Molecular biological features of Nottingham histological Grade 3 breast cancers

Abstract

Introduction:

Cancer biology dominates the behavior and prognosis of a tumor. Although Nottingham histological grade is a subjective pathological determination, it has been accepted as a surrogate model for cancer biology. As such, histologic grade was incorporated to the latest 8^th edition of the AJCC Breast Cancer Staging. Herein, we hypothesized that Grade 3 breast cancers demonstrate aggressive molecular biological profiles, reflecting worse biology and possible underlying immunogenicity.

Methods:

Transcriptomic and clinical data were obtained from Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), and the findings were validated by TCGA breast cancer cohort and GSE25066.

Results:

2876 patients were analyzed in this study. Grade 3 tumors were more common in Estrogen receptor (ER) negative, advanced stage patients, and associated with HER2 and Basal subtypes by the PAM50 classifier as well as with increased MKI67 expression (all p<0.001). Disease-free survival (DFS) was significantly worse in Grade 3 tumors (all cohorts). Gene set enrichment analysis demonstrated that Grade 3 tumors significantly enriched with not only cell proliferation and cell cycle related gene sets, but also immune activity related gene sets. CIBERSORT confirmed that Grade 3 tumors were infiltrated with macrophage M1, follicular helper T cells, and activated NK cells (all p<0.001). Furthermore, Grade 3 tumors were associated with more diverse T cell receptor (p=0.001) and increased cytolytic activity (p<0.001). Lastly, major T cell exhaustion markers were significantly elevated in Grade 3 BCs (p<0.001).

Conclusion:

Grade 3 BCs demonstrated aggressive transcriptomic features with enhanced immunogenicity and elevated T cell exhaustion markers.

Introduction:

It has been repeatedly suggested that cancer biology dominates the behavior, progression, and prognosis of a tumor (1). Histological grade has been accepted as a surrogate model for cancer biology in many cancer types (2–4), although it is a rather subjective pathological determination. In breast cancer (BC), histological grade has been recognized as a strong risk factor for clinical outcomes, along with lymph node metastases as well as tumor size (5). Among various pathological scoring system in BCs, Nottingham grading system, or Elston-Ellis modification of Scarff-Bloom-Richardson grading system (6, 7), composed of three pathological findings; degree of tubular formation, nuclear pleomorphism, and mitotic count, is currently the most validated system and demonstrated least inter-observer variability (8). Further, higher Nottingham grade is associated with shorter survival and early recurrence, irrespective of tumor size, hormone receptor status, or lymph node metastasis status (3, 9). As such, in the latest 8^th edition of the American Joint Committee on Cancer (AJCC) Breast Cancer Staging, Nottingham histologic grade is incorporated to a significant component of the tumor staging (10).

In contrast to the aggressive clinical phenotypes, higher histological grade is considered as one of predictors for pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) in triple-negative BCs (TNBCs) and HER2 positive BCs (11, 12). Similarly, tumor-infiltrating lymphocytes (TILs) density has been considered as a good predictor for pCR after NAC in BCs (13, 14) as well as an independent prognostic factor in HER2 positive BCs (13). TILs are attracted to tumor microenvironment (TME) by abundant mutant cancer peptides or neoantigens (15). Although there have been no reported association between increased neoantigens and pathological grade in BCs, significantly increased number of CD8+ TILs were identified in TME of Grade 3 BCs (16), which potentially resulted from increased neoantigens in the high-grade tumors.

Although pathological evidence appears compelling to suggest that histological grade can predict tumor biology (3, 17), transcriptomic assessment of histological grade in association with biology using RNA-sequence (RNA-Seq) has rarely been studied. With recent wide availability of RNA-Seq due to advance in the Next Generation Sequencing (NGS) technique, molecular biological data has been utilized to assess tumor biology (18). Herein, we hypothesized that Nottingham histological Grade 3 BCs would demonstrate aggressive molecular biological profiles, reflecting worse biology, as well as possible simultaneous underlying immunogenicity.

Material and Methods:

Data acquisition from METABRIC, TCGA-BRCA, and GSE25066 cohorts

Clinicopathological and transcriptomic data of breast cancer patients were acquired from Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) (19) and The Cancer Genome Atlas (TCGA)- breast invasive carcinoma (BRCA) Pan-Cancer Atlas through cBio Cancer Genomics Portal (20) and Pan-Cancer Clinical Data Resource (21), as previously described (22–29). Transcriptomic read count data was obtained from Genome Data Commons portal of National Cancer Institute (https://cancergenome.nih.gov/) (30). The data were normalized using the widely accepted trimmed mean of M-values (TMM) method. METABRIC cohort contained 1904 patients, of which 1832 patients were identified to have gene expressions from RNA-Seq and clinicopathological data including histological grade. TCGA-BRCA cohort included 1083 patients, of which 573 patients were identified to have RNA-Seq and clinicopathological data. While the median observation period of METABRIC was 112.9 months (Inter-quartile range (IQR): 59– 181 months), TCGA-BRCA was 27.2 months (IQR: 17– 51 months). Additionally, in order to exam the association between the histological grades and the effect with neoadjuvant chemotherapy (NAC), we identified three cohorts (GSE25066, GSE22226, and GSE22358) through Gene Expression Omnibus datasets (31–34). Each cohort contained 471, 123, and 111 patients. GSE25066 contained survival information (only disease-free survival (DFS)) as well as transcriptomic data of the primary tumor. Since all cohorts are de-identified publicly accessible database, review at Institutional Review Board (IRB) was exempted.

Herein, as we have three large cohorts containing transcriptomic, survival, and pathological grade information, we used the METABRIC cohort for the initial analysis and the TCGA and GSE25066 cohorts for validation of the findings.

Pathological grade information of the TCGA-BRCA cohort

Although the METABRIC cohort contained pathological grade (Grade 1, 2, and 3), the TCGA database does not include specific pathological information, such as Nottingham histological grade or lymphovascular invasion (LVI). Hence, we utilized Text Information Extraction System (TIES) Cancer Research Network to access the pathology reports of the TCGA-BRCA cohort. TIES Cancer Research Network is a federated network that facilitates data and biospecimen sharing among member institutions (35). TIES client software was used to collect Nottingham histological grade, which was recorded in 573 of the 1065 patients. Nottingham histological grading score assigns a score of 1 to 3 for each parameter; degree of tubular formation, nuclear pleomorphism, and mitosis. The final histological grade is based on a sum of the individual scores of the three parameters: 3, 4, or 5 = Grade 1; 6 or 7 = Grade 2; and 8 or 9 = Grade 3 (3, 7). While METABRIC only recorded the final grade, TCGA (via TIES) included the final grade as well as the individual scores for all three parameters.

Analysis of gene expression data

Gene Set Enrichment Analysis (GSEA) was performed comparing Grade 3 vs. Grade 1+2 tumors, using the Hallmark gene sets (36) with the software provided by the Broad Institute (https://software.broadinstitute.org/gsea/index.jsp), as described before (23–25, 29).

Measurements of immune activities, such as relative fractions of different types of immune cells in TME or T cell receptor (TCR) diversity, were estimated from tumor gene expression with CIBERSORT, a bioinformatic algorithm using the TCGA-BRCA cohort (37). The counts of neoantigen load were represented as Insertion and deletion (Indel) mutation, which was collated from the Pan-Cancer Atlas study of Thorsson et al (37). Cytolytic activity score (CYT) was defined as the geometric mean of grandzyme A (GZMA) and Perforin 1 (PRF1) expression values in transcripts per million (TPM), as described previously (38, 39). Additionally, given METABRIC and GSE25066 transcriptomes were derived from the microarray, we used the geometric mean of GZMA and PRF1 expression values to estimate cytolytic activity in these two cohorts.

Statistical Analysis

Statistical analyses were performed using R software (version 3.6.1, http://www.r-project.org/). DFS was defined as the time between the date of diagnosis of a primary tumor and the date of diagnosis of a recurrent BC, disease-specific survival (DSS) as the time from date of diagnosis to the date of death by BC, and overall survival (OS) as the time from date of diagnosis to the date of death by any cause. Kaplan-Meier method with log-rank test was performed for survival analyses. While statistical comparisons for categorical variables were performed by Fisher’s exact test, continuous values were compared by one-way ANOVA test. A two-sided p value< 0.05 was considered statistically significant.

Results:

Nottingham histological Grade 3 tumors were associated with aggressive clinical features

In all cohorts with 2876 patients, Grade 1 tumors were 274 patients (9.5%), Grade 2 tumors 1185 patients (41.2%), and Grade 3 tumors 1417 patients (49.3%). We hypothesized that Grade 3 BCs would have aggressive clinical features. Indeed, negative Estrogen Receptor (ER) status and advanced clinical stage were more common in Grade 3 tumors (Fig.1; p<0.001). Similarly, HER2 and basal molecular subtypes by the PAM50 classifier were more prevalent in Grade 3 tumors (Fig.1; p<0.001). In addition, Grade 3 tumors have significantly higher expression of MKI67 (Fig.1; p<0.001), suggesting more proliferation abilities.

Fig. 1.

Patients with ER negative status and more advanced pathological AJCC Stage were more common in Grade 3 BCs in the all cohorts. Also, Grade 3 BCs were more prevalent in Basal, HER2, and Luminal B subtypes, compared to Luminal A subtype on PAM50 classification in the all cohorts (all p<0.001). Furthermore, Grade 3 BCs demonstrated higher gene expression of MKI67 (all p<0.001).

ER, estrogen receptor; AJCC, American Joint Committee for Cancer

Grade 3 tumors were associated with worse clinical outcomes

We then further investigated if these aggressive clinical characteristics reflected on patient outcomes. Indeed, DFS was significantly shorter in the higher-grade tumors in all cohorts (Fig.2; median DFS time (METABRIC cohort): Grade 1/Grade 2/Grade 3= 22.1/20.5/19.4 months; p<0.001). Further, DSS and OS were significantly worse in the higher-grade tumors in the METABRIC cohort (Supplementary Fig.1; p<0.001, respectively). Hence, we demonstrated that Grade 3 BCs possess more aggressive biology compared to the lower-grade tumors.

Fig. 2.

Kaplan-Meier curves depicting DFS by each grade. DFS was significantly shorter in the higher-grade tumors compared to Grade 1 in all cohorts (p<0.001, p=0.028, and p=0.013, respectively).

DFS, disease free survival

Furthermore, Grade 3 tumors demonstrated significantly higher pathological complete response (pCR) rate in response to NAC in the GSE25066 cohort (Fig.3; p<0.001), which agrees with the previous report (12). Although the other two cohorts did not demonstrate significant difference likely due to the small sample size, there were trends toward higher pCR rate in Grade 3 tumors as well (Fig.3; p=0.06 and 0.09).

Fig. 3.

Nottingham Grade 3 BCs demonstrated significantly higher pCR rate in response to NAC in the GSE25066 cohort (p<0.001). Other two cohorts (GSE22226 and GSE22358) also demonstrated trends towards higher pCR rate in Grade 3 tumors (p=0.06 and 0.09, respectively).

BC, Breast cancer; pCR, pathological complete response; NAC, neoadjuvant chemotherapy

Gene Set Enrichment Analysis (GSEA) revealed that Grade 3 BCs enriched gene sets related to cell proliferation and cell cycles as well as immune activity

Given worse clinical outcomes associated with Grade 3 tumors, we hypothesized that they have aggressive transcriptomic features as well. We utilized GSEA to exam our hypothesis. Indeed, GSEA revealed that Grade 3 tumors enriched cell proliferation and cell cycle related gene sets; E2F targets (Normalized enrichment score (NES)= 2.54, p<0.001), G2M checkpoint (NES= 2.47, p<0.001), MYC Targets v1/2 (NES= 2.19/2.06, p<0.001), and mTORc1 signaling (NES= 2.10, p<0.001) (Fig.4). Above finding was validated in the TCGA and GSE25066 cohorts as well. Interestingly, Grade 3 tumors also enriched immune activity related gene sets simultaneously, such as Allograft rejection (NES= 2.33, p<0.001), Interferon gamma response (NES= 2.18, p<0.001), Interferon alpha response (NES= 2.06, p<0.001), and Inflammatory response (NES= 1.80, p<0.001) (Fig.4). GSEA results demonstrated that Grade 3 BCs were associated with aggressive transcriptomic features reflecting aggressive clinical outcomes. Additionally, Grade 3 tumors were associated with enhanced immune activities as well.

Fig. 4.

Enrichment plots of Grade 3 BCs showing association with cell proliferation/ cell cycle related gene sets by GSEA compared to Grade 1+2 BCs in the all cohorts. Grade 3 tumors were associated with E2F targets, G2M checkpoint, mTORc1 signaling and MYC targets v1 and v2 (all p<0.001) in the all cohorts. Additional enrichment plots of Grade 3 BCs showing association with immune activity related gene sets by GSEA compared to Grade 1+2 BCs in the METABRIC cohort. Grade 3 tumors were associated with Allograft rejection, INF gamma response, INF alpha response and Inflammatory response (all p<0.001). Similar trends were observed in the TCGA and GSE25066 cohorts.

BC, Breast cancer; GSEA, gene sets enrichment analysis; METABRIC, Molecular Taxonomy of Breast Cancer International Consortium; INF, Interferon; TCGA, the Cancer Genome Atlas; ES, Enrichment score

Grade 3 tumors were infiltrated with more anti-cancer immune cells in their TME

With enhanced immune activities with Grade 3 BCs on GSEA, we further hypothesized that higher number of anti-cancer immune cells would infiltrate in the TME in Grade 3 BCs. In order to test our hypothesis, immune cell composition in the TME was estimated using CIBERSORT. CIBERSORT revealed that the number of anti-cancer macrophage M1, Follicular Helper T cells, and activated NK cells significantly infiltrated in the TME of Grade 3 tumors (Fig.5; all p<0.001). CD8+ T cells were not significantly different based on the tumor grades (Fig.5; p=0.25). This finding was similar in the TCGA and GSE25066 cohorts although NK cells and CD8+ T cells infiltration results were not consistent. Anti-cancer immune cells were likely attracted to the TME due to higher neoantigen load in Grade 3 tumors (Fig.5; p=0.02). Given this higher anti-cancer immune cell composition in the TME, Grade 3 tumors were associated with more diverse T cell receptor (TCR) (TCR Shannon; Fig.5; p<0.001). Similarly, cytolytic activity score (CYT), a transcriptomic marker for anti-cancer immunity, was also significantly elevated in Grade 3 tumors (Fig.5; p<0.001).

Fig. 5.

CIBERSORT demonstrated that high infiltration of Macrophage M1 (p<0.001) and Follicular Helper T cells (p<0.001) in Grade 3 tumors in all cohorts. Infiltration of NK cells was inconsistent between the cohorts although CD8+ T cells were not significantly elevated in Grade 3 tumors. Grade 3 tumors were associated with significantly more TCR diversity (p<0.001), reflecting higher neoantigen load in Grade 3 tumors (p=0.02) in the TCGA cohort. Further, CYT was also significantly elevated in Grade 3 tumors in all cohorts.

NK cells, natural killer cells; TCR, T cell receptors; TCGA, the Cancer Genome Atlas; CYT, cytolytic activity score

As demonstrated in Fig.1, Grade 3 tumors include approximately 20–30% of TNBCs or basal molecular subtype tumors, which are typically immunogenic. To ensure no confounding factors, we performed subgroup analyses among the patients with ER positive BCs (Supplementary Fig.S2). We found that Grade 3 tumors in ER positive BCs demonstrated very similar findings on multiple analyses (survival analysis (DFS, DSS, OS), MKI67 expression, TCR diversity and CYT); thus, Grade 3 BCs possessed aggressiveness as well as immunogenicity regardless of hormone receptor or molecular subtype status.

Grade 3 tumors demonstrated significantly higher gene expression of major T cell exhaustion markers

Lastly, we hypothesized that Grade 3 BCs would express major T cell exhaustion markers to counterbalance the enhanced immunogenicity. Indeed, gene expression of major T cell exhaustion markers; programmed death ligand 1 (PD-L1), programmed death ligand 2 (PD-L2), programmed cell death 1 (PD-1), cytotoxic T-lymphocyte-associated antigen 4 (CTLA4), lymphocyte activating 3 (LAG3), indoleamine 2,3-dioxygenase 1 (IDO1), indoleamine 2,3-dioxygenase 2 (IDO2) and T cell immunoglobulin and ITIM domain (TIGIT) were significantly elevated in Grade 3 tumors (Fig.6). This increased expression was most likely a part of negative feedback loop to enhanced immunity in Grade 3 tumors. Similar trends were confirmed in the validation cohorts, although not all marker expressions were recorded in the GSE25066 cohort (Fig.6).

Fig. 6.

Grade 3 BCs were associated with significantly higher expression of T cell exhaustion markers, including PD-L1, PD-L2, PD-1, CTLA4, LAG3, IDO1, IDO2, and TIGIT in the METABRIC cohort. Similar trends were observed in the TCGA and GSE25066 cohorts as well.

BC, Breast cancer; PD-L1, programmed death ligand 1; PD-L2, programmed death ligand 2; PD-1, programmed cell death 1; cytotoxic T-lymphocyte-associated antigen 4, CTLA4; LAG3, lymphocyte activating 3; IDO1, indoleamine 2,3-dioxygenase 1; IDO2, indoleamine 2,3-dioxygenase 2; T cell immunoglobulin and ITIM domain, TIGIT; METABRIC, Molecular Taxonomy of Breast Cancer International Consortium; TCGA, the Cancer Genome Atlas

Discussions:

In the present study, we investigated if Grade 3 BCs possess molecular biologically aggressive features and possible underlying immunogenicity in approximately 3000 patients, using three different cohorts. Indeed, Grade 3 tumors demonstrated aggressive clinical characteristics, such as hormone receptor negative status, triple negative tumors, higher pathological stage, and elevated MKI67 gene expression, resulting in worse clinical outcomes, which echo the previous reports (3, 9). We found that Grade 3 BCs were strongly associated with aggressive molecular biological features, such as cell proliferation and cell cycle related gene sets by GSEA. Interestingly, Grade 3 BCs simultaneously enriched immune activity related gene sets on GSEA. This strong immunogenicity was also demonstrated by CIBERSORT, showing increased infiltration of anti-cancer macrophage M1, activated NK cells, and follicular helper T cells as well. Increased TCR diversity and elevated CYT further support strong anti-cancer immunity associated with Grade 3 tumors, likely due to increased neoantigens in the TME. T cell exhaustion markers were also significantly highly expressed in Grade 3 tumors as a negative feedback loop. Collectively, these findings suggest that Grade 3 BCs have aggressive biological features while they demonstrate strong immunogenicity and increased T cell exhaustion markers.

In the last decade, genomic signature profiling, such as Oncotype DX and MammaPrint, has been incorporated into the clinical practice of breast cancer to assess the future recurrence risk and benefit of systemic adjuvant chemotherapy (40, 41). With gene expression profiling methods, there also have been several studies to assess histological grade using genomic signatures (42, 43). Ivshina et al. stratified the patients with Grade 2 BCs into two groups (G2a and G2b), using a genetic grade signature with 232 genes, which correlated with significantly improved DFS prediction (42). Similarly, Sotiriou and colleagues established the genomic grade index (GGI), using 97- genomic signature identified from microarray technology (43). GGI was found to provide significant improved prognostic information than lymph node status or tumor size in ER-positive BCs (44). However, there are also some evidence to suggest genomic signature have limited contribution to patient care in comparison to the other clinicopathological parameters such as histological grade, tumor size, and lymph node status (45, 46). Here, with further advance in technology in RNA-Seq and computational bioinformatics methodology, we found more detailed association between Grade 3 BCs with not only enhanced cell proliferation and cell cycles but also enhanced immunity in tumor microenvironment.

Higher histological grade has been a well-known predictor for the response to NAC in TNBCs and HER2 positive BCs (11, 12). This is likely due to higher proliferation rate (12) as well as increased TILs associated with Grade 3 BCs (13). As demonstrated in the present study, TILs are attracted to TME by increased neoantigens (15, 47), which resulted in higher immunogenicity associated with Grade 3 BCs. Furthermore, it is speculated that enhanced immunity was likely due to increased anti-cancer M1 Macrophage, NK cells, and follicular helper T cells, which contributed to higher CYT in Grade 3 BCs. Additionally, as a negative feedback loop to enhanced immunity, Grade 3 BCs demonstrated significantly elevated T cell exhaustion markers. As shown in ImPassion trial and recent KEYNOTE 522 (48, 49), patients with TNBCs have revealed significant benefits from immune checkpoint inhibitors (ICIs) in addition to conventional systemic chemotherapies. Collectively, it is speculated that modifying TME in Grade 3 BCs with ICIs may be efficacious in this subset of patient populations. Further, although Nottingham grade has been accepted as a surrogate model for tumor biology, with further utilization of RNA-Seq, it might be possible to identify specific groups with even stronger immunogenicity among Grade 3 BCs.

The strength of this study is the large sample size with two validation cohorts. The TCGA-BRCA cohort does not include the detailed pathological reports, such as histological grade in the clinicopathological data. However, with TIES, we were able to manually extract the histological grade and its three individual scores from the original pathological reports, and link to the survival and transcriptomic data in TCGA.

There are a few limitations in the present study. First, this study was based on the transcriptome of the solely surgically resected primary tumor; thus, our results do not reflect on the histological grade in the metastatic sites. Second, neither METABRIC nor TCGA-BRCA cohorts contain patients who underwent NAC, hence patients with advanced disease at initial presentation might have been excluded from our analysis. However, this limitation was mitigated by GSE25066, which was consisted of patients who underwent NAC. Lastly, this study does not include any in-vitro or in-vivo experiments; therefore, all our findings are based on exclusively association. In order to further investigate the effect of Nottingham Grade 3 breast cancer, the experimental approach will be necessary.

Conclusion

In conclusion, Nottingham Grade 3 BCs demonstrated distinct transcriptomic features, reflecting their worse biology. Despite their clinical and transcriptomic aggressiveness, Grade 3 tumors were also associated with significantly enhanced anti-cancer immunity. Given strong immunogenicity associated with Grade 3 BCs, it might be possible to counterbalance their aggressiveness by modifying tumor immune microenvironment by immune checkpoint inhibition.

Supplementary Material

1

Supplementary Fig.S1 Kaplan-Meier curves depicting DSS and OS by each grade. DSS was significantly shorter in the higher-grade tumors compared to Grade 1 in all cohorts (p<0.001 and p=0.047). OS was significantly shorter in the higher-grade tumors compared to Grade 1 in the METABRIC cohort but not in the TCGA cohort (p<0.001, and p=0.293).

DSS, disease specific survival; OS, overall survival; METABRIC, Molecular Taxonomy of Breast Cancer International Consortium; TCGA, the Cancer Genome Atlas

Supplementary Fig.S2 Subgroup analysis with ER-positive patients using the METABRIC cohort (a) Kaplan-Meier curves depicting DFS, DSS and OS by each grade among the patients. DFS, DSS and OS were significantly shorter in the higher-grade tumors compared to Grade 1 in the METABRIC cohort (all p<0.001). (b) Grade 3 BCs demonstrated higher gene expression of MKI67, increased TCR diversity, and increased CYT (all p<0.001). (c) Enrichment plots of Grade 3 BCs showing association with cell proliferation/ cell cycle related gene sets by GSEA compared to Grade 1+2 BCs. Grade 3 tumors were associated with E2F targets, G2M checkpoint, mTORc1 signaling and MYC targets v1 and v2.

ER, estrogen receptor; BC, breast cancer; METABRIC, Molecular Taxonomy of Breast Cancer International Consortium; DFS, disease free survival; DSS, disease specific survival; OS, overall survival; TCR, T cell receptor; CYT, cytolytic activity score; GSEA, gene set enrichment analysis

Synopsis:

Pathological grade has been used as a surrogate model for tumor biology. With transcriptomic analysis from three large cohorts, we found that Nottingham Grade 3 breast cancers demonstrated aggressive transcriptomic features as well as enhanced immunogenicity.