Disruption of Cyclin-Dependent Kinase 12 (CDK12) is known to lead to defects in DNA repair and sensitivity to platinum salts and PARP1/2 inhibitors. However, CDK12 has also been proposed as an oncogene in breast cancer. We therefore aimed to assess the frequency and distribution of CDK12 protein expression by IHC in independent cohorts of breast cancer and correlate this with outcome and genomic status. We found that 21% of primary unselected breast cancers were CDK12 high, and 10.5% were absent, by IHC. CDK12 positivity correlated with HER2 positivity but was not an independent predictor of breast cancer–specific survival taking HER2 status into account; however, absent CDK12 protein expression significantly correlated with a triple-negative phenotype. Interestingly, CDK12 protein absence was associated with reduced expression of a number of DDR proteins including ATR, Ku70/Ku80, PARP1, DNA-PK, and γH2AX, suggesting a novel mechanism of CDK12-associated DDR dysregulation in breast cancer. Our data suggest that diagnostic IHC quantification of CDK12 in breast cancer is feasible, with CDK12 absence possibly signifying defective DDR function. This may have important therapeutic implications, particularly for triple-negative breast cancers. Mol Cancer Ther; 17(1); 306–15. ©2017 AACR.

Breast cancer is a complex disease comprising a variety of molecular and clinically distinct subtypes. Substantial progress has been made in the management of breast cancer mortality over the last 25 years, in part due to improved treatment modalities such as endocrine therapies, HER2-targeted therapy, and combination chemotherapies (1–4). However, a proportion of sporadic primary breast cancer remain difficult to treat. Hence, there is an urgent need for stratification and biomarker discovery within this cohort.

The CycK/CDK12 (Cyclin K/Cyclin dependent kinase 12) complex is involved in the regulation of RNA polymerase II (RNA pol II) and mRNA processing (5–7) and is known to protect normal cells from genomic instability by regulating the transcription of DNA damage response (DDR) genes (8). Moreover, CDK12 has been postulated as a tumor suppressor gene in high-grade serous ovarian cancer (HGSOC), where it is one of the only significantly recurrently mutated genes (9). Recurrent point mutations have been shown to abrogate the functional activity of CDK12, resulting in defects in multiple DNA repair pathways, leading to genomic instability, downregulation of some homologous recombination (HR) genes such as BRCA1, FANCI, or FANCD2 (10, 11) and selective sensitivity to both platinum agents and PARP1/2 inhibitors (12, 13). Indeed, recent data in HGSOC suggests that CDK12-inactivated tumors have a unique signature of genomic instability characterized by frequent mega-sized gains scattered over the genome, which is a result of numerous tandem duplications, indicative of gross defects in DNA repair (14). In addition, recent profiling studies have also identified CDK12 mutations in primary and castration-resistant prostate cancer that are mutually exclusive with other mutations in DNA repair genes (15, 16), and akin to HGSOC, result in large tandem duplications (14). On the other hand, in breast cancer, CDK12 gene amplification often co-occurs with ERBB2 amplification as both are colocated at locus Ch17q12 (17, 18), and CDK12 overexpression has been correlated with indicators of aggressive disease, suggesting that CDK12 could act as a oncogenic driver and prognostic biomarker in breast cancer as a result of this colocation (19).

We have previously shown that in breast cancer, CDK12 is recurrently targeted by both DNA rearrangements (13% of HER2-amplified breast cancer) and recurrent point mutations (2.6% of unselected breast cancer; ref. 13) in a similar manner to HGSOC, and that loss of CDK12 in breast cancer models confers sensitivity to PARP1/2 inhibitors in vitro through defects in HR (12, 13). Loss of CDK12 in breast cancer may therefore signify response to platinum salts and/or PARP1/2 inhibitors (12, 13).

Here we sought to (i) investigate the distribution and frequency of CDK12 protein expression in a large series of unselected and Herceptin-treated HER2-positive breast cancer, using IHC and examine any correlation with survival; (ii) evaluate CDK12 protein and mRNA expression with genomic alterations and (iii) assess whether CDK12 would constitute an oncogenic driver in CDK12-amplified tumors.

Tissue microarray patient cohorts

Unselected breast cancer.

Primary operable breast cancer cases (n = 1,650) from the Nottingham Tenovus Primary Breast Carcinoma Series were utilized as described previously (20–22). Patients were under the age of 71 years (median, 55 years), diagnosed between 1986 and 1999, and treated uniformly in a single institution. Clinicopathologic parameters for this series are summarized in Supplementary Table S1.

HER2-positive adjuvant trastuzumab series.

The HER2-positive adjuvant trastuzumab series comprises 143 primary operable breast cancer from patients presenting between 2003 and 2010 who received adjuvant trastuzumab (21). HER2 status was determined according to the American Society of Clinical Oncology (ASCO) guidelines as described previously (21). Clinicopathologic parameters for this series are summarized in Supplementary Table S2.

METABRIC Nottingham breast cancers.

This series comprised 282 primary breast cancers from Nottingham, which form part of the METABRIC cohort (ref. 23; Supplementary Table S3).

Tissue microarray construction.

Tumor samples were arrayed as described previously (22). Briefly, one core per tumor of 0.6-mm thickness was obtained from the most representative areas then reembedded in microarray blocks.

CDK12 IHC.

IHC was optimized in-house, using a standard labeled polymer technique, on 4-μm sections of formalin-fixed paraffin embedded (FFPE) normal human tonsil; cell blocks containing the MCF7 breast cancer cell line known to express CDK12 transfected with a previously validated siRNA pool targeting CDK12 or nontargeting control and BT474 cells as a positive control (13) (Fig. 1). Cells were cultured as described previously (13) and authenticated by short tandem repeat (STR) typing using the StemElite Kit (Promega). Briefly, slides were dewaxed in xylene and rehydrated through graded alcohols. Following heat-induced antigen retrieval in citrate buffer (pH 6.0), sections were incubated with a mouse anti-human CDK12 mAb (1:5,000 final dilution, Abcam clone 57311 that was raised against an immunogen peptide corresponding to amino acids 1,281–1,380 of Human CDK12) for one hour at room temperature. The staining was visualized using the Dako Flex Envision K8002 Kit (Dako), counterstained with Gill's hematoxylin (Leica). Sections were then dehydrated and mounted.

Figure 1.

Distribution of CDK12 protein expression in breast cancer. A, Modified CONSORT diagram depicting the distribution of CDK12-positive and -negative breast cancers in each of the cohorts analyzed. B, Representative micrographs of CDK12 protein expression in MCF7 cell line treated with nontargeting siRNA controls (i); MCF7 cell line treated with previously validated siRNA against CDK12 (ref. 13; ii); BT474 CDK12 amplified cells all at 400× magnification (iii); tonsil positive control (×200 magnification; iv). Representative images of staining intensity in primary breast cancers, where CDK12 expression was quantified using a modified Allred score, which assessed both intensity (highest score = 3) and percentage positivity (highest score = 5): (v) negative; (vi) 1+; (vii) 2+; and (viii) 3+; all images at 200× magnification. A score of 0 was considered absent and a score of 7 or 8, as high expression.

Figure 1.

Distribution of CDK12 protein expression in breast cancer. A, Modified CONSORT diagram depicting the distribution of CDK12-positive and -negative breast cancers in each of the cohorts analyzed. B, Representative micrographs of CDK12 protein expression in MCF7 cell line treated with nontargeting siRNA controls (i); MCF7 cell line treated with previously validated siRNA against CDK12 (ref. 13; ii); BT474 CDK12 amplified cells all at 400× magnification (iii); tonsil positive control (×200 magnification; iv). Representative images of staining intensity in primary breast cancers, where CDK12 expression was quantified using a modified Allred score, which assessed both intensity (highest score = 3) and percentage positivity (highest score = 5): (v) negative; (vi) 1+; (vii) 2+; and (viii) 3+; all images at 200× magnification. A score of 0 was considered absent and a score of 7 or 8, as high expression.

Close modal

TMAs were assessed for nuclear CDK12 protein expression in the malignant epithelium only, using a modified Allred score (14). Only technically sound cores containing >20% invasive tumor cells were included in the analysis. Cores were evaluated for both intensity (0 = no stain; 1 = mild; 2 = moderate; 3 = strong), and percentage of epithelial cells that stained positive (0 = absent; 1 = background; 2 = 1%–25%; 3 = 26%–50%; 4 = 51%–75%; 5 = >75%; Fig. 1). Scores were derived from a sum of the intensity and percentage of immunoreactive cells; an average score of 0 for each tumor was considered negative/absent; a score of 7 or 8, high; and a score of 2–6 as intermediate expression. Scores of 1 were excluded from further analysis as these equated to background nonspecific staining. IHC staining and dichotomization of the other biomarkers included in this study were as per previous publications (24–30). Scoring was performed blinded to the study endpoint.

Mining of public datasets.

To corroborate our findings, we reanalyzed publicly available data from The Cancer Genome Atlas (31, 32) and METABRIC (23) datasets, to ascertain the frequency of CDK12 copy number breakpoints, somatic mutations, and methylation and correlate these with RNA expression levels. Low and high CDK12 gene expression were defined by using an optimal threshold for dichotomizing gene expression data as described previously (33). This was carried out by a stepwise analysis from 40 to 60 percentiles at an interval of 5. The cutoffs that displayed the highest prognostic significance with log-rank test were selected. In addition, analysis of published whole genome shRNA (34) and kinome-wide siRNA (35) genetic perturbation screens was performed to correlate cell viability of breast cancer cell lines with and without CDK12 amplification after CDK12 knockdown.

Assessment of tandem duplicator phenotype.

Affymetrix SNP6.0 copy number data of 224 METABRIC samples were preprocessed using PennCNV-affy package (affy: http://penncnv.openbioinformatics.org/en/latest/misc/credit/ and segmented absolute copy number and ploidy was established with ASCAT 2.1 (36). The two tandem duplication phenotypes were established as previously described in Watkins and colleagues, 2016 (37).

Statistical analysis

Retrospective statistical analysis was performed using SPSS 21.0 statistical software (SPSS Inc.), in compliance with reporting recommendations for tumor marker prognostic studies (REMARK) criteria (38). A χ2or Fisher exact test of <0.05 was considered significant. Survival curves were analyzed by the method of Kaplan–Meier, with a P value <0.05 being considered significant with a 10-year breast cancer–specific survival as the endpoint. Multivariate survival analysis was carried out using CDK12 expression status, node status, estrogen and progesterone receptor (ER/PR) status, HER2 status, age, tumor size, and grade. A Student t test was employed to compare CDK12 expression of mined samples with genetic aberrations and normal controls. For comparisons, scores of 0 (absent) and high (7–8) were compared, given known correlations with high expression and amplification (19) and uncertainty regarding intermediate levels of expression. A P < 0.05 (two-sided) was considered statistically significant. Differential gene expression analysis of CDK12 absent versus high tumors with gene expression known to be involved in DNA repair was performed using data from METABRIC using Limma with FDR (false discovery rate) multiple correction adjustment.

Distribution of CDK12 expression and clinicopathologic correlation

In the first instance, we assessed CDK12 protein expression by IHC in 696 unselected breast cancer samples that met the inclusion criteria (described in Methods; Fig. 1). Overall 73 of 696 tumors were absent/negative for CDK12 (10.5%) by IHC, and 146 had CDK12 high expression (21%; Fig. 1A; Supplementary Table S1). Breaking this down by subtype, in ER-positive patients, 50/510 (9.8%) and 101/510 (19.8%) were CDK12 absent and high; in HER2-positive patients, 3 of 102 (2.9%) and 55 of 102 (53.9%) were CDK12 absent and high; and in triple-negative patients, 21 of 123 (17%) and 17 of 123 (13.8%) were CDK12 absent and high, respectively. Expression of CDK12 significantly correlated with HER2 expression; 96% of CDK12-absent tumors were HER2 negative and 95% of HER2-positive tumors had high CDK12 expression (P < 0.001, χ2 test). Interestingly, no significant correlation of CDK12 expression with ER or PR status was observed, but a greater proportion of CDK12-absent tumors showed a triple-negative phenotype (21/73, 29%) than CDK12-high tumors (17/143, 11.9%, P = 0.002, χ2 test; Supplementary Table S1). There was no association with CDK12-high expression and breast cancer–specific survival in this cohort [P = 0.354; HR = 1.295; 95% CI = 0.75–2.24, log-rank (Mantel–Cox), Fig. 2A; Supplementary Table S4].

Figure 2.

CDK12-positive breast cancer has a poorer survival in univariate analysis. Kaplan–Meier curves showing breast cancer–specific survival for CDK12 high (7–8) versus absent (0) breast cancers asses assessed by IHC in the Nottingham unselected primary breast cancer series (n = 203; A); tumors from METABRIC (n = 140; B); Gene expression correlations of CDK12 low versus high from METABRIC (n = 1,961; C), and HER2-positive tumors treated with Herceptin (n = 75; D).

Figure 2.

CDK12-positive breast cancer has a poorer survival in univariate analysis. Kaplan–Meier curves showing breast cancer–specific survival for CDK12 high (7–8) versus absent (0) breast cancers asses assessed by IHC in the Nottingham unselected primary breast cancer series (n = 203; A); tumors from METABRIC (n = 140; B); Gene expression correlations of CDK12 low versus high from METABRIC (n = 1,961; C), and HER2-positive tumors treated with Herceptin (n = 75; D).

Close modal

These findings were validated in a subset of tumors from the METABRIC cohort of unselected breast cancer, in which CDK12 was highly expressed in 63 of 250 tumors (25.2%) and absent in 89 of 250 tumors (35.6%; Fig. 1; Supplementary Table S3). Again, a significant correlation with HER2 status was observed, with 83 of 89 (93.3%) of CDK12-absent tumors being HER2 negative and 13 of 36 (36.1%) of HER2-amplified tumors being CDK12 high (P < 0.0001, χ2 test). CDK12 expression also conferred a significantly poorer breast cancer–specific survival [P < 0.001; HR = 3.161; 95% CI = 1.632–6.125, log-rank (Mantel–Cox)] in this cohort in univariate analysis (Fig. 2B). This was also significant in multivariate comparisons when taking account for HER2 positivity (P = 0.038; HR = 2.26; 95% CI = 1.045–4.887; Supplementary Table S4). These associations were further corroborated at the mRNA level in a larger cohort of primary tumors (n = 1,961) from METABRIC, where high CDK12 expression was associated with a worse breast cancer–specific survival [P < 0.001; HR = 1.28; 95% CI = 1.17–1.41; log-rank (Mantel–Cox); Fig. 2C]; however, this was not substantiated in multivariate analysis when taking HER2 into account (Supplementary Table S4).

We previously identified a proportion of HER2-amplified tumors harboring out-of-frame CDK12 fusion genes. As CDK12 is known to map to the smallest region within the HER2 amplicon (17, 18), these fusions are the result of an amplification breakpoint in the HER2 amplicon that converges on CDK12, disrupting its expression (13). We also found that tumor cells with loss of CDK12 due to the presence of a breakpoint in the HER2 amplicon were sensitive to PARP1/2 inhibitors suggesting that a fraction of HER2-amplified patients with CDK12 fusions might also benefit from treatment with PARP1/2 inhibitors or platinum chemotherapy (13). To ascertain the frequency of CDK12 protein absence in HER2-amplified patients, and possible associations with outcome subsequent to anti-HER2 therapy, we assessed CDK12 protein expression in HER2-positive patients who had been treated with Herceptin (21). Overall, 4 of 119 (3.4%) tumors were CDK12 absent and 71 of 119 (59.7%) tumors were CDK12 high (Fig. 1A). Lack of CDK12 expression did not improve survival in this cohort following Herceptin treatment [P = 0.586; HR= 22.636; 95% CI = 0–1718290, log-rank (Mantel–Cox); Fig. 2D].

In summary (Supplementary Table S1), overall, absence of CDK12 protein expression was seen in 10.5% of unselected breast cancers; with a similar proportion of ER+ tumors showing absent CDK12 (9.8%). Interestingly, a higher frequency of absent CDK12 was seen in TNBC within this unselected cohort (17%); however, no association with breast cancer–specific survival was seen [P = 0.577; HR= 2; 95% CI = 1.8–3.4, log-rank (Mantel–Cox)]. As expected, in all three cohorts analyzed, a lower proportion of HER2-positive tumors were CDK12 absent. Rather, most HER2-positive tumors were CDK12 high (protein and mRNA). Although high CDK12 expression was associated with a worse breast cancer–specific survival in these cohorts, this association was significant in the METABRIC dataset in multivariate analysis taking HER2 status into account, but not borne out in additional datasets.

Association of CDK12 expression and amplification

We subsequently investigated the associations of CDK12 mRNA expression with genomic status in primary breast cancer from METABRIC (23) where there is copy number and gene expression data on the same tumor specimen. Of 1,979 tumors with matched copy number and gene expression data available for CDK12, 208 (10.5%) harbored amplification encompassing the CDK12 gene and a concurrent increase in its transcript expression, (P < 0.0001; Mann–Whitney U test; Fig. 3A). Of these, 99% (205/208) were also HER2 amplified. There was a significant association with breast cancer–specific survival comparing CDK12 amplified versus nonamplified tumors [P < 0.0001, log-rank (Mantel–Cox; HR = 0.45; 95% CI = 0.35–0.57]; however, similar to CDK12 protein expression, this was lost in multivariate analysis taking HER2 status into account (Supplementary Table S4). Assessment of CDK12 protein expression by IHC in the METABRIC TMA, stratifying tumors as absent (0) versus highly expressed (7–8), however, revealed a significant association between CDK12 amplification and increased CDK12 protein expression with 12 of 16 (75%) amplified and 42 of 120 (35%) nonamplified tumors showing high protein expression, P = 0.0049, Fisher exact test, Fig. 3B).

Figure 3.

CDK12-amplified cells are not dependent on CDK12 expression for their survival. A, Scatter dot plot depicting a significant association of CDK12 transcript expression with gene amplification (n = 208) versus no amplification (n = 1,769; error bars represent median with the interquartile range). B, Bar chart depicting a significant increase of CDK12 protein expression as measured by IHC in CDK12-amplified tumors (n = 16) compared with nonamplified (n = 103). C and D, Relative cell viability after CDK12 silencing in CDK12 amplified (Campbell, n = 8; Marcotte n = 14) versus nonamplified cell lines (Campbell, n = 19; Marcotte n = 40), showing no significant difference in cell survival from Marcotte and colleagues (ref. 34; C) and Campbell and colleagues (ref. 35; D).

Figure 3.

CDK12-amplified cells are not dependent on CDK12 expression for their survival. A, Scatter dot plot depicting a significant association of CDK12 transcript expression with gene amplification (n = 208) versus no amplification (n = 1,769; error bars represent median with the interquartile range). B, Bar chart depicting a significant increase of CDK12 protein expression as measured by IHC in CDK12-amplified tumors (n = 16) compared with nonamplified (n = 103). C and D, Relative cell viability after CDK12 silencing in CDK12 amplified (Campbell, n = 8; Marcotte n = 14) versus nonamplified cell lines (Campbell, n = 19; Marcotte n = 40), showing no significant difference in cell survival from Marcotte and colleagues (ref. 34; C) and Campbell and colleagues (ref. 35; D).

Close modal

Recent evidence has pointed to the role of CDK12 as a potential oncogene, given that that its amplification is associated with increased protein expression and aggressive clinical characteristics (19) and that CDK12-amplified tumors show significantly increased CDK12 protein activation (39). To seek any evidence of oncogenic addiction to the downstream consequences of activated CDK12, we mined publicly available genetic perturbation screens in breast cancer cell line models using validated reagents (12, 13) to ascertain whether CDK12-amplified cells were addicted to CDK12 expression and downstream signaling for their survival. Analysis of both genome-wide pooled shRNA screen data (34) and siRNA kinome screen data (35) failed to identify any association between CDK12 amplification and sensitivity to shRNA or siRNA designed to target CDK12 (Fig. 3C and D).

Genetic mechanisms of absent CDK12 expression

Given our previous findings that disruption of CDK12 can occur as a result of an amplification breakpoint in the ERBB2 amplicon that converges on CDK12, disrupting its expression (13), we sought to confirm our findings in primary tumors from METABRIC (23). Of all cases that showed CDK12 amplification, 14.4% (30/208) harbored a breakpoint in CDK12 that was associated with a significant reduction in CDK12 transcript levels (P < 0.0001; Mann–Whitney U test; Fig. 4A). In HER2-amplified patients, there was no significant difference in breast cancer–specific survival between patients with a breakpoint in CDK12 and those without [P = 0.32; log-rank (Mantel–Cox) HR= 0.9897; 95% CI = 0.5107–1.918].

Figure 4.

CDK12 protein loss is associated with DNA repair defects in breast cancer. A, Box and whisker plot (min–max) showing a significant decrease in CDK12 transcript levels in HER2-amplified tumors with breakpoints (n = 30) in CDK12 compared with no breakpoints (n = 178) from METABRIC. B, Lollipop diagram depicting the distribution of CDK12 mutations in breast cancer (red= frameshift and nonsense mutations, green= nonsynonymous coding mutations). C, Scatter dot plot diagrams showing significant associations between CDK12 absent (n = 74; IHC score 0) versus CDK12 high (n = 51; IHC score 7–8) in unselected METABRIC tumors with the large tandem duplication score, indicative of gross genomic defects. D, Scatter dot plot diagrams showing significant associations between CDK12 absent (n = 8; IHC score 0) versus CDK12 high (n = 4; IHC score 7–8) in TNBC from METABRIC, with the large tandem duplication score, indicative of gross genomic defects.

Figure 4.

CDK12 protein loss is associated with DNA repair defects in breast cancer. A, Box and whisker plot (min–max) showing a significant decrease in CDK12 transcript levels in HER2-amplified tumors with breakpoints (n = 30) in CDK12 compared with no breakpoints (n = 178) from METABRIC. B, Lollipop diagram depicting the distribution of CDK12 mutations in breast cancer (red= frameshift and nonsense mutations, green= nonsynonymous coding mutations). C, Scatter dot plot diagrams showing significant associations between CDK12 absent (n = 74; IHC score 0) versus CDK12 high (n = 51; IHC score 7–8) in unselected METABRIC tumors with the large tandem duplication score, indicative of gross genomic defects. D, Scatter dot plot diagrams showing significant associations between CDK12 absent (n = 8; IHC score 0) versus CDK12 high (n = 4; IHC score 7–8) in TNBC from METABRIC, with the large tandem duplication score, indicative of gross genomic defects.

Close modal

To ascertain the frequency of absent CDK12 protein in tumors with copy number alterations in CDK12, we intersected the IHC data from a subset of the METABRIC cohort performed above with the available copy number data for CDK12. Of all CDK12-amplified tumors, 14% (4/28) showed absent CDK12 protein expression and 33% (2/6) of tumors with a breakpoint in CDK12 were CDK12 absent. This highlights that only a proportion of tumors with CDK12 genomic breakpoints lead to loss of CDK12 protein expression.

In HGSOC, CDK12-inactivating mutations have been reported to inactivate gene expression and consequently abrogate HR DNA repair pathways (10, 11, 14). Examination of DNA sequencing data from The Cancer Genome Atlas (TCGA; ref. 32) identified CDK12 mutations in 1.5% (20/1,373) of unselected breast cancer (Fig. 4B; Supplementary Table S5 and S6). These included 12 missense mutations, and 8 truncating mutations. Overall, 45% [9/20 (8 truncating and 1 missense] were predicted to disrupt protein function, and of these 25% (2/9) were seen in patients with triple-negative disease, whereas the remaining patients had ER+ or ER/HER2+ disease. In contrast with recent data in HGSOC (11), predicted deleterious mutations did not correlate with reduced transcript expression, or with downregulation of DDR genes, (Supplementary Table S7). CDK12 promoter methylation was a rare event seen in 1 of 696 (0.14%) of unselected breast cancer (Supplementary Table S6). We next looked for other genomic alterations that would lead to absent CDK12 protein expression by interrogation of METABRIC cases with copy number and miRNA expression (23, 40). This identified that heterozygous loss of CDK12 accounted for 7.1% (6/84) cases with absent CDK12 protein expression; however, this was not enriched in the CDK12-absent group compared with the CDK12-high group (P > 0.999, Fisher exact test). In addition, of the 14 of 162 miRNAs, that are known or predicted targets of CDK12 present in METABRIC (40), none showed correlation with CDK12 protein expression after multiple correction (Supplementary Table S8), suggesting there are additional mechanisms that lead to CDK12 protein loss.

Absent CDK12 protein expression is associated with reduced protein expression of genes involved in DNA repair

Although predicted deleterious mutations in CDK12 did not correlate with reduced expression of DDR genes (see above), we assessed whether loss of CDK12 protein expression correlated with a reduction in the transcript expression of DNA repair proteins or biomarkers of DNA damage in the unselected series of breast cancer cases (Table 1). Although no significant correlations were observed (after multiple testing correction) with DDR genes at the mRNA level (Supplementary Table S9), absent CDK12 protein was significantly correlated with reduced protein expression of ATR, APE1, nuclear and cytoplasmic SMC6, Ku70/Ku80, DNA-PK, and γH2AX nuclear positivity. Absent CDK12 protein expression also significantly correlated with both cleaved PARP1 (P = 0.003, χ2) and noncleaved PARP1 (P = 0.005, χ2) expression suggesting that PARP1 levels are higher in tumors with absent CDK12. Interestingly, absent CDK12 expression also correlated with a decreased expression of TP53 (P = 0.001; χ2) and RB1 (P = 0.003, χ2); however, there was no significant correlation between absent CDK12 protein and TP53 mutations in the METABRIC cohort of tumors (17.5% CDK12 absent and 26.4% CDK12-high tumors harboring TP53 mutations P = 0.2779, Fisher exact test).

Table 1.

Association of CDK12 expression with DNA repair proteins in unselected and TNBC. All P values are from χ2 or Fisher exact test. *, P < 0.05; **, P < 0.01

CDK12 Absent (0) n (%)CDK12 High (7/8) n (%)P
Unselected series 
BRCA1 (n = 170)   0.249 
 Negative 27 (51.9) 50 (42.4)  
 Positive 25 (48.1) 68 (57.6)  
PARP1 cleaved (n = 170)   0.003** 
 Negative 18 (30) 13 (11.8)  
 Positive 42 (70) 97 (88.2)  
PARP1 noncleaved (n = 175)   0.005** 
 Negative 36 (62.1) 46 (39.3)  
 Positive 22 (37.9) 71 (60.7)  
ATR (n = 173)   <0.001** 
 Negative 46 (70.8) 45 (41.7)  
 Positive 19 (29.2) 63 (58.3)  
APE1 (n = 89)   0.02* 
 Negative 10 (38.5) 10 (15.9)  
 Positive 16 (61.5) 53 (84.1)  
Ku70/Ku80 (n = 153)   0.001** 
 Negative 18 (34.0) 11 (11.0)  
 Positive 35 (66.0) 89 (89.0)  
DNA-PK (n = 156)   0.002** 
 Negative 17 (32.1) 12 (11.7)  
 Positive 36 (67.9) 91 (88.3)  
SMC6 cytoplasmic (n = 166)   0.002** 
 Negative 30 (54.5) 33 (29.7)  
 Positive 25 (45.5) 78 (70.3)  
SMC6 nuclear (n = 166)   <0.001** 
 Negative 33 (60.0) 34 (30.6)  
 Positive 22 (40.0) 77 (69.4)  
γH2AX (n = 150)   0.002** 
 Negative 24 (47.1) 22 (22.2)  
 Positive 27 (52.9) 77 (77.8)  
TP53 (n = 209)   0.001** 
 Negative 56 (80.0) 80 (57.6)  
 Positive 14 (20.0) 59 (42.4)  
CHEK1 cytoplasmic (n = 103)   0.321 
 Negative 29 (50.9) 44 (42.7)  
 Positive 28 (49.1) 59 (57.3)  
CHEK1 nuclear (n = 161)   0.927 
 Negative 48 (84.2) 87 (83.7)  
 Positive 9 (15.8) 17 (16.3)  
CHEK2 (n = 108)   0.345 
 Negative 31 (50.8) 63 (58.3)  
 Positive 30 (49.2) 45 (41.7)  
Triple-negative breast cancer 
BRCA1 (n = 28)   0.430 
 Negative 10 (71.4) 8 (57.1)  
 Positive 4 (28.6) 6 (42.9)  
PARP1 cleaved (n = 28)   0.378 
 Negative 8 (50.0) 4 (33.3)  
 Positive 8 (50.0) 8 (66.7)  
PARP1 noncleaved (n = 29)   0.103 
 Negative 11 (68.8) 5 (38.5)  
 Positive 5 (31.2) 8 (61.5)  
ATR (n = 33)   0.018* 
 Negative 18 (90.0) 7 (53.8)  
 Positive 2 (10.0) 6 (46.2)  
APE1 (n = 17)   0.453 
 Negative 2 (25.0) 1 (11.1)  
 Positive 6 (75.0) 8 (88.9)  
Ku70/Ku80 (n = 27)   0.010* 
 Negative 8 (47.1) 0 (0.0)  
 Positive 9 (52.9) 10 (100.0)  
DNA-PK (n = 31)   0.283 
 Negative 6 (37.5) 3 (25.0)  
 Positive 10 (62.5) 12 (75.0)  
SMC6 cytoplasmic (n = 31)   0.018* 
 Negative 9 (60.0) 3 (23.1)  
 Positive 6 (40.0) 13 (76.9)  
SMC6 nuclear (n = 31)   0.045* 
 Negative 11 (73.3) 6 (37.5)  
 Positive 4 (26.7) 10 (62.5)  
γH2AX (n = 29)   0.089 
 Negative 9 (60.0) 4 (28.6)  
 Positive 6 (40.0) 10 (71.4)  
TP53 (n = 36)   0.187 
 Negative 12 (63.2) 7 (41.2)  
 Positive 7 (36.8) 10 (58.8)  
CHEK1 cytoplasmic (n = 27)   0.384 
 Negative 8 (47.1) 3 (30.0)  
 Positive 9 (52.9) 7 (70.0)  
CHEK1 nuclear (n = 28)   0.068 
 Negative 17 (100.0) 9 (81.8)  
 Positive 0 (0.0) 2 (18.2)  
CHEK2 (n = 32)   0.961 
 Negative 13 (68.4) 9 (69.2)  
 Positive 6 (31.6) 4 (30.8)  
CDK12 Absent (0) n (%)CDK12 High (7/8) n (%)P
Unselected series 
BRCA1 (n = 170)   0.249 
 Negative 27 (51.9) 50 (42.4)  
 Positive 25 (48.1) 68 (57.6)  
PARP1 cleaved (n = 170)   0.003** 
 Negative 18 (30) 13 (11.8)  
 Positive 42 (70) 97 (88.2)  
PARP1 noncleaved (n = 175)   0.005** 
 Negative 36 (62.1) 46 (39.3)  
 Positive 22 (37.9) 71 (60.7)  
ATR (n = 173)   <0.001** 
 Negative 46 (70.8) 45 (41.7)  
 Positive 19 (29.2) 63 (58.3)  
APE1 (n = 89)   0.02* 
 Negative 10 (38.5) 10 (15.9)  
 Positive 16 (61.5) 53 (84.1)  
Ku70/Ku80 (n = 153)   0.001** 
 Negative 18 (34.0) 11 (11.0)  
 Positive 35 (66.0) 89 (89.0)  
DNA-PK (n = 156)   0.002** 
 Negative 17 (32.1) 12 (11.7)  
 Positive 36 (67.9) 91 (88.3)  
SMC6 cytoplasmic (n = 166)   0.002** 
 Negative 30 (54.5) 33 (29.7)  
 Positive 25 (45.5) 78 (70.3)  
SMC6 nuclear (n = 166)   <0.001** 
 Negative 33 (60.0) 34 (30.6)  
 Positive 22 (40.0) 77 (69.4)  
γH2AX (n = 150)   0.002** 
 Negative 24 (47.1) 22 (22.2)  
 Positive 27 (52.9) 77 (77.8)  
TP53 (n = 209)   0.001** 
 Negative 56 (80.0) 80 (57.6)  
 Positive 14 (20.0) 59 (42.4)  
CHEK1 cytoplasmic (n = 103)   0.321 
 Negative 29 (50.9) 44 (42.7)  
 Positive 28 (49.1) 59 (57.3)  
CHEK1 nuclear (n = 161)   0.927 
 Negative 48 (84.2) 87 (83.7)  
 Positive 9 (15.8) 17 (16.3)  
CHEK2 (n = 108)   0.345 
 Negative 31 (50.8) 63 (58.3)  
 Positive 30 (49.2) 45 (41.7)  
Triple-negative breast cancer 
BRCA1 (n = 28)   0.430 
 Negative 10 (71.4) 8 (57.1)  
 Positive 4 (28.6) 6 (42.9)  
PARP1 cleaved (n = 28)   0.378 
 Negative 8 (50.0) 4 (33.3)  
 Positive 8 (50.0) 8 (66.7)  
PARP1 noncleaved (n = 29)   0.103 
 Negative 11 (68.8) 5 (38.5)  
 Positive 5 (31.2) 8 (61.5)  
ATR (n = 33)   0.018* 
 Negative 18 (90.0) 7 (53.8)  
 Positive 2 (10.0) 6 (46.2)  
APE1 (n = 17)   0.453 
 Negative 2 (25.0) 1 (11.1)  
 Positive 6 (75.0) 8 (88.9)  
Ku70/Ku80 (n = 27)   0.010* 
 Negative 8 (47.1) 0 (0.0)  
 Positive 9 (52.9) 10 (100.0)  
DNA-PK (n = 31)   0.283 
 Negative 6 (37.5) 3 (25.0)  
 Positive 10 (62.5) 12 (75.0)  
SMC6 cytoplasmic (n = 31)   0.018* 
 Negative 9 (60.0) 3 (23.1)  
 Positive 6 (40.0) 13 (76.9)  
SMC6 nuclear (n = 31)   0.045* 
 Negative 11 (73.3) 6 (37.5)  
 Positive 4 (26.7) 10 (62.5)  
γH2AX (n = 29)   0.089 
 Negative 9 (60.0) 4 (28.6)  
 Positive 6 (40.0) 10 (71.4)  
TP53 (n = 36)   0.187 
 Negative 12 (63.2) 7 (41.2)  
 Positive 7 (36.8) 10 (58.8)  
CHEK1 cytoplasmic (n = 27)   0.384 
 Negative 8 (47.1) 3 (30.0)  
 Positive 9 (52.9) 7 (70.0)  
CHEK1 nuclear (n = 28)   0.068 
 Negative 17 (100.0) 9 (81.8)  
 Positive 0 (0.0) 2 (18.2)  
CHEK2 (n = 32)   0.961 
 Negative 13 (68.4) 9 (69.2)  
 Positive 6 (31.6) 4 (30.8)  

As absent CDK12 protein expression was seen in 17% of TNBC within our analysis, we assessed this subset of tumors for correlations with the expression of DNA repair proteins as above. Even within this relative small subset, some significant correlations were still seen with DDR genes at the protein level: ATR (P = 0.018, χ2); Ku70/Ku80 (P = 0.01, χ2); and loss or decrease in nuclear and cytoplasmic SMC6 (P = 0.045, χ2; Table 1). Of note, comparison of CDK12 absent (0) and intermediate (2–6) levels of expression together as one group versus high (7–8) failed to identify any significant associations with DNA repair genes both in unselected BC and TNBC (Supplementary Table S10). This suggests that functional loss of CDK12 is only observed in tumors with absent CDK12 protein expression in breast cancer and not low levels of CDK12 expression.

It has been shown that in HGSOC, CDK12 mutations are consistently associated with a particular genomic instability pattern characterized by hundreds of tandem duplications of up to 10 megabases (Mb) in size, dubbed the “CDK12 TD-plus phenotype” (14). Assessment of this pattern in both unselected and TNBC with absent CDK12 protein failed to identify any association with large numbers and sizes of tandem duplications (Fig. 4C and D). Together, these results suggest that absent CDK12 protein in breast cancer is associated with some defects in DNA repair–related genes; however, the resultant genomic scars are likely to be different to that seen in HGSOC (11).

In this study, we were able to analyze, for the first time, the distribution of CDK12 protein in cohorts of primary breast cancer. We show that high CDK12 expression is significantly correlated with HER2 status and that absent CDK12 is associated with a triple-negative phenotype and disruption of proteins involved in DNA repair.

The significant correlation of CDK12 with HER2 expression in these data is not surprising, as CDK12 is known to map to the smallest region within the HER2 amplicon (17, 18). Therefore, although we observed significant correlations with CDK12 expression and patient survival, this significance was subsequently lost when HER2 positivity was taken into account. However, this association was substantiated in the METABRIC cohort at the protein level, suggesting investigation of additional cohorts is warranted. It has been postulated that CDK12 may act in an oncogenic manner, given observed associations with amplification and increased transcript and protein expression and subsequent phosphorylation (19, 39). This may perhaps occur through its reported roles in transcription through phosphorylation of RNA Pol II and regulation of pre-mRNA processing (5, 7). By analyzing published siRNA and shRNA cell viability screens, using validated reagents (12, 13), we show here that amplified cells are not addicted to CDK12 for their survival. We concede the assay length (average 4–7 days) in these experiments, may be too short to detect a loss of viability to depletion of a cyclin-dependent kinase, and additional time may be required to reduce DNA repair protein expression and accumulate damage required to affect viability. Interestingly, a recent study has suggested that CDK12 does act in an oncogenic manner in HER2-amplified cells by promoting cell invasion via alternative splicing and subsequent downregulation of the long isoform of DNAJB6 (41), suggestive of a mechanism by which CDK12 acts in an oncogenic manner to promote cell invasiveness. Of note, elevated levels of CDK12 in HER2-positive tumors have been a proposed reason why tumors arising in BRCA1 carriers are usually ERBB2-negative, given elevated expression would oppose the defects in HR mediated by BRCA1 loss (34).

We have shown previously that a proportion of HER2-positive tumors harbor inactivating (out-of-frame) fusion genes in CDK12 that are due to a copy number breakpoint in the HER2 amplicon converging on CDK12, resulting in a significant decrease in both transcript and protein levels (13). Furthermore, cell lines with breakpoints in CDK12 that result in loss of protein expression are sensitive to PARP1/2 inhibitor therapy, due to impaired HR-mediated DNA repair (13). By assessing the distribution of CDK12 protein expression in CDK12-amplified breast cancers in the METABRIC cohort, we identified 14% that were CDK12 absent. Moreover, of the tumors with a breakpoint in CDK12, 33% had absent CDK12 protein expression. While the numbers we were able to assess are small, this nevertheless suggests a proportion of HER2-positive patients show absent CDK12 protein as a result of both copy number breakpoints and additional mechanisms. Overall, our data suggest there may be a small proportion of HER2-positive patients that may benefit from treatment with DNA damage response–targeted therapies such as PARP1/2 inhibitor therapy.

Recent data in HGSOC has shown that CDK12 point mutations inactivate gene expression subsequently abrogating HR DNA repair pathways (10, 11). Although in our study CDK12 mutations did not themselves correlate with decreased expression of DNA repair genes (perhaps due to residual CDK12 activity, given some of these are not associated with loss of heterozygosity), we show that breast cancer with absent CDK12 protein show downregulation of a number of genes involved in functional HR DNA repair at the protein level, suggesting that absent CDK12 protein in breast cancer may also be associated with HR defects. Although we were unable to assess all proteins known to be dysregulated in CDK12-mutant HGSOC such as RAD51, FANCI, and FANCD2, it is reasonable to postulate that CDK12 may have other targets, including DNA repair proteins themselves, phosphorylation of which is required for protein stability. Lack of any observed associations at the mRNA level, may be due to small numbers in the study, however, our observed associations warrant further investigation in larger cohorts. Recent evidence in HGSOC points to a unique signature of genomic instability in CDK12-mutated tumors that is indicative of gross defects in DNA repair characterized by tens to hundreds of large tandem duplications scattered throughout the genome, although this was not observed in CDK12-mutated breast cancers (14). Consistent with this, we observed no correlation with large tandem duplications of neither all breast cancer nor TNBC specifically. It is intriguing, however, that we observed absent CDK12 protein in 17% of TNBC from the unselected primary breast cancer analyzed in this study. Of course, this could be a consequence of the small number of cases included; however, CDK12 inactivation may play a role in a subset of TNBC's possibly through mechanisms distinct from those observed in HGSOC.

In conclusion, we have shown that a subset of HER2-positive patients show absent CDK12 protein expression and have shown an enrichment of absent CDK12 protein expression in TNBC. Moreover, we have provided evidence that absent CDK12 expression is associated with defects in DNA repair proteins. These results suggest that CDK12 IHC could potentially be useful, once validated in sufficiently powered studies, for stratification of patients for treatment with DDR-targeted therapies.

C.J. Lord is a named inventor on patents describing the use of PARP inhibitors and stand to gain from their use as part of the ICR “Rewards to Inventors” scheme. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Naidoo, E.A. Rakha, R. Natrajan

Development of methodology: K. Naidoo, F. Daley, E.A. Rakha, I.O. Ellis, C.J. Lord, R. Natrajan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Naidoo, P.T. Wai, A. Tutt, P.M. Moseley, T.M.A. Abdel-Fatah, S.Y.T. Chan, S. Madhusudan, E.A. Rakha, I.O. Ellis, C.J. Lord, A.R. Green

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Naidoo, S.L. Maguire, S. Haider, H. Mirza, A. Grigoriadis, S.Y.T. Chan, S. Madhusudan, C.J. Lord, Y. Yuan, A.R. Green

Writing, review, and/or revision of the manuscript: K. Naidoo, D. Kriplani, A. Tutt, P.M. Moseley, S.Y.T. Chan, S. Madhusudan, E.A. Rakha, I.O. Ellis, C.J. Lord, A.R. Green, R. Natrajan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Daley, D. Kriplani, J. Campbell, P.M. Moseley, S.Y.T. Chan, S. Madhusudan, E.A. Rakha

Study supervision: S.Y.T. Chan, R. Natrajan

The results shown here are in part based upon data generated by the TCGA Research Network: http://cancergenome.nih.gov/. This study makes use of METABRIC data generated by the Molecular Taxonomy of Breast Cancer International Consortium with funding provided by Cancer Research UK and the British Columbia Cancer Agency Branch. We acknowledge NHS funding to the ICR/Royal Marsden Hospital Biomedical Research Centre. We thank J. Watkins for providing the tandem duplication R script. We also thank the Breast Cancer Now Tissue Bank/Nottingham Health Science Biobank (NHSB) for tissue provision. This study was funded by Breast Cancer Now. R. Natrajan is the recipient of a Breast Cancer Now Career Development Fellowship (2011MaySF01). H. Mirza, A.R. Green, and A. Tutt are supported by the Breast Cancer Now Research Unit (former Breakthrough Breast Cancer Research) funding at King's College London.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Early Breast Cancer Trialists' Collaborative Group
,
Peto
R
,
Davies
C
,
Godwin
J
,
Gray
R
,
Pan
HC
, et al
Comparisons between different polychemotherapy regimens for early breast cancer: meta-analyses of long-term outcome among 100,000 women in 123 randomised trials
.
Lancet
2012
;
379
:
432
44
.
2.
Early Breast Cancer Trialists' Collaborative Group
,
Davies
C
,
Godwin
J
,
Gray
R
,
Clarke
M
,
Cutter
D
, et al
Relevance of breast cancer hormone receptors and other factors to the efficacy of adjuvant tamoxifen: patient-level meta-analysis of randomised trials
.
Lancet
2011
;
378
:
771
84
.
3.
Cuzick
J
,
Sestak
I
,
Baum
M
,
Buzdar
A
,
Howell
A
,
Dowsett
M
, et al
Effect of anastrozole and tamoxifen as adjuvant treatment for early-stage breast cancer: 10-year analysis of the ATAC trial
.
Lancet Oncol
2010
;
11
:
1135
41
.
4.
Gianni
L
,
Dafni
U
,
Gelber
RD
,
Azambuja
E
,
Muehlbauer
S
,
Goldhirsch
A
, et al
Treatment with trastuzumab for 1 year after adjuvant chemotherapy in patients with HER2-positive early breast cancer: a 4-year follow-up of a randomised controlled trial
.
Lancet Oncol
2011
;
12
:
236
44
.
5.
Cheng
SW
,
Kuzyk
MA
,
Moradian
A
,
Ichu
TA
,
Chang
VC
,
Tien
JF
, et al
Interaction of cyclin-dependent kinase 12/CrkRS with cyclin K1 is required for the phosphorylation of the C-terminal domain of RNA polymerase II
.
Mol Cell Biol
2012
;
32
:
4691
704
.
6.
Chen
HH
,
Wang
YC
,
Fann
MJ
. 
Identification and characterization of the CDK12/cyclin L1 complex involved in alternative splicing regulation
.
Mol Cell Biol
2006
;
26
:
2736
45
.
7.
Liang
K
,
Gao
X
,
Gilmore
JM
,
Florens
L
,
Washburn
MP
,
Smith
E
, et al
Characterization of human cyclin-dependent kinase 12 (CDK12) and CDK13 complexes in C-terminal domain phosphorylation, gene transcription, and RNA processing
.
Mol Cell Biol
2015
;
35
:
928
38
.
8.
Blazek
D
,
Kohoutek
J
,
Bartholomeeusen
K
,
Johansen
E
,
Hulinkova
P
,
Luo
Z
, et al
The Cyclin K/Cdk12 complex maintains genomic stability via regulation of expression of DNA damage response genes
.
Genes Dev
2011
;
25
:
2158
72
.
9.
Cancer Genome Atlas Research Network
. 
Integrated genomic analyses of ovarian carcinoma
.
Nature
2011
;
474
:
609
15
.
10.
Joshi
PM
,
Sutor
SL
,
Huntoon
CJ
,
Karnitz
LM
. 
Ovarian cancer-associated mutations disable catalytic activity of CDK12, a kinase that promotes homologous recombination repair and resistance to cisplatin and poly(ADP-ribose) polymerase inhibitors
.
J Biol Chem
2014
;
289
:
9247
53
.
11.
Ekumi
KM
,
Paculova
H
,
Lenasi
T
,
Pospichalova
V
,
Bosken
CA
,
Rybarikova
J
, et al
Ovarian carcinoma CDK12 mutations misregulate expression of DNA repair genes via deficient formation and function of the Cdk12/CycK complex
.
Nucleic Acids Res
2015
;
43
:
2575
89
.
12.
Bajrami
I
,
Frankum
JR
,
Konde
A
,
Miller
RE
,
Rehman
FL
,
Brough
R
, et al
Genome-wide profiling of genetic synthetic lethality identifies CDK12 as a novel determinant of PARP1/2 inhibitor sensitivity
.
Cancer Res
2014
;
74
:
287
97
.
13.
Natrajan
R
,
Wilkerson
PM
,
Marchio
C
,
Piscuoglio
S
,
Ng
CK
,
Wai
P
, et al
Characterization of the genomic features and expressed fusion genes in micropapillary carcinomas of the breast
.
J Pathol
2014
;
232
:
553
65
.
14.
Popova
T
,
Manie
E
,
Boeva
V
,
Battistella
A
,
Goundiam
O
,
Smith
NK
, et al
Ovarian cancers harboring inactivating mutations in CDK12 display a distinct genomic instability pattern characterized by large tandem duplications
.
Cancer Res
2016
;
76
:
1882
91
.
15.
Grasso
CS
,
Wu
YM
,
Robinson
DR
,
Cao
X
,
Dhanasekaran
SM
,
Khan
AP
, et al
The mutational landscape of lethal castration-resistant prostate cancer
.
Nature
2012
;
487
:
239
43
.
16.
Robinson
D
,
Van Allen
EM
,
Wu
YM
,
Schultz
N
,
Lonigro
RJ
,
Mosquera
JM
, et al
Integrative clinical genomics of advanced prostate cancer
.
Cell
2015
;
161
:
1215
28
.
17.
Arriola
E
,
Marchio
C
,
Tan
DS
,
Drury
SC
,
Lambros
MB
,
Natrajan
R
, et al
Genomic analysis of the HER2/TOP2A amplicon in breast cancer and breast cancer cell lines
.
Lab Invest
2008
;
88
:
491
503
.
18.
Sircoulomb
F
,
Bekhouche
I
,
Finetti
P
,
Adelaide
J
,
Ben Hamida
A
,
Bonansea
J
, et al
Genome profiling of ERBB2-amplified breast cancers
.
BMC Cancer
2010
;
10
:
539
.
19.
Capra
M
,
Nuciforo
PG
,
Confalonieri
S
,
Quarto
M
,
Bianchi
M
,
Nebuloni
M
, et al
Frequent alterations in the expression of serine/threonine kinases in human cancers
.
Cancer Res
2006
;
66
:
8147
54
.
20.
Elston
CW
,
Ellis
IO
. 
Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up
.
Histopathology
2002
;
41
:
154
61
.
21.
Green
AR
,
Barros
FF
,
Abdel-Fatah
TM
,
Moseley
P
,
Nolan
CC
,
Durham
AC
, et al
HER2/HER3 heterodimers and p21 expression are capable of predicting adjuvant trastuzumab response in HER2+ breast cancer
.
Breast Cancer Res Treat
2014
;
145
:
33
44
.
22.
Abd El-Rehim
DM
,
Ball
G
,
Pinder
SE
,
Rakha
E
,
Paish
C
,
Robertson
JF
, et al
High-throughput protein expression analysis using tissue microarray technology of a large well-characterised series identifies biologically distinct classes of breast cancer confirming recent cDNA expression analyses
.
Int J Cancer
2005
;
116
:
340
50
.
23.
Curtis
C
,
Shah
SP
,
Chin
SF
,
Turashvili
G
,
Rueda
OM
,
Dunning
MJ
, et al
The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups
.
Nature
2012
;
486
:
346
52
.
24.
Green
AR
,
Aleskandarany
MA
,
Agarwal
D
,
Elsheikh
S
,
Nolan
CC
,
Diez-Rodriguez
M
, et al
MYC functions are specific in biological subtypes of breast cancer and confers resistance to endocrine therapy in luminal tumours
.
Br J Cancer
2016
;
114
:
917
28
.
25.
Rakha
EA
,
El-Sayed
ME
,
Green
AR
,
Lee
AH
,
Robertson
JF
,
Ellis
IO
. 
Prognostic markers in triple-negative breast cancer
.
Cancer
2007
;
109
:
25
32
.
26.
Rakha
EA
,
Elsheikh
SE
,
Aleskandarany
MA
,
Habashi
HO
,
Green
AR
,
Powe
DG
, et al
Triple-negative breast cancer: distinguishing between basal and nonbasal subtypes
.
Clin Cancer Res
2009
;
15
:
2302
10
.
27.
Aleskandarany
MA
,
Green
AR
,
Benhasouna
AA
,
Barros
FF
,
Neal
K
,
Reis-Filho
JS
, et al
Prognostic value of proliferation assay in the luminal, HER2-positive, and triple-negative biologic classes of breast cancer
.
Breast Cancer Res
2012
;
14
:
R3
.
28.
Aleskandarany
MA
,
Green
AR
,
Rakha
EA
,
Mohammed
RA
,
Elsheikh
SE
,
Powe
DG
, et al
Growth fraction as a predictor of response to chemotherapy in node-negative breast cancer
.
Int J Cancer
2010
;
126
:
1761
9
.
29.
Aleskandarany
MA
,
Rakha
EA
,
Ahmed
MA
,
Powe
DG
,
Ellis
IO
,
Green
AR
. 
Clinicopathologic and molecular significance of phospho-Akt expression in early invasive breast cancer
.
Breast Cancer Res Treat
2011
;
127
:
407
16
.
30.
Aleskandarany
MA
,
Rakha
EA
,
Ahmed
MA
,
Powe
DG
,
Paish
EC
,
Macmillan
RD
, et al
PIK3CA expression in invasive breast cancer: a biomarker of poor prognosis
.
Breast Cancer Res Treat
2010
;
122
:
45
53
.
31.
Cancer Genome Atlas Network
. 
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
32.
Ciriello
G
,
Gatza
ML
,
Beck
AH
,
Wilkerson
MD
,
Rhie
SK
,
Pastore
A
, et al
Comprehensive molecular portraits of invasive lobular breast cancer
.
Cell
2015
;
163
:
506
19
.
33.
Todd
JR
,
Ryall
KA
,
Vyse
S
,
Wong
JP
,
Natrajan
RC
,
Yuan
Y
, et al
Systematic analysis of tumour cell-extracellular matrix adhesion identifies independent prognostic factors in breast cancer
.
Oncotarget
2016
;
7
:
62939
53
.
34.
Marcotte
R
,
Sayad
A
,
Brown
KR
,
Sanchez-Garcia
F
,
Reimand
J
,
Haider
M
, et al
Functional genomic landscape of human breast cancer drivers, vulnerabilities, and resistance
.
Cell
2016
;
164
:
293
309
.
35.
Campbell
J
,
Ryan
CJ
,
Brough
R
,
Bajrami
I
,
Pemberton
HN
,
Chong
IY
, et al
Large-scale profiling of kinase dependencies in cancer cell lines
.
Cell Rep
2016
;
14
:
2490
501
.
36.
Van Loo
P
,
Nordgard
SH
,
Lingjaerde
OC
,
Russnes
HG
,
Rye
IH
,
Sun
W
, et al
Allele-specific copy number analysis of tumors
.
Proc Natl Acad Sci U S A
2010
;
107
:
16910
5
.
37.
Watkins
J
,
Tutt
A
,
Grigoriadis
A
. 
Tandem duplications contribute to not one but two distinct phenotypes
.
Proc Natl Acad Sci U S A
2016
;
113
:
E5257
8
.
38.
McShane
LM
,
Altman
DG
,
Sauerbrei
W
,
Taube
SE
,
Gion
M
,
Clark
GM
, et al
REporting recommendations for tumor MARKer prognostic studies (REMARK)
.
Breast Cancer Res Treat
2006
;
100
:
229
35
.
39.
Mertins
P
,
Mani
DR
,
Ruggles
KV
,
Gillette
MA
,
Clauser
KR
,
Wang
P
, et al
Proteogenomics connects somatic mutations to signalling in breast cancer
.
Nature
2016
;
534
:
55
62
.
40.
Dvinge
H
,
Git
A
,
Graf
S
,
Salmon-Divon
M
,
Curtis
C
,
Sottoriva
A
, et al
The shaping and functional consequences of the microRNA landscape in breast cancer
.
Nature
2013
;
497
:
378
82
.
41.
Tien
JF
,
Mazloomian
A
,
Cheng
SG
,
Hughes
CS
,
Chow
CC
,
Canapi
LT
, et al
CDK12 regulates alternative last exon mRNA splicing and promotes breast cancer cell invasion
.
Nucleic Acids Res
2017
;
45
:
6698
6716
.

Supplementary data