p=0.005). High CXCL16 expression was detected in 65 patients (50.78%), with a median H-score of 98.42 (range: 0–265). Baseline characteristics were well balanced between the CXCL16-high and CXCL16-low groups (all p>0.05; Supplementary Table S1).

Notably, 38 patients (29.69%) exhibited high expression of both chemokines, while 29 patients (22.66%) showed low expression of both. The correlation between CXCL10 and CXCL16 expression was moderate (Spearman’s ρ=0.425, p<0.001), suggesting partially overlapping but distinct regulatory mechanisms. Representative immunohistochemical staining patterns of negative, low, and high expression of CXCL10 and CXCL16 are shown in Figure 1.

Figure 1.

Representative immunohistochemical staining of CXCL10 and CXCL16 in TNBC tissues. (A) CXCL10 negative; (B) CXCL10 low expression; (C) CXCL10 high expression; (D) CXCL16 negative; (E) CXCL16 low expression; (F) CXCL16 high expression. Original magnification ×200; scale bar, 50 μm.

Association of CXCL10 and CXCL16 Expression with Objective Response to ICI-Based Therapy

Among the 68 patients who received ICI-based therapy, those with high CXCL10 expression demonstrated significantly improved treatment outcomes (Table 2). The ORR was 44.44% (16/36) in the CXCL10-high group compared to 18.75% (6/32) in the CXCL10-low group (p=0.024). Similarly, the DCR was superior in CXCL10-high patients (77.78% vs 53.13%, p=0.043).

Table 2.

Immunotherapy Response by Chemokine Expression Status

Response	CXCL10-High (n=36)	CXCL10-Low (n=32)	p	CXCL16-High (n=35)	CXCL16-Low (n=33)	p
ORR, n (%)	16 (44.44)	6 (18.75)	0.024	15 (42.86)	7 (21.21)	0.058
DCR, n (%)	28 (77.78)	17 (53.13)	0.043	26 (74.29)	19 (57.58)	0.146
CR, n (%)	3 (8.33)	0 (0.00)	0.093	3 (8.57)	0 (0.00)	0.084
PR, n (%)	13 (36.11)	6 (18.75)	0.111	12 (34.29)	7 (21.21)	0.23
SD, n (%)	12 (33.33)	11 (34.38)	0.93	11 (31.43)	12 (36.36)	0.667
PD, n (%)	8 (22.22)	15 (46.88)	0.043	9 (25.71)	14 (42.42)	0.146

Patients with combined high expression of both CXCL10 and CXCL16 (n=19 in the immunotherapy cohort) achieved the highest ORR of 52.63% (10/19) and DCR of 84.21% (16/19), suggesting synergistic predictive value.

Progression-Free and Overall Survival According to CXCL10 and CXCL16 Expression

With a median follow-up of 28.62 months (range: 6.12–48.35), survival analyses revealed significant associations between chemokine expression and clinical outcomes. During follow-up, 68 progression-free survival events were recorded in the immunotherapy cohort, and 83 deaths occurred in the entire cohort. In the immunotherapy cohort, patients with high CXCL10 expression had significantly longer median PFS compared to those with low expression (8.62 vs 4.35 months, HR=0.467, 95% CI: 0.268–0.814, p=0.007) (Figure 2).

Figure 2.

Kaplan-Meier curves of progression-free survival (PFS) according to CXCL10 expression in the immunotherapy cohort (n=68). Patients with high CXCL10 expression (H-score ≥100; n=36) had significantly longer median PFS than those with low expression (n=32) (8.62 vs 4.35 months; HR=0.467, 95% CI: 0.268–0.814; log-rank p=0.007). Tick marks indicate censored observations, and the number of patients at risk is shown below the curves.

High CXCL16 expression was also associated with improved PFS (7.84 vs 4.82 months, HR=0.582, 95% CI: 0.341–0.993, p=0.046) (Figure 3). The combination of high CXCL10 and CXCL16 expression conferred the most favorable outcomes, with median PFS of 10.84 months compared to 3.65 months for patients with low expression of both markers (HR=0.328, 95% CI: 0.168–0.641, p=0.001) (Figure 4).

Figure 3.

Kaplan-Meier curves of progression-free survival (PFS) according to CXCL16 expression in the immunotherapy cohort (n=68). High CXCL16 expression (H-score ≥100; n=35) was associated with longer median PFS than low expression (n=33) (7.84 vs 4.82 months; HR=0.582, 95% CI: 0.341–0.993; log-rank p=0.046). Tick marks indicate censored observations, and the number of patients at risk is shown below the curves.

Figure 4.

Kaplan-Meier curves of progression-free survival (PFS) according to combined CXCL10 and CXCL16 expression in the immunotherapy cohort (n=68). Patients with high expression of both chemokines (double-high; n=19) had the most favorable median PFS compared with patients with low expression of both markers (10.84 vs 3.65 months; HR=0.328, 95% CI: 0.168–0.641; log-rank p=0.001). Tick marks indicate censored observations, and the number of patients at risk is shown below the curves.

Overall survival analysis in the entire cohort (n=128) showed that high CXCL10 expression was associated with longer median OS (36.28 vs 25.14 months, HR=0.564, 95% CI: 0.342–0.930, p=0.024) (Figure 5). The OS benefit was more pronounced in patients receiving immunotherapy (median OS not reached vs 18.62 months, HR=0.412, 95% CI: 0.198–0.857, p=0.017). High CXCL16 expression was also associated with prolonged OS in the entire cohort (36.28 vs 28.63 months, HR=0.588, 95% CI: 0.379–0.912, p=0.017) (Figure 6).

Figure 5.

Kaplan-Meier curves of overall survival (OS) according to CXCL10 expression in the entire cohort (n=128). High CXCL10 expression (n=72) was associated with longer median OS than low expression (n=56) (36.28 vs 25.14 months; HR=0.564, 95% CI: 0.342–0.930; log-rank p=0.024). Tick marks indicate censored observations, and the number of patients at risk is shown below the curves.

Figure 6.

Kaplan-Meier curves of overall survival (OS) according to CXCL16 expression in the entire cohort (n=128), stratified into high (H-score ≥100; n=65) and low (n=63) expression groups. High CXCL16 expression was associated with longer median OS than low expression (36.28 vs 28.63 months; HR=0.588, 95% CI: 0.379–0.912; log-rank p=0.017). Tick marks indicate censored observations, and the number of patients at risk is shown below the curves.

Multivariate Cox Regression Analysis Identifying Independent Predictors of PFS

Cox proportional hazards regression analysis was performed to identify independent prognostic factors (Table 3). After adjusting for age, stage, grade, Ki-67 index, and PD-L1 expression, both CXCL10 (HR=0.512, 95% CI: 0.312–0.841, p=0.008) and CXCL16 (HR=0.623, 95% CI: 0.404–0.961, p=0.032) remained independent predictors of improved PFS in the immunotherapy cohort. The model showed modest discrimination (apparent C-index 0.695; optimism-corrected C-index 0.656 on bootstrap internal validation with 1000 resamples), and the proportional hazards assumption was satisfied (global Schoenfeld test p=0.295). Further details of model diagnostics and internal validation are provided in Supplementary Note 1.

Table 3.

Multivariate Cox Regression Analysis for Progression-Free Survival (Immunotherapy Cohort)

Variable	HR	95% CI	p-value
Age (≥50 vs <50 years)	1.124	0.687–1.839	0.642
Stage (III–IV vs I–II)	2.358	1.426–3.902	0.001
Grade (3 vs 2)	1.542	0.894–2.659	0.12
Ki-67 (≥30% vs <30%)	1.368	0.825–2.269	0.225
PD-L1 CPS (≥10 vs <10)	0.685	0.418–1.122	0.134
CXCL10 (high vs low)	0.512	0.312–0.841	0.008
CXCL16 (high vs low)	0.623	0.404–0.961	0.032

Interaction Analysis of Combined CXCL10 and CXCL16 Expression

To evaluate the interaction between CXCL10 and CXCL16 expression, we performed stratified analyses. The predictive value of CXCL10 for immunotherapy response tended to be more pronounced in patients with concurrent high CXCL16 expression, although the interaction term did not reach statistical significance (interaction p=0.095). This synergistic effect was reflected in both ORR (52.63% for double-high vs 6.25% for double-low) and PFS outcomes.

Biomarker-Treatment Interaction and Sensitivity Analyses in the Entire Cohort

To distinguish predictive from prognostic effects, biomarker × treatment interaction analyses were performed in the entire cohort (n=128). Because progression-free survival was not systematically assessed in the chemotherapy-alone cohort, overall survival was used as the endpoint for these analyses. The OS benefit of ICI-based therapy over chemotherapy alone was significant in patients with high CXCL10 expression (HR=0.418, 95% CI: 0.212–0.823, p=0.012) but not in those with low CXCL10 expression (HR=0.913, 95% CI: 0.508–1.642, p=0.761); however, the formal CXCL10 × treatment interaction did not reach statistical significance (interaction HR=0.462, 95% CI: 0.189–1.129, p=0.090). No CXCL16 × treatment interaction was observed (interaction HR=0.815, 95% CI: 0.340–1.954, p=0.646). Sensitivity analyses using inverse probability of treatment weighting (IPTW) to account for treatment-selection bias yielded consistent results (CXCL10 × treatment interaction HR=0.476, p=0.128; CXCL16 × treatment interaction HR=0.609, p=0.303). These findings support a predictive trend for CXCL10 that warrants confirmation in larger, randomized cohorts. Detailed results of the biomarker × treatment interaction and IPTW sensitivity analyses are provided in Supplementary Table S2.

Correlation of CXCL10 and CXCL16 Expression with Tumor Immune Cell Infiltration

Exploratory analysis of 42 patients with available tumor-infiltrating lymphocyte (TIL) data revealed that high CXCL10 expression correlated with increased stromal TILs (median 45.25% vs 22.18%, p<0.001) and CD8+ T cell density (median 125.42 vs 58.36 cells/mm², p=0.002). Similarly, CXCL16 expression was associated with NK cell infiltration (median 32.15 vs 14.82 cells/mm², p=0.008). These correlations are summarized in Table 4.

Table 4.

Correlation Between Chemokine Expression and Tumor Immune Infiltration (n=42)

Immune Parameter	Stratification Marker	High Expression	Low Expression	p-value
Stromal TILs, median %	CXCL10	45.25	22.18	<0.001
CD8+ T-cell density, median cells/mm2	CXCL10	125.42	58.36	0.002
NK cell density, median cells/mm2	CXCL16	32.15	14.82	0.008

Discussion

This study provides compelling clinical evidence that CXCL10 and CXCL16 expression levels serve as prognostic biomarkers with potential predictive value for immunotherapy response in TNBC patients. Our findings demonstrate that high expression of these chemokines, particularly in combination, identifies a subset of patients with significantly improved treatment outcomes, offering a potential strategy to refine patient selection beyond current PD-L1-based approaches.

The observed association between CXCL10 expression and immunotherapy efficacy aligns with emerging mechanistic understanding of its role in anti-tumor immunity. CXCL10, through its interaction with CXCR3, orchestrates the recruitment of effector T cells to tumor sites, a fundamental process to checkpoint inhibitor efficacy.^9,19 Our results showing 44.44% ORR in CXCL10-high patients compared to 18.75% in CXCL10-low patients are consistent with preclinical studies demonstrating that CXCL10 potentiates immune checkpoint blockade, particularly in homologous recombination-deficient tumors.¹¹

The prognostic significance of CXCL16 in our cohort extends previous observations from smaller studies. The ELK3-CXCL16 regulatory axis has been shown to modulate NK cell cytotoxicity in TNBC,¹³ and our finding that high CXCL16 expression correlates with improved PFS (HR=0.623, p=0.032) supports its functional importance. Notably, CXCL16 expression showed stronger associations with NK cell infiltration than with T cell populations, suggesting distinct but complementary immune activation mechanisms compared to CXCL10.

The synergistic predictive value of combined CXCL10/CXCL16 assessment represents a novel finding with immediate translational implications. Patients with high expression of both chemokines achieved an ORR of 52.63% and median PFS of 10.84 months, outcomes that compare favorably to current biomarker-selected populations in pivotal trials.^5,6 This suggests that multi-parametric chemokine profiling may identify “immune-hot” tumors most likely to benefit from checkpoint inhibition.

Our multivariate analysis confirmed that CXCL10 and CXCL16 provide prognostic information independent of established factors including PD-L1 expression. This is particularly relevant given the limitations of PD-L1 as a sole biomarker, where many PD-L1-negative patients still derive benefit from immunotherapy while some PD-L1-positive patients show primary resistance.^20,21 The ongoing identification of immune targets beyond the PD-1/PD-L1 axis, including emerging checkpoints such as TIM-3, LAG-3, and TIGIT and signaling pathways such as JAK/STAT, MAPK, and PI3K-AKT-mTOR, highlights the need for comprehensive biomarker screening strategies incorporating chemokine profiling.²² The moderate correlation between CXCL10 and CXCL16 expression (ρ=0.425) suggests they capture partially distinct aspects of immune activation, supporting their combined assessment.

The association between chemokine expression and immune infiltration patterns provides mechanistic validation for our clinical observations. High CXCL10 expression correlated with increased stromal TILs and CD8+ T cell density, consistent with its role in T cell trafficking.²³ The distinct association of CXCL16 with NK cell infiltration highlights the importance of innate immunity in immunotherapy responses, an aspect often overlooked in T cell-centric biomarker strategies.²⁴ Conversely, immunosuppressive molecules within the tumor microenvironment may counteract chemokine-driven immune recruitment. For example, ENPP1, an ectonucleotidase that hydrolyzes extracellular cGAMP and attenuates STING-dependent production of T cell-attracting chemokines such as CXCL10, has been shown to promote immune suppression, drug resistance, and adverse outcomes, and represents a druggable target whose inhibition may restore chemokine-mediated antitumor immunity.²⁵

Several limitations warrant consideration. The retrospective design introduces potential selection bias, though consecutive patient inclusion and standardized treatment protocols mitigate this concern. The relatively modest sample size, particularly for subgroup analyses, necessitates validation in larger cohorts. First, the IHC cutoff values were derived from ROC curve analysis within the same cohort subsequently used for survival testing, which may introduce optimism bias; validation with prespecified cutoffs in independent TNBC immunotherapy cohorts is therefore essential. Second, the immunotherapy cohort comprised only 68 patients, with only 19 patients in the double-high subgroup, and the multivariate model included seven variables relative to a limited number of PFS events; although the proportional hazards assumption was verified and bootstrap internal validation was performed, the risk of model overfitting remains. Third, although biomarker × treatment interaction analyses were performed across the entire cohort with inverse probability of treatment weighting applied as sensitivity analyses, the CXCL10 × treatment interaction showed only a nonsignificant trend (p=0.090), and these analyses were underpowered given the limited number of events; definitive distinction between predictive and prognostic effects therefore requires larger, randomized cohorts. Accordingly, we have tempered our conclusions to describe CXCL10 and CXCL16 as prognostic markers with potential predictive value. Finally, germline mutational background (eg, BRCA1/2 and other homologous recombination-related genes), which is highly prevalent in TNBC and can substantially influence both immunotherapy sensitivity and prognosis, was not assessed in this study and represents an important potential confounder.²⁶ Additionally, the heterogeneity in immunotherapy regimens, while reflecting real-world practice, may influence the generalizability of specific response rates.

The clinical implementation of CXCL10/CXCL16 assessment faces practical considerations. Unlike PD-L1, standardized scoring systems for these chemokines remain to be established. Our H-score methodology, while reproducible, requires validation across institutions. The optimal cutoff values may vary depending on antibody clones, staining protocols, and patient populations, highlighting the need for harmonization efforts similar to those undertaken for PD-L1 testing.²⁷

Future research directions should include prospective validation in randomized controlled trials, ideally with predefined chemokine expression stratification. The development of multiplex immunofluorescence or spatial transcriptomic approaches could provide more nuanced assessment of chemokine gradients and cellular sources within the tumor microenvironment.^28,29 Additionally, investigating the dynamic changes in chemokine expression during treatment may identify mechanisms of acquired resistance.

The potential for therapeutic targeting of chemokine pathways represents an exciting avenue for combination strategies. Recent preclinical evidence suggests that STAT1 inhibition combined with anti-PD-1 therapy and CXCL16 modulation can overcome resistance mechanisms.¹² Our clinical data showing superior outcomes in chemokine-high patients provide rationale for exploring chemokine pathway enhancement as a therapeutic strategy in chemokine-low tumors.

Integration with emerging biomarkers and technologies offers opportunities for comprehensive immune profiling. Liquid biopsy approaches for circulating chemokine detection could enable non-invasive monitoring and early resistance detection.³⁰ Machine learning algorithms incorporating chemokine expression with genomic, transcriptomic, and clinical features may further refine predictive models.^31,32

Conclusions

This study demonstrates that CXCL10 and CXCL16 are clinically relevant prognostic biomarkers associated with immunotherapy response in TNBC. High expression of these chemokines, particularly in combination, identifies patients with favorable treatment outcomes independent of PD-L1 status. These findings support the incorporation of chemokine assessment into biomarker panels for optimizing patient selection for ICI-based therapy. Prospective validation and standardization efforts are warranted to translate these observations into clinical practice, potentially expanding the population of TNBC patients who benefit from immunotherapy.

Funding Statement

This research was funded by Science and Technology Innovation Project of Xuzhou Municipal Health Commission (XWKYHT20240061).

Data Sharing Statement

All data generated or analyzed in this study are included in the present manuscript, and its Supplementary Material.

Ethics and Informed Consent Statement

This retrospective cohort study was conducted at Xuzhou Cancer Hospital with approval from the institutional review board (IRB No. XCH-2023-087). The study was conducted in accordance with the Declaration of Helsinki. Informed consent was waived for this retrospective study due to the exclusive use of de-identified patient data, which posed no potential harm or impact on patient care.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors have no conflicts of interest to declare.