Protein analysis of extracellular vesicles to monitor and predict therapeutic response in metastatic breast cancer

Abstract

Molecular profiling of circulating extracellular vesicles (EVs) provides a promising noninvasive means to diagnose, monitor, and predict the course of metastatic breast cancer (MBC). However, the analysis of EV protein markers has been confounded by the presence of soluble protein counterparts in peripheral blood. Here we use a rapid, sensitive, and low-cost thermophoretic aptasensor (TAS) to profile cancer-associated protein profiles of plasma EVs without the interference of soluble proteins. We show that the EV signature (a weighted sum of eight EV protein markers) has a high accuracy (91.1 %) for discrimination of MBC, non-metastatic breast cancer (NMBC), and healthy donors (HD). For MBC patients undergoing therapies, the EV signature can accurately monitor the treatment response across the training, validation, and prospective cohorts, and serve as an independent prognostic factor for progression free survival in MBC patients. Together, this work highlights the potential clinical utility of EVs in management of MBC.

Fig. 1. Thermophoretic aptasensor (TAS) for detecting protein markers of EVs.

a Schematic of the TAS procedure. Clinical plasma samples (1 µL, diluted by 100-folds) were incubated with Cy5-conjugated aptamers to bind to target proteins on EVs, and then subjected to thermophoretic accumulation to amplify the fluorescence signal of aptamer-bound EVs, enabling rapid and sensitive detection of EV protein markers. b Size distribution of EVs derived from three BC cell lines SK-BR-3 (red line), BT-474 (blue line), and MDA-MB-231 (green line), and benign mammary epithelial cell line (MCF-10A, violet line) using nanoparticle tracking analysis (NTA). Size modes are indicated. c Wide-field TEM image of EVs. The representative image is shown from three independent repeats. Scale bar, 500 nm. d TAS and ELISA measurement of expression levels of CA 15-3 (red bars), CA 125 (blue bars), and CEA (green bars) in three types of samples: (i) EV-depleted plasma diluted by 100-folds in 1× PBS; (ii) the diluted EV-depleted plasma spiked with soluble proteins; (iii) the sample ii spiked with plasma EVs (2 × 10⁹ mL⁻¹, n = 3 samples for each protein marker). Statistical difference was determined by a two-sided, parametric t test. P value is indicated in the chart. e Sensitivity of TAS (red dots) and ELISA (blue dots) for the detection of plasma EVs incubated with CEA aptamer (0.1 μM) (n = 3 samples for each EV concentration). R square (R²) is indicated. f Fluorescence images of aptamer-labeled EVs (10¹⁰ mL⁻¹) after thermophoretic accumulation showing elevated levels of eight EV protein markers from the three BC cell lines, SK-BR-3, BT-474, and MDA-MB-231, compared to MCF-10A. Scale bar, 50 μm. Images are shown from a single measurement. g Radar plot showing TAS analyses of 8 EV protein markers from the four different cell lines (BT-474 is represented by red dots, SK-BR-3 by blue dots, MDA-MB-231 by green dots, and MCF-10A by violet dots). Error bars represent the mean ± s.d. in (d, e). Source data are provided as a Source Data file.

Fig. 2. Profiling of plasma EV proteins from MBC, NMBC, and HD by TAS.

a Fluorescence images showing elevated levels of 8 EV protein markers (CA 15-3, CA 125, CEA, HER2, EGFR, PSMA, EpCAM, and VEGF) in plasma samples from a MBC patient (HR + subtype) and a NMBC patient (HER2 + subtype) compared to HD. Images are shown from a single measurement. Scale bar, 50 μm. b Radar plot showing TAS analyses of eight EV protein markers from MBC (red dots), NMBC (blue dots), and HD (gray dots). c Heatmap of EV protein profiles from 36 MBC patients, 21 NMBC patients, and 66 HD. d Precision-Recall Curves (PRC) for single EV protein markers to differentiate between BC and HD, as well as MBC and NMBC. e Summary of AUPRC (area under the PRC) of single EV protein markers for BC versus HD discrimination and MBC versus NMBC discrimination. f Pearson correlation matrix showing weak correlations between any pair of EV protein markers, as well as between EV markers and plasma CA 15-3. Plasma CA 15-3 is represented as P CA 15-3. Pearson correlation coefficient (r) is indicated. Source data are provided as a Source Data file.

Fig. 3. EV^DX signature for differentiation of MBC, NMBC, and HD.

a, b PRC for the EV^DX signature (a weighted sum of eight EV markers by LDA, solid lines) and SUM (unweighted sum of eight EV markers, dashed lines). c Confusion matrix showing an overall accuracy of 91.1% across MBC, NMBC, and HD. d Hierarchical clustering of individual EV protein markers and the EV^DX signature showing no segregation according to metastasis sites of MBC patients. e Significant (P = 0.0408) elevation of EGFR level on EVs from MBC patients with lung metastasis (n = 12 patients, red dots), as compared to MBC patients without lung metastasis (n = 21 patients, gray dots). Statistical difference was determined by a two-sided, nonparametric Mann–Whitney test. P value is indicated in the chart. f Concordance between tumor size and the EV^TS signature identified using multivariate linear regression (MLR). Mean square errors (MSE) and R squared (R²) are indicated. Linear regression result is indicated by the dashed line. Error bars represent the mean ± s.d. in (e). Source data are provided as a Source Data file.

Fig. 4. EV analyses for classification of MBC treatment response in the training and validation cohorts.

a Heatmap of ΔIntensity of EV protein markers for the training cohort involving patients with PR (n = 18), SD (n = 17), or PD (n = 10) and the validation cohort involving patients with PR (n = 13), SD (n = 13), or PD (n = 7). b Values of the EV^M signature (the weighted sum of ΔIntensity of eight markers by LDA) in the training cohort (PR/SD: n = 35 samples, void blue dots; PD: n = 10 samples, void red dots) and the validation cohort (PR/SD: n = 26 samples, solid blue dots; PD: n = 7 samples, solid red dots) c, d Confusion matrix showing that the EV^M signature had an accuracy of 88.9 % in differentiating PD from PR/SD for the training cohort (c) and 87.9% for the validation cohort (d). Statistical differences were determined by two-sided, nonparametric Mann–Whitney test (b). P values are indicated in the charts. Error bars represent the mean ± s.d. in (b). Source data are provided as a Source Data file.

Fig. 5. Prospective cohort of MBC patients for treatment response monitoring.

a Heatmap of ΔIntensity of EV protein markers in patients with PR (n = 11), SD (n = 9), or PD (n = 7). b Values of the EV^M signature (the weighted sum of ΔIntensity of eight markers by LDA) for SD/PR (n = 20 samples, blue dots) and PD (n = 7 samples, red dots) groups. c Confusion matrix showing that the EV^M signature had an accuracy of 85.2% in differentiating PD from PR/SD. d EV^M signature values in the HR + BC group (PR/SD: n = 26 samples, void red dots; PD: n = 9 samples, solid red dots), the HER2 + BC group (PR/SD: n = 33 samples, void blue dots; PD: n = 8 samples, solid blue dots), and the TNBC group (PR/SD: n = 22 samples, void green dots; PD: n = 7 samples, solid green dots). e ROC curves showing similar performance of the EV^M signature for classifying treatment response in different BC subtypes (HR + , red line; HER2 + , blue line; TNBC, green line). Statistical differences were determined by two-sided, nonparametric Mann–Whitney test (b, d). P values are indicated in the charts. Error bars represent the mean ± s.d. in (b, d). Source data are provided as a Source Data file.

Fig. 6. Comparison of the EV^M signature and plasma CA 15-3 in longitudinal monitoring for MBC.

The EV^M signature, CA 15-3 level, and therapeutic response were summarized for MBC patients in different BC subtypes (n = 5 for each subtype). The cumulative sum of the EV^M signature at each time point was calculated and plasma CA 15-3 level was normalized by the cutoff value of 25 U mL⁻¹. The solid gray, green, blue, and red dots indicate the EV^M signature at the baseline, PR, SD, and PD, respectively. The void gray, green, blue, and red dots indicate the plasma CA 15-3 at the baseline, PR, SD, and PD, respectively. Source data are provided as a Source Data file.

Fig. 7. EV protein markers for prediction of progression-free survival (PFS) in a MBC cohort.

Kaplan–Meier curves showing PFS of 59 MBC patients according to the EV^P signature (a), CA 15-3 (b), CA 125 (c), CEA (d), HER2 (e), EGFR (f), PSMA (g), EpCAM (h), and VEGF (i) on EVs before treatment (baseline). The baseline level (high or low) was stratified according to the median value. The significance of the difference was calculated by a two-sided log-rank test. Hazard ratio (HR) and 95% CI were calculated using Cox proportional-hazard regression with a univariate model. Source data are provided as a Source Data file.