ATM mutations in locally advanced breast cancer
ATM gene alterations recorded through sequencing (n
= 149 tumors) and MLPA analysis (n
= 69 tumors) in different breast cancer cohorts are summarized in Table (see Materials and methods for cohort details). Most alterations found were of germline origin and previously observed by others [26
] in multiple patients, indicating these changes to be polymorphic variants rather than mutations contributing to a malignant phenotype.
Seven of the mutations observed have, to our knowledge, not previously been reported (Table ). Each of these variants was observed in a single patient only. Peripheral blood lymphocytes were available for four of these patients. Out of these four, one mutation proved to be somatic, whereas three were also found in lymphocyte DNA, indicating the mutations to be of germline origin.
ATM mutations are not associated with chemo-resistance
We compared gene alterations in tumors progressing on therapy (PD) versus tumors not progressing (stable disease or an objective response) classified according to the UICC criteria [30
] as previously described [10
]. The frequency of ATM
mutations was similar among patients with progressive disease (PD) upon treatment and those responding to therapy in any of the cohorts analyzed (data not shown). Stratifying tumors into wild-type versus TP53
mutated ones did not reveal any imbalance regarding ATM mutation incidence between these subgroups.
Assessing the possibility that some ATM variants in particular could be associated with therapy resistance, we compared individual mutations observed among PD-patients with those observed in responders. However, none of the individual mutations (Ser49Cys, Asp1853Asn, Phe858Leu or Pro1054Arg) was found at higher incidence among patients progressing on therapy as compared to responders.
We further assessed the potential impact of ATM mutations on response to paclitaxel monotherapy. Contrasting what has been recorded for the anthracyclines, TP53
mutations do not predict therapy resistance towards the taxanes [23
]. Five out of the 11 analyzed patients displaying progressive disease upon paclitaxel monotherapy treatment harbored ATM
mutations (compared to 10 out of 27 responders). Thus, no association between ATM
mutations and resistance to paclitaxel therapy was recorded.
Low ATM expression levels predict chemo-resistance to doxorubicin and mitomycin but not to paclitaxel in tumors wild-type for TP53 and CHEK2
ATM mRNA levels were determined by qPCR in 69 out of the 71 doxorubicin or 5-fluorouracil/mitomycin treated patients (Cohort 1) from whom RNA were available (for two patients, one with partial response and one with stable disease upon treatment, sufficient amounts of RNA was lacking; Table ). Results from these analyses revealed large differences in ATM mRNA levels in the cohort, with a 56.9-fold ratio between the highest and the lowest value recorded (Figure ).
Figure 1 Relative intratumor ATM mRNA levels among patients with locally advanced breast cancer. (A) ATM mRNA levels among patients receiving neoadjuvant doxorubicin or 5-fluorouracil/mitomycin (n = 69; Cohort 1). Blue bars represent patients displaying stable (more ...)
No association between ATM mRNA levels and ATM mutation status was observed (P
> 0.5; Mann-Whitney rank test). Further, since previous studies have indicated that concomitant ATM and p53 inactivation is underrepresented in breast tumors [22
], we assessed the ATM levels among TP53
mutated versus wild-type tumors; no difference between the two groups was observed (P
Considering the 18 patients with progressive disease, these patients displayed a non-significant trend towards lower ATM mRNA levels as compared to the responders (P = 0.104; Mann-Whitney rank test), with 12 out of 18 expressing ATM levels below the median value of the cohort (P = 0.168; Fischer exact test).
To test the hypothesis that low ATM expression may be an alternative mechanism inactivating the p53 pathway, we compared ATM mRNA expression levels in tumors resistant to chemotherapy despite harboring wild-type TP53/CHEK2 (Group A Table ; n = 5) to the other tumors (Group B + C + D; n = 64) in the same cohort (Cohort 1). ATM-levels were lower among tumors in Group A as compared to the tumors in the other three groups (P = 0.012). Stratifying the latter tumors (n = 64) into TP53/CHEK2 mutated (Group B + D; n = 40) and TP53/CHEK2 wild-type (Group C; n = 24) tumors, the tumors in the A group expressed lower ATM levels when compared to each of these two subgroups (P = 0.010 and 0.028, respectively). Notably, each of the tumors in Group A revealed an ATM expression level in the lower tertile of the total cohort.
Grouping of tumors used for evaluation of ATM's impact on resistance to chemotherapy
In contrast, no difference in ATM-levels between PD-tumors harboring TP53 or CHEK2 mutations (Group B) and the other tumors in this cohort was found (P > 0.5); neither did we record any difference in ATM levels between mutated tumors progressing on chemotherapy and mutated tumors responding to treatment (Group B vs. Group D; P > 0.5)
In the epirubicin treated validation cohort (n
= 107, Cohort 2, Table ), 6 out of 10 patients with progressive disease on therapy previously were found mutated in the TP53
genes (Group B) [8
], limiting the number of patients with a PD despite harboring wild-type TP53
(Group A) to 4. Still, these four patients each revealed low ATM expression levels as compared to the rest of the tumors (n
= 103; P
= 0.092; Figure ), as well as when compared to the subgroups of other wild-type TP53
= 79; P
= 0.097) or tumors harboring TP53
= 25; P
To evaluate whether the effect of low ATM status was specific to DNA-damaging chemotherapy, we analyzed the predictive impact of ATM expression levels on response to paclitaxel monotherapy (Cohort 3, Table ). Here, we observed no difference in ATM expression levels between patients revealing primary resistance to paclitaxel with (n = 5) or without (n = 7) concomitant TP53/CHEK2 mutations and patients obtaining an objective response/stable disease (n = 102; P > 0.2 for both comparisons).
Low ATM expression levels may substitute for TP53/CHEK2 mutations as a cause of chemo-resistance
Postulating low ATM expression and mutations affecting TP53 and CHEK2 to be alternative mechanisms inactivating the p53 pathway (Figure ), we compared the frequency of tumors having a "hit" in this pathway (either a TP53 (L2/L3) or CHEK2 mutation or low ATM expression) among chemo-resistant versus tumors responding to anthracycline/mitomycin chemotherapy. Defining low ATM as the levels expressed by the lower 20% percentile of the patients in the cohort, a "hit" in this pathway (either low ATM expression level or a CHEK2/TP53 mutation) correlated to therapy resistance (P = 0.0267; Table ). Further, we evaluated the robustness of the model by performing repeated analysis using ATM cut-off values ranging between the 5% and 50% percentile of the cohort defining tumors with "low expression" for ATM (Figure ). Notably, the different models all revealed a statistically significant correlation between defects in the p53 pathway (defined as L2/L3-TP53/CHEK2 mutations or low level ATM expression) and therapy resistance defined as PD on treatment (P-values varying from 0.001 to 0.027; Figure ). In a multivariate analysis (logistic regression), L2/L3-TP53/CHEK2 mutations or low level ATM expression was also confirmed to be significantly associated with resistance to therapy (overall test of the model, P = 0.010).
Figure 2 Alterations in the p53 functional pathway predicts chemoresistance. (A) Schematic illustration of the three central players (ATM, Chk2 and p53), activated in response to chemotherapy induced double stranded DNA breaks. All patients in this study with (more ...)
Correlation between alterations in TP53/CHEK2/ATM ("hit") and in vivo resistance to doxorubicin/FUMI
To confirm this observation, similar analyses were performed on a validation cohort of patients having epirubicin monotherapy (Cohort 2, Table ). We confirmed the observation that low ATM expression levels or mutations affecting either TP53 (L2/L3) or CHEK2 to be associated with anthracycline resistance (P = 0.0074; Table ). Using different cut-off values classifying between 5% and 50% of the tumors as "low ATM expressors", we confirmed the model to be robust in this cohort as well (P-values varying from 0.20 to < 0.01; notably, setting a cut-off between 5% and 25%, all P-values were < 0.05; Figure ; multivariate analysis: overall test of the model; P = 0.007).
Correlation between alterations in TP53/CHEK2/ATM ("hit") and in vivo resistance to epirubicin
To evaluate whether the differences in ATM mRNA levels were reflected at the protein level, we performed immunohistochemical staining on samples from the initial cohort (Cohort 1, Table ). While most samples revealing high mRNA levels displayed strong ATM protein staining, interestingly, some samples stained strongly despite expressing low mRNA levels; thus, there was a lack of statistical correlation between ATM mRNA and IHC-staining levels (P > 0.5). However, in the group of tumors with progressive disease and wild-type TP53 and CHEK2, one tumor only was found to stain strongly for ATM.
ATM gene copy number
Out of the 70 patients from Cohort 1 sequenced for point mutations, material for MLPA-analyses was available from 66 (all included among the 69 analyzed for ATM expression levels). While no larger intragenetic deletions or duplications were observed, 9 out of the 66 tumors harbored a reduced copy number for the entire ATM locus. No association between ATM reduced copy number and either ATM mRNA levels or response to therapy was observed (P > 0.2 and P > 0.4, respectively).
ATM promoter mutations and hypermethylations
The observed difference in ATM mRNA levels could be due to different mechanisms of promoter inactivation, including mutations or hypermethylations.
Screening for potential ATM promoter hypermethylation in the patients treated with epirubicin (Cohort 2; n = 109), none displayed methylation of the ATM promoter.
Next, we sequenced the promoter region from position -661 to +105 relative to the transcriptional start site (sequence NT_033899; [32
]) in 70 tumors from Cohort 1, no mutations were recorded. The promoter area was found to exist as two distinct haplotypes differing in positions -635 (rs228589) and -10 (rs189037) relative to the transcriptional start site. Homozygosity for the NT_033899 haplotype -635A/-10G was observed in 15 patients (21.4%), 22 patients (31.4%) were homozygous for the -635T/-10A haplotype while 33 patients (47.1%) were heterozygotes. No difference in ATM expression levels between patients harboring the different genotypes were observed (P
> 0.2; Kruskal-Wallis rank test). This finding was confirmed in Cohort 2 (data not shown).
Impact of c-myc amplifications
Our data indicate that events other than promoter alterations, like deregulation of trans-acting factors, may be responsible for the alterations in ATM expression levels. N-myc induces expression of miR-421, which in turn suppresses ATM levels [33
]. However, while we found the C-myc gene to be amplified in 12 out 69 doxorubicin/FUMI-treated tumors, no correlation between C-myc amplification status and ATM expression levels were recorded (P
> 0.4; Mann-Whitney test).
ATM mRNA levels are not associated with breast cancer subclasses
An interesting question is whether low ATM expression may correlate to other tumor characteristics. Among doxorubicin/mitomycin treated tumors analyzed here, 64 have previously been classified according to gene expression profiling [34
], defining 25 and 11 tumors belonging to the Luminal A and B class, respectively, 12 belonging to the ERBB2+ class, 11 basal-like, while 5 were found belonging to the normal cell-like class. No difference in ATM expression levels between tumors belonging to the different subclasses were recorded (data not shown).
ATM mRNA levels predict overall survival in breast cancer patients
Recently, Jiang and colleges suggested low ATM levels on a TP53
wild-type background to be associated with poor survival while low ATM predicted improved survival in patients harboring TP53
]. While we detected a non-significant trend in Cohort 1 (including 69 patients only), we confirmed low ATM levels to predict a poor outcome in patients with tumors wild-type for TP53
but to improve long-term outcome among patients with tumors harboring TP53
mutations (Figure ). Further, we observed a significant differential effect of ATM levels on survival pending on TP53/CHEK2
mutation status in our larger confirmatory set (Cohort 2; Figure ; P
= 0.007; interaction between TP53
status and ATM levels on survival P
Figure 3 Impact of ATM expression levels on long-term survival. (A) Kaplan-Meier curves showing long-term survival among breast cancer patients treated with doxorubicin or a combination of 5-fluorouracil and mitomycin in the neoadjuvant setting (Cohort 1), stratified (more ...)
In order to investigate whether these effects on long-term overall survival were of general prognostic nature or related to DNA damaging chemotherapy, we performed a similar analysis on patients enrolled in the paclitaxel treatment arm (Cohort 3). Here, we observed no effect of low ATM levels on prognoses either among patients with tumors wild-type or mutated for TP53/CHEK2 (data not shown). This suggests the effect of ATM level on long-term outcome may be related to treatment with DNA-damaging agents.