Cell-cycle-phase progression analysis identifies unique phenotypes of major prognostic and predictive significance in breast cancer

Abstract

Multiparameter analysis of core regulatory proteins involved in G1-S and G2-M cell-cycle transitions provides a powerful biomarker readout for assessment of the cell-cycle state. We have applied this algorithm to breast cancer to investigate how the cell cycle impacts on disease progression. Protein expression profiles of key constituents of the DNA replication licensing pathway (Mcm2, geminin) and mitotic machinery (Plk1, Aurora A and the Aurora substrate histone H3S10ph) were generated for a cohort of 182 patients and linked to clinicopathological parameters. Arrested differentiation and genomic instability were associated with an increased engagement of cells into the cell division cycle (P<0.0001). Three unique cell-cycle phenotypes were identified: (1) well-differentiated tumours composed predominantly of Mcm2-negative cells, indicative of an out-of-cycle state (18% of cases); (2) high Mcm2-expressing tumours but with low geminin, Aurora A, Plk1 and H3S10ph levels (S-G2-M progression markers), indicative of a G1-delayed/arrested state (24% cases); and (3) high Mcm2-expressing tumours and also expressing high levels of the S-G2-M progression markers, indicative of accelerated cell-cycle progression (58% of cases). The active cell-cycle progression phenotype had a higher risk of relapse when compared with out-of-cycle and G1-delayed/arrested phenotypes (HR=3.90 (1.81-8.40, P<0.001)), and was associated with Her-2 and triple negative subtypes (P<0.001). It is of note that high-grade tumours with the G1-delayed/arrested phenotype showed an identical low risk of relapse compared with well-differentiated out-of-cycle tumours (HR=1.00 (0.22-4.46), P=0.99). Our biomarker algorithm provides novel insights into the cell-cycle state of dynamic tumour cell populations in vivo. This information is of major prognostic significance and may impact on individualised therapeutic decisions. Patients with an accelerated phenotype are more likely to derive benefit from S- and M-phase-directed chemotherapeutic agents.

Figure 1

(A) Diagrammatic representation of the mitotic cell division cycle. (B) Phase-specific distribution of cell-cycle biomarkers in proliferating cells and out-of-cycle states. (C) Cell-cycle-phase-specific chemotherapeutic and mechanistic agents. Drugs highlighted in green are commonly used in the treatment of breast cancer.

Figure 2

Kaplan–Meier curves showing association between NPI, Plk1 and Aurora A and disease-free survival (days from diagnosis to death, recurrence or last follow-up) across the whole series. (A) NPI and disease-free survival segregated into the three decision group categories, P<0.001. (B) Aurora A and disease-free survival stratified by those patients lying above or below the median LI; HR=3.31 (1.67–6.57), P<0.001. (C) Plk1 and disease-free survival stratified by those patients lying above or below the median LI; HR=4.48 (2.21–9.09), P<0.001. (D) Effect of Plk1 after adjustment for NPI; HR=3.31 (1.57–6.97) risk above Plk1 median relative to below median, P=0.002.

Figure 3

Effect of cell-cycle biomarkers on disease-free survival before and after adjusting for NPI. All biomarkers are associated with disease-free survival. The effects remain statistically significant after adjusting for NPI, so that they are predictive, independent of NPI.

Figure 4

Distribution of cell-cycle biomarker expression defines three distinct cell-cycle phenotypes: (I) out-of-cycle state, (II) in cycle G1-delayed/arrested state and (III) actively cycling state. The median (solid black line), interquartile range (boxed) and robust range excluding outlying cases (enclosed by lines) are shown. Outlying cases are shown by isolated points (LI, labelling index).

Figure 5

Kaplan–Meier curves showing association between cell-cycle phenotype and disease-free survival. (I) Out-of-cycle state, (II) in cycle G1-delayed/arrested state and (III) actively cycling state. On univariate analysis, comparing phenotype III with phenotypes I and II combined, HR=3.90 (1.81–8.40), P<0.001. On multivariate analysis, adjusted for NPI, HR=2.71 (1.18–6.23), P=0.019.

Figure 6

Relationship between cell-cycle phenotype and breast cancer subtypes. The panels show the proportion of each breast cancer subtype, which display cell-cycle phenotypes I (out of cycle), II (G1-delayed/arrested) and III (actively cycling). It is of note that the majority of Her-2 and triple negative tumours display the actively cycling phenotype (III).

Figure 7

Proposed prognostic and predictive cell-cycle-phase algorithm in breast cancer. Three distinct cell-cycle phenotypes characterised by the differential expression of cell-cycle biomarkers Mcm2, Aurora A, geminin, Plk1 and H3S10ph. Prognosis and treatment response can be predicted from the distinct immunoexpression profile displayed by each tumour.