Basic characteristics of the study cohort

A total of 252 HER2 + MBC patients were enrolled in this study. The clinical characteristics of the cohort are presented in Table 1. In summary, the median follow-up time of the HER2 + MBC cohort was 29.6 months (range: 3.7–95.0 months), and the median age at diagnosis was 52 years (range: 25–82). At the time of treatment, primary stage IV breast cancer was observed in 67 patients (26.6%), while recurrent breast cancer was present in 185 patients (71.9%). All individuals of the cohort received first-line therapy. Out of the 252 patients, 151 (60.0%) received chemotherapy plus trastuzumab, 49 (19.4%) received chemotherapy in combination with trastuzumab and pertuzumab, 23 (9.1%) received chemotherapy plus tyrosine kinases inhibitors (TKIs) and 29 (11.5%) received chemotherapy alone. In those who received TKIs, all had previously undergone adjuvant and/or neoadjuvant anti-HER2 therapy involving trastuzumab and/or pertuzumab.

Table 1 Clinical characteristics of the study cohort (n = 252).The prognostic value of baseline pTCD8+CD28- level

The 252 patients with HER2 + MBC were randomly allocated into two groups: the training set (n = 126) and the validation set (n = 126). Using X-tile software, we evaluated the prognostic potential of each pL subtype in terms of first-line PFS within the training set. Consequently, only the pTCD8+CD28− level demonstrated a significant association with first-line PFS in HER2 + MBC patients (Supplementary Fig. S3). The optimal cut-off value for pTCD8+CD28- was determined to be 18.0% (ranging from 16.5 to 19.0%) with a significant P value of 0.028, as shown in Fig. 2a. This identified cut-off value was subsequently validated in the independent validation set (P = 0.033) (Fig. 2b). Apart from pTCD8+CD28- level, other baseline characteristics that may contribute to first-line PFS were also assessed through univariate analysis. These characteristics included the primary lymph node stage (P = 0.017), the presence of uncommon metastatic sites (P = 0.001) and the utilisation of HER2-targeted therapy (P = 0.034) (Supplementary Table S2). The final Cox regression model in multivariate analysis was adjusted to include all variables that showed statistical significance in the aforementioned univariate analysis. Consistent with expectations, the Cox regression model confirmed that the baseline pTCD8+CD28− level served as an independent prognostic factor for first-line PFS among HER2 + MBC patients, and high pTCD8+CD28- level (≥18.0%) was associated with a prolonged median PFS (mPFS) compared to low level (10.5 vs. 17.4 months, P = 0.002) (Fig. 2c).

Fig. 2: The prognostic role of baseline pTcd8+cd28- in HER2 + MBC patients regarding first-line PFS.

Determine the prognostic role of pTcd8+cd28− in training set (a) and retest in the validation set (b). Cox regression analysis was performed for pTcd8+cd28−, including other factors contributing to the PFS. The HR values for each factor derived from the multivariate analysis are also presented (c). The prognostic role of pTcd8+cd28− in the chemotherapy plus trastuzumab subgroup (d), in the chemotherapy plus trastuzumab & pertuzumab subgroup (e), in the chemotherapy plus TKIs subgroup (f) and in chemotherapy-alone subgroup (g); The prognostic role of pTcd8+cd28− in patients who received anti-HER2 based therapy (h). pTcd8+cd28− peripheral CD8 + CD28- T cell, TKIs tyrosine kinase inhibitors.

The prognostic value of baseline pTCD8+CD28- level in different therapeutic subgroups

Within the cohort, HER2 + MBC patients were stratified into four subgroups based on their therapeutic regimens. In the subgroup receiving chemotherapy plus trastuzumab (n = 151), individuals with of pTCD8+CD28- high exhibited a prolonged mPFS compared to those with lower levels (10.5 vs. 12.7 months, P = 0.008), which aligns with our overall cohort findings (Fig. 2d). However, the significant difference in mPFS between pTCD8+CD28- high and pTCD8+CD28- low was not found in the subgroups receiving chemotherapy in combination with trastuzumab and pertuzumab (not reached vs. 14.7 months, P = 0.285), chemotherapy plus tyrosine kinase inhibitor (TKI) subgroup (16.8 vs. 23.2 months, P = 0.158) and chemotherapy-alone subgroup (9.2 vs. 13.8 months, P = 0.680) (Fig. 2e–g). Nevertheless, for HER2-targeting-based therapy (n = 223), the pTCD8+CD28- High at baseline was associated with prolonged mPFS (11.1 vs. 18.9 months, P = 0.001) (Fig. 2h).

The prognostic value of dynamic pTCD8+CD28− levels

Considering the aforementioned results, we sought to determine the relationship between the dynamic alteration of pTCD8+CD28- and first-line PFS. Based on variations in pTCD8+CD28-, patients who underwent anti-HER2-based therapy were categorised into four groups: high-level, low-level, reducing and enhancing groups respectively. The high-level group consistently maintained a high level of pTCD8+CD28- throughout the entire therapeutic course, while the low-level group consistently maintained a low level. The reducing group initially exhibited a high level but subsequently showed a low level at each follow-up visit, whereas the enhancing group demonstrated the opposite pattern (Fig. 3a). Consequently, we included a total of 139 patients who met these criteria in Kaplan–Meier analysis regarding first-line median PFS. Among them, the high-level group (n = 71) had the longest mPFS of 15.5 months, while the low-level group (n = 25) had the shortest first-line mPFS of 7.7 months. Meanwhile, both the reducing group (n = 6, mPFS = 11.4 months) and the enhancing group (n = 37, mPFS = 11.1 months) displayed similar mPFS durations. In summary, we demonstrated that the dynamic changes in pTCD8+CD28- levels were also associated with first-line PFS in patients with HER2 + MBC who received anti-HER2-based therapy (P = 0.000427) (Fig. 3b). However, it is important to note that this prognostic value was observed within a limited sample size (n = 139); therefore, further validation in an independent cohort is warranted for future studies.

Fig. 3: The prognostic value of dynamic change of pTcd8+cd28- regarding first-line PFS.

The diagram for grouping dynamic change of pTcd8+cd28− (a). The prognostic role of dynamic change of pTcd8+cd28− was assessed using K–M analysis (b).

Association of pTCD8+CD28- level with other clinical characteristics

In addition to assessing the prognostic value of pTCD8+CD28-, we also investigated its correlation with other clinical characteristics in the cohort. The baseline pTCD8+CD28- level exhibited a significant negative correlation with bone metastasis (coefficient = −0.189, P = 0.003), lymph node metastasis (coefficient = −0.152, P = 0.016), and the number of metastatic sites (coefficient = −0.200, P = 0.001) (Table 2). Furthermore, there was a negative correlation between changes in pTCD8+CD28- and bone metastasis (coefficient = −0.218, P = 0.006) (Supplementary Table S3). Notably, both bone and lymph node are all critical immune-related organs. Compared to health controls, HER2 + MBC patients exhibited a unique positive correlation between pTCD8+CD28- and total T cells (coefficient = 0.134, P = 0.033) as well as natural killer T cells (coefficient = 0.143, P = 0.024) (Table 3). These findings, however, require further validation and should be cautiously interpreted due to the low coefficients.

Table 2 Association of pTCD8+CD28− with the clinical characteristics.Table 3 Correlation of pTCD8+CD28− level with other peripheral lymphocyte subtypes.The cytotoxic potential of pTCD8+CD28- in HER2-positive MBC

To identify the in vivo cytotoxic potential of pTCD8+CD28−, we assessed the secretion of cytotoxic effectors, T-cell receptor (TCR) clonality, and transcriptome of pTCD8+CD28- derived from the enrolled patients. Initially, significantly higher levels of perforin (84.29 ± 3.3 vs. 19.14 ± 2.0%, P <0.001) and granzyme B (76.10 ± 4.2 vs. 16.65 ± 2.4%, P <0.001) were observed in pTCD8+CD28- in comparison to its precursor cell (pTCD8+CD28+) (Fig. 4a). Using single-cell RNA sequencing, seven T-cell subtypes were identified across pTCD8+CD28+ and pTCD8+CD28-, based on well-known expression patterns of cell markers (Fig. 4b, left). As compared to pTCD8+CD28+, the predominant cell subtypes of pTCD8+CD28- composed of CD8+ effector T cells and CD8+ effector memory T cells (92.48% vs. 27.92%) (Fig. 4b, right). These two subtypes exhibited higher cytotoxic scores among others (Fig. 4b, middle), indicating the cytotoxicity of pTCD8+CD28-. TCR repertoire analysis further revealed a lower diversity-higher clonality of pTCD8+CD28- as compared to pTCD8+CD28+, particularly for those effector T-cell subtypes of pTCD8+CD28- (Fig. 4c, upper). Meanwhile, in pTCD8+CD28-, we observed the expansion of most shared TCR clones between pTCD8+CD28+ and pTCD8+CD28- as shown in the Sankey diagram (Fig. 4c, lower). Collectively, the aforementioned data demonstrate that pTCD8+CD28- functions as an antigen-experienced effector T cell in HER2 + MBC.

Fig. 4: The assessment of the cytotoxic capacity of pTcd8+cd28-.

Detection of expression of Perforin and Granzyme B in pTcd8+cd28- and pTcd8+cd28+ (a). Comparison of cell cluster composition between pTcd8+cd28- and pTcd8+cd28+ using single-cell RNA sequencing (b). Cell cluster annotation based on well-known cell markers (b, left). Cytotoxic score evaluation of annotated T-cell clusters (b, middle). Cell cluster distributions of pTcd8+cd28- and pTcd8+cd28+ (b, right); TCR repertoire comparison between pTcd8+cd28- and pTcd8+cd28+ (c). TCR clonality distribution in pTcd8+cd28- and pTcd8+cd28+ (c, upper). The shared clones between pTcd8+cd28- and pTcd8+cd28+ and clone expansion in pTcd8+cd28- (c, lower).

Association of pTCD8+CD28- with the tumour immunity

To ascertain the relationship between pTCD8+CD28- levels and patients’ tumour immunity, we initially compared the TILs infiltration in individuals with high and low levels of pTCD8+CD28-. Considering the variability in TILs infiltration among different metastatic sites in HER2 + MBC (Fig. 5a, left), we separately analysed the TILs scores as depicted in right panel of Fig. 5a. Despite observing higher TILs scores among patients with pTCD8+CD28- high in the lymph and/or lung metastasis group (19.6 ± 3.9 vs. 16.5 ± 4.0%) and other metastasis sites group (6.7 ± 4.8 vs. 5.7 ± 1.2%), none of these differences reached statistical significance (Fig. 5a, right). Confocal immunofluorescence analysis revealed a positive quantitative correlation between pTCD8+CD28- and infiltrated CD8 + CD28- T cells in the paired tumour lesions (P = 0.037, Fig. 5b, right). We then compared the serum level of various inflammatory cytokines between high and low level of pTCD8+CD28-. Patients with pTCD8+CD28- high exhibited elevated IL-2 levels (386.6 ± 73.7 vs. 206.0 ± 28.5 pg/ml, P = 0.034) and decreased TGF-β levels (23.7 ± 2.7 vs. 34.2 ± 3.1 ng/ml, P = 0.016) comparing to pTCD8+CD28- low. However, no statistical differences were observed for IFN-γ or TNF-α (Fig. 5c). In addition, cfDNA-based sequencing revealed a higher prevalence of CD274 (PD-L1) deletion in pTCD8+CD28- high as compared to pTCD8+CD28- low (P = 0.041, Supplementary Fig. S4).

Fig. 5: The association of pTcd8+cd28- level with tumour immunity.

Comparison of tumour-infiltrating lymphocyte level based on metastatic sites (a, left) and pTcd8+cd28- levels across different metastatic sites (a, right); Comparison of infiltrated CD8 + CD28- T cells from patients with pTcd8+cd28- high and pTcd8+cd28- low (b). Serum levels of IFN-γ, IL-2, TNF-α, and TGF-β in patients with pTcd8+cd28- high and pTcd8+cd28- low (c). The changes in peripheral CD3 + CD8 + T cells (left), CD8 + PD1 + T cells (middle) and CD8 + CD28- T cells (right) following combination therapy with anti-PD1 and MASCT (d).

To further substantiate the association between pTCD8+CD28- and tumour immunity, we investigated the variations of pTCD8+CD28- mediated by immunotherapy. Regrettably, immunotherapy is not currently employed as a standard regimen for HER2 + MBC patients. Alternatively, we evaluated our hypothesis in the context of metastatic triple-negative breast cancer (mTNBC) patients who are eligible for immunotherapy. We enrolled five mTNBCs patients received combination therapy of anti-PD1 and multiple antigen-specific cell therapy (MASCT) at different therapeutic lines. Specifically, patients with PD-L1-positive tumours underwent a dose-escalation study using intravenous camrelizumab (3 mg/kg, every 2 weeks), along with intravenous infusion of T cells and dendritic cells (DCs) every 27–36 days. Meanwhile, the baseline and subsequent follow-up assessments until disease progression revealed a decrease in peripheral levels of CD8 + PD1 + T cells (Fig. 5d, left) and an increase in pTCD8+CD28- levels (Fig. 5d, middle) across all five patients. In contrast, the fluctuation in peripheral CD3 + CD8 + T-cell levels did not demonstrate a consistent pattern, thereby ruling out the possibility of false-positive amplification of pTCD8+CD28- due to T-cell infusion (Fig. 5d, right).