Introduction

Checkpoint kinase 2 (CHK2) is a serine/threonine protein kinase encoded by the CHEK2 gene which maintains genomic stability in the setting of DNA damage [1,2,3]. CHK2 is activated by the ATM kinase and in turn, interacts with BRCA1, BRCA2, p53, CDC25, and other effectors to initiate DNA repair, apoptosis, and cell cycle arrest [4,5,6]. Germline CHEK2 variants were first described in patients without germline TP53 mutations who were thought to meet clinical criteria for Li-Fraumeni syndrome [7, 8], although the association with Li-Fraumeni syndrome has since been negated [9]. Germline CHEK2 variants have now been identified in ~ 1% of the population [9], and pathogenic variants (PVs) are causally linked to multiple types of cancer, including colon cancer and breast cancer (BC), further highlighting the functional relevance of CHEK2 in safeguarding genome integrity [3] [10]. Germline CHEK2 PVs confer a lifetime BC risk ranging from 20 to 44%, depending on family history of BC [11,12,13]. Furthermore, individuals with the higher risk c.1100delC allele carry a BC risk of up to two times higher in women and ten times higher in men when compared to non-carriers [14,15,16]. Accordingly, clinical guidelines currently recommend enhanced BC screening in individuals with protein truncating CHEK2 variants [17].

Investigations into the phenotype and clinical behavior of CHEK2-associated BCs have been limited. CHEK2-associated BCs tend to occur in younger women, with higher rates of bilateral disease [12], and frequently express estrogen receptor (ER) and/or progesterone receptor (PR) [12, 13]. Interestingly, in contrast to BRCA1/2-related cancers, CHEK2-associated BCs appear to lack the genetic characteristics of homologous recombination DNA repair deficiency [18, 19]. Biallelic CHEK2 inactivation has been observed in some but not all CHEK2-associated BCs [19]. The precise mechanistic relationship between CHEK2 and BC development currently remains uncertain.

In order to characterize BC that develop in the setting of germline CHEK2 variants, we explored the clinicopathologic, histologic, and genetic features of BCs arising in a consecutive series of patients with pathogenic or likely pathogenic germline CHEK2 variants (P/LPVs) at our institution.

Methods

Study population

The institutional review board of the University of California San Francisco (UCSF) approved this study. Patients with P/LPVs in CHEK2 who underwent breast tissue sampling were retrospectively identified over a 15-year period (2008 to 2023) using a pathology database. All germline variants in this cohort (n = 35) with invasive BC were reviewed by a board-certified genetic counselor (AB) and classified as P/LPVs according to the ClinVar database [20].

Study design

All slides were reviewed by at least one breast pathologist (CJS, YYC, and/or GK). For primary BCs, ER, PR, and human epidermal growth factor receptor 2 (HER2) were evaluated and scored according to ASCO/CAP guidelines [21, 22]. Ki67 was scored according to recommendations of the International Ki67 in Breast Cancer Working Group [23]. Immunohistochemical surrogates of molecular subtype were classified based on St. Galen criteria as previously described using either 14% or 20% cut-offs for Ki-67 [24]. Tumor-infiltrating lymphocytes (TILs) were assessed based on recommendations by the TILs Working Group [25]. For neoadjuvant chemotherapy (NACT)-treated tumors, assessment of tumor cellularity and residual cancer burden (RCB) index were calculated according to methods previously described [26]. Clinicopathologic and outcome data were retrieved from the electronic medical record.

Targeted next-generation DNA sequencing

Tumor tissues were selected from primary BCs (n = 21), lymph node metastasis (n = 1), lung metastasis (n = 1), and liver metastasis (n = 1) for capture-based targeted next-generation DNA sequencing (NGS). Sequencing was performed at the UCSF Clinical Cancer Genomics Laboratory using an assay that targets all coding regions of ~ 500 cancer-related genes, select introns from 40 genes, and TERT promoter, and analyzed according to methods previously described [27].

Results

Study overview

The workflow for this study is depicted in Fig. 1. We identified 67 patients with CHEK2 P/LPVs who underwent core biopsy of the breast. Most patients (50/67, 75%) had intraductal epithelial atypia or in situ or invasive carcinoma on core biopsy. Of these, noninvasive disease included atypical ductal hyperplasia (4%, 3/67), lobular neoplasia (atypical lobular hyperplasia or lobular carcinoma in situ) (6%, 4/67), and ductal carcinoma in situ (12%, 8/67). Invasive BC was identified in 52% of patients (35/67), and these patients comprised the subsequent study population. In total, 44 invasive BCs were analyzed from these 35 patients. Most patients (88%, 15/17) without atypia or carcinoma on core biopsy were under active surveillance, with two patients opting for bilateral risk-reducing mastectomy (Fig. 1).

Fig. 1
figure 1

Schematic workflow for identifying breast cancers arising in individuals with CHEK2 pathogenic and likely pathogenic variants

Full size image

Indications for genetic testing and description of germline CHEK2 variants in patients with invasive breast cancer

Genetic testing followed the personal diagnosis of BC in 83% (29/35) of patients (Table 1). A smaller subset (11%, 4/35) underwent genetic testing after a germline CHEK2 variant was identified in a family member. Most patients (79%, 27/34) had a family history of BC, with 41% (14/34) occurring in a first degree relative (Table 1). Three patients had a previous history of BC prior to presentation at our institution (9%, 3/35).

Table 1 Indication for genetic testing and cancer history in patients with CHEK2-associated invasive breast cancer
Full size table

A summary of the germline CHEK2 P/LPVs in patients that developed invasive BC and associated clinical characteristics are detailed in Supplementary Table S1. CHEK2 P/LPVs consisted of frameshift (17/35, 49%), missense (26%, 9/35), and splice site (11%, 4/35) mutations, as well as exonic deletions (14%, 5/35) (Table 2). The cohort was enriched for CHEK2 c.1100delC (p.T367fs) variants (40%, 14/35), which is thought to portend the highest BC risk [14,15,16]. The second most common variant was c.1283C > T (p.S428F) (9%, 3/35), a low penetrance missense mutation thought to abrogate the kinase function of CHK2 [28]. Nearly all patients (33/35, 94%) were heterozygous CHEK2 carriers (Table 2). Two patients (patient 2, c.1100delC and patient 32, c.499G > A) were homozygous with a family history of BC in a first-degree relative. Five patients (14%) additionally harbored PVs in other genes associated with BC risk, including ATM (c.1027_1030del, p.E343Ifs and c.237del, p.Lys79fs), PALB2 (c.2827_2830del, p.E943Sfs), RAD50 (c.2517dupA, p.D480fs) and MUTYH (c.536A > G, p.Y179C) (Supplementary Table S1).

Table 2 Germline pathogenic/likely pathogenic variants in patients with invasive breast cancer
Full size table

Clinicopathologic characteristics of CHEK2-associated breast cancers

The clinicopathologic features of CHEK2-associated invasive BCs are shown in Table 3. All patients were women. The median age at diagnosis was 45 years (range 25–75, Table 3). Most presented with a palpable mass (54%, 19/35). The remainder had either a mammographically detected mass (29%, 10/35) or calcifications (17%, 6/35). Four patients had synchronous bilateral BCs (11%, 4/35) and seven had multifocal BC (20%, 7/35). The mean tumor size in non-treated cases was 1.3 cm (range < 0.1 to 5.4) (Table 3). Lymph node metastases were present in 38% of cases (13/34).

Table 3 Clinicopathologic characteristics of breast cancers arising in patients with CHEK2 pathogenic/likely pathogenic germline alterations
Full size table

BCs were predominantly invasive ductal carcinomas of no special type (IDC-NST) (86%, 38/44). Other histologic types were invasive lobular carcinoma (7%, 3/44, including solid [1], alveolar [1], and classic [1] patterns), invasive carcinoma with ductal and lobular features (5%, 2/44), and invasive papillary carcinoma (2%, 1/44). BCs were most often Nottingham grade 2 (57%, 25/44), with 25% (11/44) grade 3 and 18% (8/44) grade 1. TILs were sparse (< 10%) in most (89%, 34/38) cases and > 20% in only one case (3%, 1/34) (Supplemental Table S2).

Nearly all BCs were ER + (95%, 41/43), including 79% (34/43) ER + HER2- and 16% (7/43) ER + HER2 + . Both ER- tumors were HER2 + (5%, 2/43). Of the nine total HER2 + BCs, seven (78%) arose in a background of CHEK2 c.1100delC, and, conversely, 43% (6/14) of BCs arising in patients with CHEK2 c.1100delC were HER2 + . No patients developed triple negative BC. The Ki67 proliferation index was ≥ 30% in 32% (10/31), 6–29% in 58% (18/31), and ≤ 5% in 10% (3/31) of cases. Using immunohistochemical surrogates for molecular subtyping, most (69%, 25/36) BCs were classified as luminal B (50%, 18/36 Luminal B HER2-, 19%, 7/36 Luminal B HER2 + ; using ≥ 14% Ki-67 cutpoint) [24]. One-quarter (25%, 9/36) were luminal A, and only 6% (2/36) were HER2-enriched. Results were similar when using ≥ 20% Ki-67 cutpoint for Luminal B (67% luminal B) (Supplemental Table S2).

OncotypeDX scores were available for 12 patients and were variable, with most (83%) scores falling into the low (5/12) or intermediate categories (5/12). Of the two patients with high OncotypeDX scores (16%, 2/12), one developed lung metastasis. MammaPrint assay scores were available for 13 patients and stratified nearly equally into low-risk (46%, 6/13) and high-risk categories (54%, 7/13).

Next-generation DNA sequencing

Targeted next-generation DNA sequencing was performed on 23 BCs (24 samples) from 20 patients, including 21 primary BCs, one lymph node metastasis,one liver metastasis, and one lung metastasis. The results are shown in Fig. 2. Sequenced primary BCs included synchronous unilateral tumors from two patients and synchronous bilateral tumors from one patient. Most (92%, 21/23) sequenced tumors were IDC-NST. All were ER + , and 22% (5/23) were HER2 + by immunohistochemistry and/or fluorescence in situ hybridization, including a metachronous ER + HER2 + lung metastasis in Patient 11.

Fig. 2
figure 2

Genetic profiling of invasive breast cancers arising in patients with germline CHEK2 variants

Full size image

Biallelic CHEK2 alteration was detected in 13 of 23 BCs (57%), which was due to somatic loss of heterozygosity (LOH) of the germline CHEK2 allele in all cases. All sequenced BCs arising in patients with germline CHEK2 c.1100delC (n = 6) showed biallelic CHEK2 inactivation via LOH. In one case (patient 26), an additional somatic missense CHEK2 mutation (p.K373E) was identified in an invasive lobular carcinoma (ILC) arising in a 70-year-old germline CHEK2 exon 9 deletion carrier. Whether this second CHEK2 hit was in cis or trans with the germline allele could not be determined. A synchronous ipsilateral ILC in the same patient was heterozygous for the germline CHEK2 mutation with no additional CHEK2 alterations. The genetic profiles of the synchronous ILCs in this patient were also otherwise distinct, including different inactivating CDH1 mutations (p.T115fs and p.N315fs) among other alterations (Fig. 2).

Aside from biallelic alteration of CHEK2, recurrent alterations included inactivating GATA3 mutations (6/23, 26%), PIK3CA hotspot (p.H1047R) mutations or amplification (6/23, 26%), and amplifications of CCND1 (5/23, 22%), FGFR1 (5/23, 22%), ERBB2 (4/23, 17%), GAB2 (4/23, 17%), ZNF703 (4/23, 17%), GPR124 (4/23, 17%), PPM1D (2/23, 9%), and MDM2 (2/23, 9%). TP53 mutation were infrequent (2/23, 9%).

Invasive BCs from 3 of the 5 patients with concurrent PVs in additional BC risk genes were sequenced (Fig. 2). Patient 13 (CHEK2 c.1100delC and PALB2 c.2827_2830del) presented with a 2 cm Nottingham grade 3 ER + HER2- IDC-NST at age 25 (Oncotype 21). Tumor sequencing revealed LOH in CHEK2 and PALB2, as well as CCND1,FGFR1, GPR124, ZNF217, ZNF703, and PTPN1 amplifications. Patient 22 (CHEK2 c.1100delC and ATM c.237del) presented with a Nottingham grade 3 ER + HER2- IDC-NST at age 30 (MammaPrint high-risk). Tumor sequencing revealed LOH in CHEK2 and ATM, as well as CCND1 and GAB2 amplifications. This patient received neoadjuvant chemotherapy and was one of only three patients to achieve pCR. Patient 31 (CHEK2 c.707 T > C (p. Leu236Pro and RAD50 c.2517dupA, p. D480fs) presented at age 45 with synchronous bilateral IDC-NST. No additional somatic alterations were identified on sequencing of either tumor.

Treatment and Outcome

Treatment and outcomes of CHEK2-associated BCs are shown in Table 4. Most patients underwent mastectomy (69%, 24/35) compared to lumpectomy (31%, 11/35). Fourteen patients received NACT (40%, 14/35), and five patients received neoadjuvant hormonal therapy (NAHT) (14%, 5/35). In line with the high frequency of ER + BCs, nearly all (88%, 30/34) patients received adjuvant hormonal therapy. Twelve patients received adjuvant radiotherapy (35%), and 13 patients received adjuvant chemotherapy (38%).

Table 4 Treatment, follow-up and outcomes of patients with CHEK2-associated breast cancers
Full size table

Clinical follow-up was available for 34 patients (97%, 34/35) with a mean interval of 50 months. One patient (3%, 1/34) developed local recurrence, which was an axillary recurrence of ER + IDC-NST 72 months after bilateral mastectomy. Five patients (15%) developed distant metastases, with metastatic sites including liver (2), bone (2) and lung (1) (Supplementary Table S2). The majority of patients (85%, 29/34) had no evidence of disease at time of study, whereas 12% (4/34) are alive with disease. Only one patient (3%) died of disease following a poor response to NACT, liver metastasis at 58 months, and death at 117 months. Among the 11 patients who opted for breast conservation, none developed local recurrence during the follow-up period.

Response to neoadjuvant chemotherapy

Treatment responses to NACT of 18 CHEK2-associated BCs arising in 14 patients are summarized in Supplemental Table S3. All BCs (100%) showed a clinical reduction in tumor size based on tumor size at pathologic examination compared to pre-treatment magnetic resonance imaging (MRI). All BCs with available data showed a post-treatment reduction in Ki67 proliferation index compared to pre-treatment scores. By RCB calculation, most treated BCs were RCB-II (11/18, 61%) or RCB-I (4/18, 22%), with no RCB-III tumors. Pathologic complete response (pCR) was observed in only three patients (3/14, 21%), two of which had HER2 + BC. In cases with residual cancer, invasive tumor cellularity ranged from < 1% to 50% and was ≥ 30% in 50% (7/14) of tumors.

Discussion

In summary, CHEK2-associated invasive BCs tended to arise in younger women with a strong family history of BC and were often multifocal and bilateral. The tumors were mostly immune-poor ER + IDC-NST of predominantly luminal B subtype and were heterogeneous in terms of histologic grade, HER2 status, proliferation index, and multi-parametric molecular testing. NGS overall showed luminal-like genetics and biallelic CHEK2 alteration via LOH as an oncogenic driver in most cases. BCs arising in patients with concurrent germline PVs in the DNA damage genes PALB2 and ATM (but not RAD50) also showed LOH in these genes. Most patients with CHEK2-associated BCs had partial responses to NACT, with few achieving pCR.

Most but not all CHEK2-associated BCs showed biallelic CHEK2 alteration via LOH, which is in keeping with incomplete rates of LOH seen in prior studies [18, 19, 29, 30]. A recent study found high rates of CHEK2 LOH in BCs arising in c.1100delC carriers compared to lower rates in other CHEK2 variants [19]. Consistent with this, we found CHEK2 LOH in all BCs arising in c.1100delC carriers (6/6) but only in 47% (8/17) of BCs arising in other variants. Taken together, our findings and those of others raise consideration that at least some CHEK2 variants may function via haploinsufficiency in driving BC, whereas others (such as c.1100delC) may require bilallelic inactivation. This is in contrast to most other germline BC risk genes, such as BRCA1/2, PALB2, and ATM, PVs of which often show biallelic loss of function in BC [29]. Alternatively, CHK2 inactivation may also be due to non-genetic mechanisms in cases without LOH, and future studies may help address this issue.

Most BCs arising in our population of CHEK2 carriers were ER + IDC-NST, predominantly of luminal B subtype, with a minority of tumors being luminal A or HER2-enriched, and none being triple negative. These findings are consistent with previous reports showing a predominance of ER/luminal BC and HER2 overexpression in this setting [9, 12, 19, 31, 32]. Aside from biallelic CHEK2 inactivation in the majority of cases, the mutational repertoire of the sequenced luminal BCs in our series was otherwise similar to other luminal BCs and included recurrent aberrations in genes such as PIK3CA, GATA3, CCND1, FGFR1, and ERBB2, with a low frequency of TP53 mutations. Our results are thus overall consistent with those of previous studies [18, 19, 31, 33,34,35] and highlight CHEK2 as a dominant oncogenic driver in these tumors while also supporting a model in which CHEK2 may act as a facilitator or accelerator of tumorigenesis that is otherwise subtype-specific. This model of CHEK2-facilitated tumorigenesis was previously put forth by Massink et al. in their study describing luminal-like copy number profiles of BCs arising in CHEK2 c.1100delC carriers [18].

Concurrent germline P/LPVs in multiple hereditary BC risk genes (so-called double heterozygotes) are rare, and data regarding cancer risk and clinical implications are scarce. The majority of reported double heterozygotes in the literature include BRCA1 and/or BRCA2, with only few studies and case reports describing double heterozygosity exclusively involving other BC risk genes, such as CHEK2, ATM, PALB2, NBS1, and BLM [36,37,38,39,40,41,42,43,44]. Given the rarity of these events, conclusions regarding cancer risk, cancer phenotype, patient surveillance, and clinical management of these cases are limited. It has been suggested that BRCA1/2 and CHEK2 double heterozygotes do not have increased BC risk or a different BC phenotype beyond what is expected for BRCA1/2 variants [39, 40, 44]. Whether a similar principle holds for double heterozygotes involving CHEK2 and non-BRCA1/2 BC risk genes is not known due to paucity of data. Taken as a whole, BCs arising in double heterozygotes do not appear to present at younger age or with increased bilaterality compared to single heterozygotes [39]. Of the five double heterozygotes involving CHEK2 with ATM, PALB2, RAD50, or MUTYH in our study, we noted a younger age of BC presentation than the cohort median in three cases (CHEK2 c.433delC/MUTYH c.536A > G, age 34; CHEK2 c.1100delC/PALB2 c.2827_2830del, age 25; CHEK2 c.1100delC/ATM c.237del, age 29), and bilateral and multifocal tumors in only one patient (CHEK2 c.707 T > C/RAD50 c.2517dupA). We therefore find no compelling evidence of a synergistic effect compared to CHEK2 carriers alone, but note that the small number of such cases in our series significantly limits meaningful conclusions. There is similarly little data on BC risk and phenotype in CHEK2 homozygotes or compound heterozygotes, although available data suggests that these patients may have higher BC risk, present at younger age, and more often have a second BC diagnosis compared to heterozygotes [45,46,47]. Our study included two CHEK2 homozygotes (c.1100delC and c.499G > A), who presented with BC at 45 and 39 years of age, respectively, and have not developed additional BCs at time of study.

Ours is one of few studies assessing the response of CHEK2-associated BCs to NACT. Although decreased post-treatment tumor size and Ki67 proliferative index reflected treatment effect in all evaluable cases, only three patients in our series achieved pCR. Two tumors with pCR were HER2 + (Luminal B HER2 + and HER2 enriched). These results are not surprising, given the overall lower pCR rates of ER + /luminal tumors [48, 49] as well as the absence of robust homologous recombination DNA repair deficiency in CHEK2-associated BC [18, 19, 30]. Indeed, our results are similar to a prior study of poor NACT responses in eight CHEK2 patients with BC, in which responses to NACT were also found to be worse in CHEK2 carriers when compared to non-carriers [49]. Larger studies are required to confidently compare NACT sensitivities of CHEK2-associated ER + BCs to other BCs. We speculate that the absence of significant immune cell infiltration may also impact response rates to NACT.

Guidelines recommend annual mammography with consideration for magnetic resonance imaging starting at age 40 in patients with CHEK2 P/LPVs. However, unlike for patients with germline BRCA1 or BRCA2 alterations, there are currently no guidelines regarding risk-reducing mastectomy in CHEK2 carriers, and decision making is generally based on individualized and family risk [17]. There is a paucity of data evaluating recurrence rates after breast conservation versus unilateral or bilateral mastectomy in CHEK2 carriers with BC, although some studies have suggested a higher risk of contralateral BC [32, 50, 51]. In our study, only 11 patients with BC opted for breast conservation, and none experienced local recurrence. On the other hand, we found higher than expected rates of bilateral and multifocal tumors compared to the general population, which is consistent with prior studies [52] and has implications arguing against breast conservation. Aside from CHEK2 germline status and type of CHEK2 variant (i.e. c.1100delC vs other), patient age at presentation, family history, and patient preferences, should also be considered when planning surgical intervention and surveillance for these patients.