Premenopausal women with breast cancer in the early post-partum period show molecular profiles of invasion and are associated with poor prognosis

We accessed women with BC aged ≤ 45 years from a retrospective series of 777 tumors. Among these, 155 (20%) were ≤ 45 years with a mean age ± SD of 38.9 ± 5 years. 48% had tumors size > 3 cm and nearly half (45%) were grade 3. Most tumors were lymph node positive (60%) and had a higher proportion of triple-negative BC (29%).

Categorization based on the time since LCB showed 15% (23/155) were within the first 5 years (PPBC1), 25% (38/155) were between 6 and 10 years (PPBC2), 46% (72/155) had tumors 10 years after most recent childbirth (PPBC3), and 14% (22/155) were NPBC.

Association of reproductive and clinicopathological features among different groups

Examination of the association of reproductive features between the 3 PPBC groups showed PPBC1 patients had a significantly older age at FCB and LCB compared to PPBC2 and PPBC3 (FCB, p = 0.007 and p < 0.0001; LCB, p = 0.020 and p < 0.0001, respectively). We also noticed that PPBC2 patients had significantly higher age at FCB and LCB than PPBC3 patients (FCB, p = 0.003; LCB, p < 0.001). In addition, within the PPBC1 group, 91% of patients were comparatively younger at menarche (Fisher exact test, p = 0.047). No association with the number of live births (calculated as parity) was observed among the parous groups. Results are represented in Table 1. There was no association with the family history of BC.

Table 1 Association of parous and nulliparous groups with reproductive features

Next, we examined the differences in clinicopathological characteristics such as tumor size, grade, lymph node status, lymphovascular invasion, the density of tumor-infiltrating lymphocytes, estrogen receptor, progesterone receptor, and HER2 status between the parous and nulliparous groups. No significant difference was observed among the groups (Supplementary Table 1, Additional file 1).

Association of post-partum tumors with prognosis

Further, we examined the disease-free survival between the parous and nulliparous groups. 141/155 (91%) had complete follow-up information, with the longest and shortest of 102 and 19 months, respectively. Kaplan–Meier survival analysis showed no difference between parous and nulliparous groups (mean survival time ± SD, 72.6 ± 3.3 vs 68.5 ± 4 months, log-rank test, p = 0.52). Similarly, no difference in survival probability was observed between the PPBC subgroups and NPBC (mean survival time ± SD for PPBC1, PPBC2, PPBC3, and NPBC was 61.8 ± 8.2, 71.3 ± 3.7, 68.9 ± 4 and, 68.5 ± 4 months, respectively, log-rank, p = 0.314) (Fig. 1A).

Fig. 1figure 1

Kaplan–Meier survival analysis for disease-free survival (DFS). A Comparison of DFS among four groups in all tumors (n = 141). B Comparison of DFS among four groups in lymph node (LN)-positive tumors (n = 78) showing PPBC1 tumors are associated with poor prognosis compared to PPBC2 and NPBC tumors

Next, we performed subset analysis within lymph node-positive tumors alone. 78/141 (56%) were lymph node-positive. Among these, PPBC1 patients (n = 14) had a worse prognosis compared to PPBC2 (n = 20) and NPBC (n = 9) patients (p = 0.015 and p = 0.026 and mean survival time ± SD for PPBC1, PPBC2, and NPBC was 40.2 ± 2.7, 65.3 ± 5, and 64.3 ± 5.5 months, respectively). Similarly, PPBC3 patients (n = 35) had better survival (mean survival time ± SD = 60.3 ± 6 months) compared to PPBC1 patients although not statistically significant (p = 0.60) (Fig. 1B).

In concordance with the Kaplan–Meier analysis within lymph node-positive tumors, cox proportional hazard analysis showed that PPBC1 tumors had a higher hazard ratio compared to PPBC2 tumors [HR = 4.83 (95% CI = 1.12–19.61) (p = 0.028)] and NPBC tumors [HR = 5.30 (95% CI − 1.006 to 27.92) (p = 0.049)] on univariate analysis and [HR = 4.91 (95% CI − 0.68 to 35.13) (p = 0.11)] on multivariate analysis.

Differentially regulated pathways between early and late post-partum groups

Further, to understand the biology of tumors in the early and late post-partum period, we carried out RNA sequencing of the PPBC1 (n = 3), PPBC2 (n = 3) and PPBC3 (n = 4) tumor samples. The clinical characteristics of these ten tumors are mentioned in supplementary table 2 (Additional file 1). We obtained 587 DEGs between PPBC1 vs PPBC2 (280 genes upregulated and 307 genes downregulated). Similarly, 894 DEGs were identified between PPBC1 vs PPBC3 (435 genes upregulated, 459 genes downregulated). Unsupervised hierarchical clustering of these ten tumors showed a distinct separation of 8 tumors into three PPBC groups (Fig. 2A). In addition, these samples clustered better based on post-partum duration than ER status or the molecular subtypes (Supplementary Fig. S1, Additional file 1), suggesting that the post-partum period significantly impacts tumor gene expression.

Fig. 2figure 2

A Unsupervised hierarchical clustering of 10 samples across three PPBC groups. B Gene Ontologies from enrichment analysis in early PPBC (E-PPBC) tumors (n = 3) compared to late PPBC (L-PPBC) tumors (n = 7) showing major differentially regulated biological processes. The size of the bubble corresponds to the number of genes involved. C. Curated pathways by gene set enrichment analysis

Further, to understand the expression of genes that are concurrently expressed and regulated in the post-partum period since the LCB, we compared the DEGs between PPBC1 vs PPBC2 and PPBC1 vs PPBC3 (Supplementary Fig. S2, Additional file 1). We identified 178 DEGs that were common and overlapping between the groups. Unsupervised hierarchical clustering showed that the expression pattern of these selected DEGs was similar in PPBC2 and PPBC3 (Supplementary Fig. S3A, Additional file 1), and the PPBC1 tumors had a distinct pattern compared to the others. Further, PPBC1 tumors clustered separately in principal component analysis (PCA) (Supplementary Fig. S3B, Additional file 1) from the other two groups (PPBC2 and PPBC3). These results indicated that the post-partum tumors in the immediate vicinity of 5 years, since the LCB were distinctly different in their gene expression from those in the later post-partum period. Hence, we labeled PPBC1 as early PPBC (E-PPBC, post-partum period ≤ 5 years) and PPBC2 and PPBC3 tumors were combined and regrouped as late PPBC tumors (L-PPBC, post-partum duration > 6 years) for further analysis.

Next, we checked the differential pathways regulated between E-PPBC and L-PPBC tumors. There were 745 DEGs between the two groups (287 DEGs were upregulated and 458 DEGs were downregulated). Functional enrichment analysis demonstrated that the E-PPBC tumors were enriched in immune-related processes, ECM organization/degradation, and tumor cell invasion. Gene ontology-based analysis showed that E-PPBC compared to L-PPBC tumors were enriched for immune-related terms, such as leucocyte/lymphocyte/T cell activation, innate and adaptive immune response, and cytokine-mediated signaling, among others. Apart from the immune-related terms, we observed other upregulated processes related to angiogenesis, ECM organization, cell migration, wound healing, and BMP signaling. On the other hand, downregulated processes were cell morphogenesis and differentiation, cell adhesion, ion transport, lipid metabolic processes, response to estrogen and progesterone, regulation of MAPK cascade, and ERBB2-ERBB3 signaling pathway. Figure 2B represents major differentially regulated biological processes in E-PPBC compared to L-PPBC. Curated pathway analysis by GSEA showed that E-PPBC were enriched in tumor invasion along with ECM and its component degradation. In line with the gene ontology analysis, pathways related to cell cycle, DNA repair, and response to steroid hormones were downregulated (Fig. 2C).

To examine if the biological processes observed through gene ontologies were related to the involution process during the normal post-partum period, we obtained gene expression data from normal breast [15] at various time points after childbirth. Based on the time since the LCB, we categorized them into early and late post-partum periods using a similar cut-off of ≤ 5 years. We noticed that most gene ontologies, such as immune-related pathways, cell adhesion, response to steroid hormones, ERBB signaling, cell differentiation and morphogenesis, and angiogenesis were among the commonly enriched ontologies specific to early post-partum duration irrespective of cancerous and non-cancerous conditions (Supplementary Fig. S4, Additional file 1). The major differences observed in the early post-partum period in cancerous conditions were upregulation of angiogenesis, enrichment of macrophage markers, interleukins 4 and 13 signaling, cell migration, and wound healing.

Estimation of immune cell subtypes between the E-PPBC and L-PPBC groups

As GSEA analysis showed predominant upregulation of immune-related pathways in E-PPBC tumors compared to L-PPBC, we examined the immune cell subtypes within each group of tumors. We estimated the immune cell infiltration between the groups using deconvolution-based methods using CIBERSORT, xCell, EPIC, MCP-counter, and quanTIseq. Estimating the cell components is important to understand the distinct tumor immune microenvironment. Comparison of the cell types between the two groups showed that the significantly higher cell types (p < 0.05) in E-PPBC were M0 (CIBERSORT, xCell, MCP-counter), M1 (xCell, quanTIseq) and M2 (xCell) macrophages, and natural killer cells (xCell, EPIC). The proportion of immature dendritic cells (xCell) was significantly (p < 0.05) reduced in E-PPBC tumor samples. The distribution of M0 macrophages in E-PPBC and L-PPBC tumors is represented in Fig. 3A. Further, the proportion of tumor-associated macrophages (TAM) between E-PPBC and L-PPBC tumors was analyzed using 36 gene TAM signatures [16] (Fig. 3B). The distribution of other cell types did not differ between the two groups.

Fig. 3figure 3

Distribution of immune cell type and immune signatures between early PPBC (E-PPBC) and late PPBC (L-PPBC) groups. A shows an increased proportion of M0 macrophages. B Increased tumor-associated macrophage (TAM) signature. C Increased T cell exhaustion (T cell exhaustion signature included PD-1, PDL1, PDL2, CTLA4, TIGIT, IDO1, IDO2, and TIM3 expression) in E-PPBC

The analysis of differential pathways showed that T cell activation, proliferation, and differentiation were significantly upregulated (Fig. 2B) in E-PPBC tumors. To assess the T cell functionality, we used the T cell exhaustion signature (PD-1, PDL1, PDL2, CTLA4, TIGIT, IDO1, IDO2, and TIM3) [17]. A comparison of the score distribution between the two groups showed E-PPBC tumors had significantly higher T cell exhaustion scores suggesting that these tumors are likely to have pro-tumorigenic microenvironment (Fig. 3C).

Alterations in key pathways involved in tumor progression in E-PPBC

We further examined important cancer hallmark pathways contributing to an aggressive disease, such as ECM remodeling, angiogenesis, epithelial-to-mesenchymal transition (EMT), hypoxia, and stem cell phenotype, using the pathway-specific genes from the molecular signature database. The DEGs mapped with the signature gene sets are mentioned in supplementary Table 3 (Additional file 1). We evaluated ECM remodeling between E-PPBC and L-PPBC groups. DEGs that mapped with selected ECM gene signatures included ECM proteins, ECM regulators, and secreted factors involved in remodeling. E-PPBC tumors were significantly enriched (p = 0.017) in the ECM remodeling signature (Fig. 4A).

Fig. 4figure 4

Bar diagrams showing key cancer hallmark pathways upregulated in early PPBC (E-PPBC) tumors compared to late PPBC (L-PPBC) tumors. A Extracellular remodeling. B Angiogenesis. C Epithelial-to-mesenchymal transition

Tumor growth and metastasis depend on angiogenesis. As mentioned in Fig. 2B, the regulation of angiogenesis was an upregulated process in E-PPBC by GSEA analysis. We further validated this biological process using the angiogenesis gene signature. There was significantly increased (p = 0.033) angiogenesis in E-PPBC compared to L-PPBC tumors (Fig. 4B). Additionally, we observed E-PPBC tumors were more mesenchymal, which was indicated by a significantly high EMT score (p = 0.036) (Fig. 4C). Further, we also checked expressions of hypoxia-related and stemness gene signatures, and no significant difference was noticed between the two groups.

Invasive nature of the E-PPBC tumors

On enrichment analysis, we identified genes annotated with the term “breast cancer ductal invasive” as being significantly (q = 3 × 10–40) upregulated, followed by “multicancer invasiveness signature” (q = 6.72 × 10–12). The top 10 up- and downregulated pathways enriched using GSEA are represented in Fig. 2C. Enriched terms are given in Additional file 2.

To derive invasion-specific gene signatures in E-PPBC that can predict patient prognosis, 183 leading-edge genes (those genes that “drive” the enrichment score in a GSEA analysis) of “breast cancer ductal invasive” from GSEA analysis were collected. Out of these, 34 upregulated genes overlapped with DEGs of E-PPBC vs L-PPBC. Further, we considered genes falling in the top 25 percentile to derive a smaller subset. Based on this analysis, we derived ten genes, among them nine genes mapped with a public database (METABRIC) (listed in supplementary Table 4, Additional file 1). The gene score was calculated based on the average expression of these nine genes across all samples. We categorized the tumors into high- and low-expression groups using a mean score. Examination of the relapse-free survival in external datasets (METABRIC) in women ≤ 45 years showed tumors with high invasive score had poor survival (mean survival time, high vs low overall survival, 122 vs 99 months, p = 0.046 and relapse-free survival, 117 vs 88 months, p = 0.023) (Fig. 5). These results further confirmed that invasive signature derived from E-PPBC is associated with poor prognosis.

Fig. 5figure 5

Validation of the invasive signature using the external dataset (METABRIC) by Kaplan–Meier survival analysis in BC patients aged ≤ 45 years. Association of high expression of derived invasive signature with poor outcome. A Overall survival. B Relapse-free survival

To check if the gene signature would vary by subtype of breast cancer, we examined the pattern of expression of these genes in tumors (n = 10). Further subgrouping was identified by the IHC classification and PAM50 subtypes. Supplementary Fig. S5A (Additional File 1) shows the distribution of the tumors within the subclasses. As seen in the figure, distribution of the genes was clearly different between E-PPBC and L-PPBC and not different between ER + and TNBC tumors. PAM50 subtypes were also distributed in both E-PPBC and L-PPBC tumors without any subtype specificity. We further explored a larger set of tumors, [METABRIC, less than ≤ 45 years (n = 248)] and examined the expression of nine chosen genes in these tumors. Tumors were divided into invasive signature high and low (based on median value of mean of 9 genes) due to the lack of information on LCB. Clear pattern emerged between gene signature high and low tumors (Supplementary Fig. S5B, Additional File 1), while no pattern was visible by the ER + /TNBC classes or PAM50 subtypes. These results indicate that invasive signature does not vary by the subtype of the tumors.

留言 (0)

沒有登入
gif