The genomic landscape of metastatic histologic special types of invasive breast cancer

Histologic special types of breast cancer (BC) account for ~20% of BCs. Large sequencing studies of metastatic BC have focused on invasive ductal carcinomas of no special type (IDC-NSTs). We sought to define the repertoire of somatic genetic alterations of metastatic histologic special types of BC. We reanalyzed targeted capture sequencing data of 309 special types of BC, including metastatic and primary invasive lobular carcinomas (ILCs; n = 132 and n = 127, respectively), mixed mucinous (n = 5 metastatic and n = 14 primary), micropapillary (n = 12 metastatic and n = 8 primary), and metaplastic BCs (n = 6 metastatic and n = 5 primary), and compared metastatic histologic special types of BC to metastatic IDC-NSTs matched according to clinicopathologic characteristics and to primary special type BCs. The genomic profiles of metastatic and primary special types of BC were similar. Important differences, however, were noted: metastatic ILCs harbored a higher frequency of genetic alterations in TP53, ESR1, FAT1, RFWD2, and NF1 than primary ILCs, and in CDH1, PIK3CA, ERBB2, TBX3, NCOR1, and RFWD2 than metastatic IDC-NSTs. Metastatic ILCs displayed a higher mutational burden, and more frequently dominant APOBEC mutational signatures than primary ILCs and matched metastatic IDC-NSTs. ESR1 and NCOR mutations were frequently detected in metastatic mixed mucinous BCs, whereas PIK3CA and TP53 were the most frequently altered genes in metastatic micropapillary and metaplastic BCs, respectively. Taken together, primary and metastatic BCs histologic special types have remarkably similar repertoires of somatic genetic alterations. Metastatic ILCs more frequently harbor APOBEC mutational signatures than primary ILCs and metastatic IDC-NSTs.


INTRODUCTION
Breast cancer (BC) is heterogeneous and comprises various entities with divergent phenotype, biology, and clinical presentation 1,2 . There are over 20 histologic special types of BC recognized by The World Health Organization (WHO), accounting for~20% of all BCs 3 . Large sequencing studies have focused on invasive ductal carcinoma of no special type (IDC-NSTs), the most common histologic form of BC [4][5][6][7][8][9] , and data on the genomic landscape of histologic special types, particularly in the metastatic setting, are scarce. These studies have shown that although the repertoire of somatic genetic alterations found in metastatic BCs is remarkably similar to that of primary tumors, TP53, ESR1, ARID1A, ERBB2, GATA3, KMT2C, NCOR1, NF1, and RB1 have been found to be significantly more frequently mutated in metastatic disease 7,10,11 . In addition, estrogen receptor (ER)-positive metastatic BCs have been shown to more frequently display the APOBEC mutagenesis and homologous recombination DNA repair deficiency (HRD) processes than primary ER-positive disease 7,10 .
Here, through the reanalysis of targeted sequencing data generated with an FDA-approved multigene sequencing assay 11 , we sought to define the repertoire of somatic genetic alterations of metastatic ILCs, mixed mucinous, micropapillary, and metaplastic BCs, and determine whether the landscape of somatic mutations and copy number alterations (CNAs) of metastatic special types of BC is distinct from that of their primary counterparts or of metastatic IDC-NSTs.

Repertoire of somatic genetic alterations in primary and metastatic metaplastic BCs
The metaplastic BCs analyzed in this study were histologically heterogeneous. Out of the six metastatic metaplastic BCs, three displayed predominantly chondroid and three predominantly squamous differentiation, whereas the five primary metaplastic BCs exhibited predominantly squamous (n = 2), chondroid (n = 2), or spindle (n = 1) differentiation (Fig. 4a). The non-synonymous TMB of metastatic metaplastic BCs (median = 2.1, 95% CI = 0.9-4.8) was significantly lower than that of metastatic IDC-NSTs matched by age, menopausal status, and ER/HER2 status (median = 4.6, 95% CI = 0.8-17.7; P = 4 × 10 −2 , Mann-Whitney U test), but comparable to that of primary metaplastic BCs (median = 2.5, 95% CI = 1-3.2; P = 1.0; Mann-Whitney U test; Fig. 4b). TP53 was the most frequently altered gene in metastatic (83%) and primary (100%) metaplastic BCs. Genes altered in more than one sample in metastatic metaplastic BCs, and not altered in primary metaplastic BCs, included CCND1 and SOX9 (33% in metastatic vs 0% in primary, each, Fig. 4c). Akin to primary metaplastic BCs 22 , metastatic metaplastic BCs harbored mutations affecting Notch pathway genes, such as FBXW7, and PI3K/AKT pathway genes, such as PTEN and PIK3R1 (Fig. 4c). No statistically significant differences were observed in metastatic metaplastic BCs compared to primary metaplastic BCs or to metastatic IDC-NSTs matched by clinical features (Fig. 4c), potentially due to the small sample size of metaplastic BCs included in the study. We have previously reported that PIK3CA mutations are enriched in primary metaplastic BCs with predominant squamous or spindle cell differentiation, and absent or remarkably rare in primary metaplastic BCs with chondroid differentiation 22 . In line with these findings, we identified one primary metaplastic BC with squamous differentiation harboring a PIK3CA C420R hotspot mutation, whereas the primary chondroid metaplastic BCs studied here were PIK3CA wild type. In contrast, we observed PIK3CA mutations in metastatic metaplastic BCs with squamous (1/3) or chondroid (2/3) differentiation (Fig. 4b). Taken together, these findings suggest that PIK3CA mutations, albeit rare in primary metaplastic BCs with chondroid differentiation, may occur a subset of these tumors in the metastatic setting. Due to the limited number of metaplastic BCs studied here, these findings should be considered hypothesis generating.

Comparative analysis of gene CNAs between special types of BC and IDC-NSTs
We observed no significant differences in the frequency of gains/ losses and amplifications/homozygous deletions in metastatic ILCs, mixed mucinous, micropapillary, and metaplastic BCs, when compared to primary tumors of their respective histologic type, or to metastatic IDC-NSTs matched by clinicopathologic characteristics ( Supplementary Fig. 3a, b, e-j). Nonetheless, the fraction of the genome altered (FGA) of metastatic ILCs was found to be significantly higher than that of primary ILCs (P = 9.7 × 10 −3 , Mann-Whitney U test; Supplementary Fig. 4a). Compared to metastatic IDC-NSTs matched by clinicopathologic characteristics, however, metastatic ILCs displayed a significantly lower FGA (P = 2.5 × 10 −5 , Mann-Whitney U test; Supplementary Fig. 4a). Despite a lower FGA in the cohort of combined primary and metastatic ILCs compared to the combined primary and metastatic IDC-NSTs (P = 4.8 × 10 −9 ), no differences in the frequency of amplifications and homozygous deletions between the two groups were identified ( Supplementary Figs 3d and 4b). Nonetheless, we observed that the combined cohort of primary and metastatic ILCs   harbored a higher frequency of 1q gains and 16q losses than combined primary and metastatic IDC-NSTs matched by clinicopathologic characteristics and sample type ( Supplementary Fig.  3c). No statistically significant differences were observed in the FGA of metastatic mixed mucinous, micropapillary, or metaplastic BCs compared to their primary counterparts, or to age-, menopausal status-, and ER/HER2 status-matched metastatic IDC-NSTs ( Supplementary Fig. 4c-e), potentially due to the small sample size of special types of BC other than ILCs analyzed in the study.

Mutational signatures in special types of BC
There is evidence to suggest that the mutational processes underpinning metastatic BCs may differ from those of primary BCs 7,10 . Hence, we sought to determine whether the mutational signatures of metastatic forms of special histologic subtypes of BC would differ from those of primary tumors and from common forms of BC. We inferred the dominant mutational signatures using SigMA 34 , using all synonymous and non-synonymous somatic mutations in those cases with at least five single nucleotide variants (SNVs) for an accurate signature inference (n = 202), as previously described 35 . We observed a significant enrichment for APOBEC mutational signatures 2 and 13 in metastatic ILCs (51%, 53/103) compared to primary ILCs (35%, 25/72, P = 3.2 × 10 −2 ; Fisher's exact test), and to metastatic IDC-NSTs matched by clinicopathologic features (28%, 47/170; P = 9.7 × 10 −5 , Fisher's exact test; Fig. 5a and Supplementary Fig. 5a).

DISCUSSION
Here, through the reanalysis of targeted sequencing data of primary and metastatic forms of histologic special types of BC, we have demonstrated that the repertoire of somatic genetic alterations in metastatic forms of histologic special subtypes of BC is generally similar to that of their primary counterparts. Notable differences were, however, were observed, such as an enrichment for genetic alterations affecting ESR1, mainly as hotspot mutations, in metastatic ILCs and metastatic mixed mucinous carcinomas. We also observed a higher frequency of ERBB2 mutations in metastatic ILCs compared to primary ILCs, in agreement with previous studies of metastatic and relapsed ILCs 37,38 , and compared to age, menopausal status, and ER/HER2 status-matched metastatic IDC-NSTs. In addition, the spectrum of ERBB2 mutations differed between metastatic ILCs and metastatic IDC-NSTs matched by clinical characteristics, given that 36% of ERBB2 mutations targeted the L755 hotspot locus in metastatic ILCs, whereas this mutation accounted for only 17% in ERBB2 mutations in metastatic IDC-NSTs. The basis for the apparent enrichment for L755 ERBB2 mutations in metastatic ILCs warrants further investigation. The L755S mutation, however, has been shown to confer resistance to the tyrosine kinase inhibitor lapatinib, but not to the irreversible inhibitor neratinib 39,40 . In addition to the known ERRB2 hotspot mutations, we detected the ERRB2 X1097 splice mutation, whose biological impact and clinical significance remain to be determined. In agreement with previous studies 39,41 , all but two (89%) ERBB2 mutated metastatic ILCs were HER2-negative by immunohistochemistry and/or fluorescence in situ hybridization (FISH), highlighting the need of molecularly stratified clinical trials in the metastatic setting.
The enrichment for genetic alterations affecting TP53 and RFWD2, a ubiquitin ligase that targets p53 for degradation 42 , observed in metastatic ILCs compared to primary tumors might be reflective of the advanced stage of these patients. FAT1, a tumor suppressor that confers resistance to CDK4/6 inhibitors when inactivated 43 , was also found to be altered more frequently in metastatic ILCs. In addition, NF1 genetic alterations were more frequent in metastatic ILCs than in primary ILCs, in agreement with Sokol et al. 37 , who reported on the presence of genetic alterations targeting NF1 arising in the setting of relapse on endocrine therapy, indicating that these alterations likely constitute a mechanism of endocrine resistance 37 . In our study, although the number of NF1-mutant metastatic ILCs was insufficient for a formal mutual exclusivity analysis with ESR1 ligand-binding domain mutations, we observed that none of the metastatic ILCs with NF1 biallelic inactivation harbored ESR1 mutations. These findings provide further evidence supporting the notion that NF1 mutations may constitute a mechanism of resistance to endocrine therapy 37 .
We observed a higher mutational burden in metastatic ILCs than in primary ILCs and in age-, menopausal status-, and ER/HER2 status-matched metastatic IDC-NSTs, in agreement with the study by Sokol et al. 37 . Consistent with previous studies reporting that metastatic ER-positive BCs in general have an enrichment for the APOBEC mutagenesis process 7,10 , we observed an enrichment in APOBEC mutational signatures in metastatic ILCs as compared to primary ILCs and metastatic IDC-NSTs matched to the metastatic ILCs according to clinicopathologic features. Of note, APOBEC genes are not part of the MSK-IMPACT panel, and genetic alterations affecting APOBEC genes were not investigated. APOBEC processes have been implicated in tumor hypermutation 44 and likely play a role in resistance to endocrine therapy. Further studies to determine the role of APOBEC signatures in the clinical behavior of metastatic ILCs and the potential utility of the detection of APOBEC mutagenesis, as a biomarker of resistance to endocrine therapy are warranted.
Our study has limitations, including the small sample size of the metastatic mixed mucinous, micropapillary, and metaplastic BCs, owing to their rarity, which may limit the identification of statistically significant differences in the comparisons performed. Hence, the negative conclusions related to these histologic special types need to be interpreted with caution, as we cannot rule out type II or β errors. Furthermore, our study is based on the reanalysis of targeted sequencing data, and we cannot rule out differences between primary and metastatic special histologic types of BC outside of the genes captured by MSK-IMPACT. Hence, whole-exome and/or whole-genome analyses of metastatic special types of BCs are warranted. Moreover, the primary and metastatic special histologic subtypes of BC were not matched lesions from the same patients.
Notwithstanding these limitations, our study indicates that the repertoire of genetic alterations in primary and metastatic forms of special histologic types of BC is remarkably similar; however, key differences exist, such as higher mutational burden and an enrichment for the APOBEC mutational processes in metastatic ILCs. Our findings also suggest that ERBB2 and ESR1 mutations should be considered as potential mechanisms of resistance to endocrine therapy and druggable targets in clinically HER2negative metastatic ILCs.

Cases and study population
The study was approved by Memorial Sloan Kettering Cancer Center Institutional Review Board as part of the project whose findings were initially published by Razavi et al. 11 . Informed consent was provided in the original study by Razavi et al. 11 . Targeted massively parallel sequencing data of primary and metastatic BCs were obtained from the study by Razavi et al. 11 . All cases had been previously subjected to targeted capture massively parallel sequencing using the MSK-IMPACT sequencing assay from the study by Razavi et al. 11 (Supplementary Table 2). Following the criteria put forward by the WHO 3 , 309 BCs were classified as of one of the special histologic types included in this study: 259 were classified as classic ILCs (n = 127 metastatic and n = 132 primary), 19 as mixed (i.e., >50% but <90% mucinous component) mucinous carcinomas (n = 5 metastatic and n = 14 primary), 20 as pure micropapillary carcinomas (n = 12 metastatic and n = 8 primary), and 11 as metaplastic BCs (n = 6 metastatic and n = 5 primary, Supplementary Table 2). The initial diagnosis of a given special histologic type of BC was retrieved from Razavi et al. 11 , and cases for which the histologic material of the sample subjected to sequencing was available (n = 265) were reviewed centrally by a board-certified breast pathologist (F.P.) for diagnosis confirmation. Pleomorphic ILCs (metastatic, n = 6; primary, n = 8) were excluded from further analyses. ER and HER2 status had been assessed by immunohistochemistry and/or FISH, as previously described 11 , following the American Society of Clinical Oncology/College of American Pathologists guidelines 45,46 . Comparison with common forms of breast cancer For the comparison of non-synonymous TMB, FGA, frequency of nonsynonymous somatic mutations, and CNAs, metastatic BCs of special histologic subtype were compared to those of IDC-NSTs included in the same study 11 , matched by age (20-year intervals), menopausal status, and ER/HER2 status and to those of their primary counterparts. Metastatic ILCs were matched to metastatic IDC-NSTs from the study by Razavi et al. 11 previously subjected to MSK-IMPACT at a 1:2 ratio, whereas mixed mucinous BCs, micropapillary BCs, and metaplastic BCs were matched to IDC-NSTs at a 1:3 ratio. No statistically significant differences were observed in the therapy received prior to tumor sampling between the metastatic BCs of special histologic types and metastatic IDC-NSTs matched by clinicopathologic characteristics in the cohorts analyzed in this study (Supplementary Table 6). Lollipop plots were produced using MutationMapper on cBioPortal 47 (http://www.cbioportal.org), manually curated and mutation types were color-coded as follows: splice-site SNV (yellow), missense SNV (green), truncating SNV (black), in-frame insertion/ deletion (brown), and hotspot mutation (orange).

Targeted massively parallel sequencing analysis
All samples included in this study were subjected to targeted sequencing using the FDA-approved MSK-IMPACT assay 48 , as part of the study by Razavi et al. 11 . Non-synonymous somatic mutations, amplifications, and homozygous deletions were retrieved from the original study 11 . The raw MSK-IMPACT sequencing data (i.e., FASTQ files) were reprocessed using our validated bioinformatics pipeline, as previously described 49,50 , for the inference of copy number gains and losses, and loss of heterozygosity of genes targeted by somatic mutations and mutational signatures. Mutations affecting hotspot codons were annotated as described. Non-synonymous TMB was calculated as the number of non-synonymous mutations divided by the total genomic region assessed by MSK-IMPACT, per megabase. The FGA, defined as the number of base pairs which are not copy neutral divided by the size of genome assayed, was retrieved from the original study by Razavi et al. 11 . Mutational signatures were defined using SigMA 34 using all synonymous and non-synonymous somatic mutations of cases with at least F. Pareja et al.

Assessment of TILs infiltration
Histologic assessment of TILs infiltration in primary and metastatic ILCs with a sufficient number of SNVs (≥5) for accurate assessment of mutational signatures by SigMA, and available hematoxylin and eosin (H&E) slides was performed. The assessment of TILs infiltration was conducted following the guidelines described by the International TIL working group 52 . In brief, following the examination of one representative section, the intratumoral stromal area covered by mononuclear cells, including lymphocytes and plasma cells, was recorded.

Statistical analysis
Statistical analyses were conducted using R v3.1.2. Fisher's exact tests were employed for comparisons between categorical variables, and Mann-Whitney U test were used for continuous variables. All tests were two-sided and P values < 0.05 were considered statistically significant. We performed multiple testing correction using the Benjamini-Hochberg procedure to control for the false discovery rate (q values; Supplementary Table 3). To assess the mutual exclusivity between ERBB2 and ESR1 mutations (hotspot mutations and non-hotspot pathogenic mutations) in ER-positive metastatic ILC and IDC-NST using CoMEt 53 .

Reporting summary
Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.