Establishment of SE-i•FISH for breast cancer CTCs in situ phenotype and karyotype identification
SE-i•FISH was optimized to monitor breast cancer CTCs expressing tumor biomarkers and Chr8 aneuploidy. Immunofluorescence staining of CTCs and CECs with anti-EpCAM and mesenchymal marker vimentin showed distinct intracellular staining for both EpCAM and vimentin. (Fig. 1).
Analysis of CTCs detected by patient classification
The CTC positive detection rate was 43/45 in LABC patients (95.6%) before NCT, and 100% following the first course of NCT and post-NCT.
Patients were divided into two groups according to changes in Ki-67. The High-R group was comprised of 22 patients (48.9%) and the Low-R group of 23 patients (51.1%). According to the Miller-Payne system, 8/45 patients obtained > 90% loss of tumor cells (grade 4–5), while the other 37 patients were assigned to grade 1–3. All eight grade 4–5 patients belonged to the High-R group when sorted according to Ki-67 changes.
Patient clinicopathological features and their association with CTC detection are shown in Table 1. The CTC positive rate was not correlated with age, HER-2 status, or lymph node metastasis. However, patients with more than two lymph node metastases exhibited significantly higher numbers of CTCs (repetitive measurement deviation analysis, p = 0.029). Further, during NCT, CTC numbers increased in patients with more than two metastatic lymph nodes (p < 0.001 and 0.043 for post-1st NCT and post-NCT, respectively, additional file 1, Figure S1A). Specifically, the CTC number was high only during NCT in the High-R group, yet was continuously elevated in the Low-R group. Alternatively, for patients with two or less metastatic lymph nodes, the CTC number showed a downward trend following the initial rise in the High-R group; while in the Low-R group the number was found to steadily increase (Additional file 1, Figure S1A and 1B).
Table 1
CTC numbers of patients with different clinical characteristics
Factors | Total (No.) | High-R (No.) | Low-R (No.) | | CTC number | | |
| | | | p value | pre-NCT | post-1st NCT | post-NCT | p value1 | p value2 |
Total | 45 | 22 | 23 | | | | | | |
Age | | | | 0.181 | | | | < 0.001 | 0.545 |
< 50 | 22 | 13 | 9 | | 8.18 ± 7.82 | 111.18 ± 160.25 | 32.55 ± 45.48 | | |
≥ 50 | 23 | 9 | 14 | | 6.44 ± 5.53 | 49.78 ± 63.22 | 66.91 ± 93.04 | | |
| | | | | | | | | |
Her-2 status | | | | 0.848 | | | | 0.001 | 0.111 |
Negative | 28 | 14 | 14 | | 6.72 ± 6.18 | 70.14 ± 130.58 | 40.28 ± 61.69 | | |
Positive | 17 | 8 | 9 | | 8.31 ± 7.73 | 97.31 ± 110.92 | 67.94 ± 94.06 | | |
| | | | | | | | | |
Molecular subtype | | | | 0.815 | | | | 0.002 | 0.117 |
Hormone + Her-2-/+ | 33 | 17 | 16 | | 7.30 ± 6.82 | 58.49 ± 76.48 | 47.12 ± 77.27 | | |
TNBC | 10 | 4 | 6 | | 7.25 ± 6.74 | 138.42 ± 196.83 | 58.33 ± 70.71 | | |
Hormone-Her-2+ | 2 | 1 | 1 | | 8.50 ± 3.54 | 140.00 ± 195.16 | 124.00 ± 4.24 | | |
| | | | | | | | | |
Lymph node | | | | 0.100 | | | | < 0.001 | 0.029 |
≤ 2 | 23 | 14 | 9 | | 6.63 ± 6.77 | 48.92 ± 52.36 | 34.83 ± 57.57 | | |
> 2 | 22 | 8 | 14 | | 8.05 ± 6.76 | 115.81 ± 166.65 | 67.57 ± 89.14 | | |
P value1: different timepoints |
P value2: clinical characteristics |
CTC numbers analyzed by patient group: correlation with NCT effect
A typical fluorescent photograph of a CTC is shown in Fig. 2A. Generally, the number of CTCs increased after the first NCT cycle compared to pre-NCT levels (p < 0.001). This trend continued post-NCT (p = 0.001, Fig. 2B). The number of CTCs was higher in the Low-R group (repetitive measurement deviation analysis, p = 0.042). There were no significant differences in CTC number between High-R and Low-R groups before NCT or after the first course of NCT (Fig. 2C). However, following the eighth NCT course, the difference between the two groups was significant (p = 0.028). The CTC number increased slightly (p = 0.051), then decreased to baseline in the High-R group, while in the Low-R group, the number of CTCs continuously increased continuously during NCT (Fig. 2D).
According to the Miller-Payne system, the CTC number of grade 1–3 patients increased significantly and continuously following all NCT courses compared to grade 4–5 patients (Fig. 2E and 2F).
Chromosome 8 karyotype changes between patients with different NCT responses
The existence of heterogeneous polysomic Chr8 confirmed the marked heterogeneity of breast cancer CTCs (Fig. 3A); while the ratios of CTCs with different Chr8 ploidies changed during treatment (Fig. 3B). The frequencies of CTCs with triploid Chr8 copy numbers were 28, 34, and 45% at the three successive time points, and the corresponding tetraploid Chr8 numbers were 14, 27, and 24%, respectively. Analysis of the occurrence of triploid, tetraploid, pentaploid, or higher Chr8 copies in CTCs according to NCT response showed that the number of CTCs containing Chr8 triploidy and tetraploidy was higher in the Low-R group than in the High-R group following NCT, however, this effect was not observed pre- or post-initial NCT (p = 0.017 and 0.009 for triploidy and tetraploidy, respectively, Fig. 3C and 3G). Compared to the CTC number before NCT, an increase in Chr8 triploidy and tetraploidy was observed in the Low-R group post-first NCT (p = 0.003 and 0.010, respectively) and post-NCT (p = 0.002 both in triploid and tetraploid Chr8); while in the High-R group, Chr8 triploidy only increased post-first NCT, after which it returned to baseline. No increase in tetraploid Chr8 was observed in CTCs following NCT in High-R group (Fig. 3D and 3H). Moreover, no significant differences were observed between the Low-R and High-R groups during NCT in the numbers of multiploid (≥ pentasomy 8) CTCs for either of the time points (Fig. 3K). Following NCT, the number and proportion of triploid and tetraploid CTCs rose continuously in only the Low-R group, not the High-R group.
In addition, no significant intragroup differences were observed in multiploid CTCs (Fig. 3L). When patients were classified by the Miller-Payne system, the trends observed in CTC Chr8 triploidy, tetraploidy and multiploid variations were similar to the ki-67 group mode (Fig. 3E-F, I-J, M-N).
Correlation of the number of small size CTCs with NCT response
A vimentin + small size CTC is depicted in Fig. 4A (left). The number of small CTCs (< 5 µm) increased after the first NCT cycle (p < 0.001), after which it remained elevated post-NCT (p = 0.001), consistent with the overall changes observed in CTCs (Fig. 4B). Further, the number of small CTCs in Low-R patients was significantly higher during the final time point after the eighth course of NCT (p = 0.038; Fig. 4C). This difference, however, reflected growth in the number of small CTCs after the first course (p = 0.010), which then plateaued at a comparatively high level (p = 0.003). Unlike the Low-R group, the CTC level of the High-R group remained approximately constant throughout the NCT process (Fig. 4D). The percentage of small CTCs in the High-R group were 30.5%, 30.4%, and 35.9%, respectively, at the three timepoints (Fig. 4G). The corresponding data for the Low-R group were 14.5%, 32.3%, and 46.7%. Additionally, the patients designated as Miller-Payne grade 1–3 exhibited the same trend in variation as that observed in the Low-R group, while grade 4–5 was observed to be the same as the High-R group (Fig. 4E-F).
CTM of LABC patients during NCT
A vimentin + CTM is depicted in Fig. 4A (right). No significant changes were observed in CTM numbers following the eighth course of NCT (repetitive measurement deviation analysis). However, the results in Low-R patients following the first course of treatment showed apparently higher numbers of CTM compared to those before NCT (Fig. 4I). No significant differences were observed between High-R and Low-R groups at the other timepoints; the CTM increase in the Low-R group only slightly exceeded the significance level (p ≤ 0.05) at the second time point, post-first course (p = 0.067) (Fig. 4H).
CTC, mesenchymal phenotypes, and NCT sensitivity
Next, CTCs were grouped according to the presence or absence of the mesenchymal marker vimentin (Fig. 5A). The number of total and small vimentin− CTCs increased during NCTl. Further, the vimentin+ CTCs increased after the first NCT course, and decreased after the last course, whereas small vimentin+ CTCs increased after the first course, then remained approximately constant (Fig. 5B and 5C). Patients in the Low-R group displayed higher numbers of vimentin− CTCs and small CTCs after NCT (p = 0.060 and 0.038 for CTCs and small CTCs, respectively, Fig. 5H and 5J). Conversely, no difference was found in vimentin+ CTCs, while an increased number of vimentin+ small CTCs was detected in the High-R group following the first course of NCT (Fig. 5D and 5F). Further, vimentin− CTCs remained constant over the course of NCT treatment in the High-R group and increased in the Low-R group at which level it remained relatively constant until surgery (p = 0.002 and 0.008 for the first and eighth courses of NCT, respectively, Fig. 5I). In contrast, vimentin+ CTCs showed no significant changes during NCT (Fig. 5E). In terms of vimentin+ small CTCs, no significant changes in number were detected in the Low-R group during NCT. In contrast, vimentin+ small CTCs increased post-first NCT course after which they returned to baseline in the High-R group (Fig. 5G). Vimentin− small CTCs showed no significant change in the High-R group, but in the Low-R group an obvious increase was observed after the first course of NCT, then levels remained high (Fig. 5K).
Cumulatively, these data indicate a possible correlation between low response to NCT and increased vimentin− CTCs with no reduction in vimentin+ CTC numbers. The proportion of total vimentin− CTCs confirmed this conclusion. In the Low-R group, the percentage of vimentin− small CTCs rose from 10% before NCT to 46% after completion (Fig. 5L).
Correlation between CTCs and non-cancer cells (lymphocyte, neutrophils and platelets), CTCs and cancer biomarkers during NCT
The relationships between aneuploid CTCs and non-cancer cells (lymphocyte, neutrophils and platelets) were shown in Fig S2 (Additional file 2). The number of CTCs was positively correlated with neutrophils (p = 005, r = 0.416) and platelets (p = 0.421, r = 0.004) before treatment, however, these correlations were not apparent following the first course of NCT. Moreover, no correlation was observed between CTCs and the lymphocyte, tumor markers CEA, CA12-5, and CA15-3 (Additional files 2 and 3, Fig S2 and S3).
Predictive value of CTC numbers
ROC curve analyses indicated that the CTC total, as well as the individual subtypes, effectively predicted the NCT response after the eighth course of NCT. The AUCs for CTCs with different subtypes after the eighth course of NCT were 0.770 (95% confidence interval (CI) 0.630–0.909), 0.776 (95% CI 0.642–0.910), 0.777 (95% CI 0.640–0.913), 0.784 (95% CI 0.649–0.918), 0.772 (95% CI 0.632–0.911) and 0.792 (95% CI 0.663–0.922) for overall CTCs, small CTCs, trisomy 8 CTCs, tetrasomy 8 CTCs, vimentin− CTCs, and vimentin− small CTCs, respectively (all p values < 0.05, Fig. 6A). Similarly, the ∆value2 (difference between the third and the first measurements) was an effective predictor of outcome. The AUCs for this measurement were 0.797 (95% CI 0.670–0.925), 0.822 (95% CI 0.702–0.942), 0.776 (95% CI 0.641–0.910), 0.804 (95% CI 0.676–0.933), 0.794 (95% CI 0.666–0.923), and 0.845 (95% CI 0.732–0.958), respectively (all p values < 0.05, Fig. 6B). Considering the samples after only the first course of treatment, only the ∆value1 (difference between the second and the first measurements) of small vimentin− CTCs exhibited significant diagnostic value (AUC 0.566; 95% CI 0.396–0.737; Fig. 6C). However, a lower degree of diagnostic efficiency was achieved when the Miller-Payne system was adopted, which may be related to the unbalanced sample size (Additional file 4, Fig S4).
A three-year follow-up of LABC patients was conducted to evaluate the curative effects of NCT and the rationality of the surrogate endpoints including Ki-67 changes. We set the media value of CTC numbers as a threshold, and divided subjects into a CTC-High and CTC-Low group. The comparison of progression-free survival (PFS) and overall survival (OS) between the two groups was performed using Log-rank test. Results show that patients with higher CTC numbers exhibited a significantly shorter PFS and OS compared to those in the CTC-Low group after the 8th NCT course (Fig. 6D and 6E). In addition, we evaluated the survival of patients with High-R and Low-R (according to the Ki-67 index) and found patients in the High-R group had significantly higher PFS and OS than patients in the Low-R group (Additional file 5, Fig S5), which confirms the high reliability of the grouping methods.