Background
Metastatic breast cancer is a major cause of global cancer mortality and, despite several advances in recent years, is still largely incurable [
1]. Critically, little progress has been made in the past decade in the evolution of chemotherapeutic or endocrine therapies to improve overall survival in patients. Nevertheless, targeted therapies, such as those directed against tumors overexpressing human epidermal growth factor receptor 2 (HER2), have improved patient outcomes [
2]. Moreover, molecular-characterization studies in breast cancer have revealed that, in addition to HER2 amplification, tumors may possess numerous other genomic alterations located in oncogenes or tumor suppressor genes [
3,
4]. As specific oncogenic events may be blocked by targeted therapies, screening for targetable genomic alterations may help to identify subpopulations of patients for whom specific targeted therapy would be beneficial.
One such targetable alteration resides in the v-akt murine thymoma viral oncogene (
AKT).
AKT1 is a member of the serine-threonine kinase class that plays a key role in cellular processes, including growth, proliferation, survival, and angiogenesis. It is a downstream mediator of phosphatidylinositol 3-kinase which, along with
AKT1, is a key mediator of proliferation and survival pathways frequently activated in cancer [
5‐
10]. Tumors from patients with breast, colorectal, ovarian, and leukemic cancers have been shown to harbor activating somatic mutations in
AKT1 [
5,
9]. The activation of
AKT1 is driven by membrane localization which, in turn, is initiated by the binding of the pleckstrin homology domain to phosphatidylinositol-3,4,5-trisphosphate or phosphatidylinositol-3,4-bisphosphate, followed by phosphorylation of the regulatory amino acids serine 473 and threonine 308 [
7,
11].
Genetic mutations in the
AKT pleckstrin homology domain have been reported to disturb the localization behavior and loss of sensitivity towards phosphatidylinositols, and to have major consequences in
AKT functional behavior [
5]. For instance, a somatic point mutation at nucleotide 49 introduces a lysine substitution for glutamic acid at amino acid 17 (
AKT1
E17K), resulting in a pathologic association of
AKT1 with the plasma membrane and constitutive activation of the enzyme which, in turn, results in an increased level of
AKT phosphorylation and downstream molecules independent of upstream, e.g. stimulation of growth factor.
For breast cancer patients,
AKT1
E17K mutation frequencies between 1.4 % and 8.2 %, with a mean mutation frequency of 3.8 %, have been described [
12]. Moreover, the
AKT1
E17K mutation appears to be restricted to ductal and lobular histotypes, and hormone receptor (HR)-positive breast tumors [
13‐
15]. Interestingly, higher incidences of
AKT1
E17K mutations have been reported to occur in benign papillomas (33 %; 20/61 [defined as papillomas without atypia]), compared with papillary carcinoma (10 %; 1/10) [
16].
Several studies have indicated
PTEN,
PIK3CA, and
AKT1 mutations to be mutually exclusive (i.e. not co-occurring in the same tumor tissue sample) in individual tumors [
5,
13,
17], suggesting that mutational activation of the phosphatidylinositol 3-kinase pathway by any one of these means is biologically equivalent. Alterations in all three are considered to be potential drivers of human breast cancer [
4,
18]. However, particularly for
AKT1
E17K mutation, the precise role in cancer development and progression in the clinical context is still largely unknown.
To better understand the role of the AKT1
E17K mutation in breast cancer, more than 600 tumor samples from breast cancer patients were profiled for presence of the AKT1
E17K mutation using Beads, Emulsions, Amplification, and Magnetics technology (BEAMing; Sysmex Inostics GmbH, Hamburg, Germany) in tissue and circulating tumor DNA (ctDNA). Additionally, targeted exome sequencing was conducted on tumor tissues to reveal any co-existence of the AKT1
E17K mutation with other oncogenic alterations.
Discussion
The single hotspot mutation in the pleckstrin homology domain of the
AKT1 gene [G49A:E17K] has been described in human breast, colon, and ovarian cancers, with the highest incidence observed in breast cancer [
5]. Including all recently published studies, the reported frequency of this mutation ranges from 1.4 % to 12.5 %, with a mean frequency of 3.1 % [
4,
5,
9,
13,
14,
24‐
37] (see in Additional file
1: Table S3). However, the majority of these studies (14/19) had relatively small sample sizes (under 200) and were associated with broad confidence intervals. While meta-analysis can support the identification of actual mutation frequencies, drawing conclusions about any association of mutation frequency and clinical parameters in this way is challenging, as available clinical information varies widely across studies.
We assessed the prevalence of the
AKT1
E17K mutation in a large breast cancer cohort (over 600 cases, cohort A). Our data indicate a prevalence of 6.3 % (39/619; 95 % confidence interval 4.5–8.54) using BEAMing technology on tumor tissue, which is in agreement with previous studies (see in Additional file
1: Table S3). Assessment of ctDNA by BEAMing has been shown to reliably facilitate analysis of a cancer patient’s mutational status [
38,
39]. However, using matched liquid biopsy samples, approximately 6 % prevalence could not be confirmed in this cohort, and the mutation capture rate in ctDNA was only 11.4 % (4/35). We obtained similar results for
PIK3CA mutation. In samples from patients with advanced metastatic breast cancer (cohorts B and C), we obtained 98.0 % and 97.1 % concordance and an
AKT1
E17K mutation capture rate of 66.7 % and 85.7 % for ctDNA, and 100 % concordance and a
PIK3CA mutation capture rate of 100 % (cohort B). This is in line with concordance for
AKT1
E17K (100 % in three patients) observed by Perkins et al. in patients eligible for a phase I study [
25]. Interestingly, tissue samples from two patients in our cohort were collected several years before blood samples, thereby providing the first hint that the
AKT1
E17K mutation is stable during disease development.
Blood samples for cohort A were serum, compared with plasma used in cohorts B and C. Thus, differences in mutation detection based on sample type cannot be excluded. However, considering all data obtained so far, the low mutation capture rate for ctDNA within cohort A is likely based on the fact that the majority of patients had primary-diagnosed early-stage disease without previous treatment (approximately 55 % for UICC stages I and II, and 7 % for UICC stage IV). Thus, these data are in line with previous studies indicating a stage-dependent limitation for mutation detection by ctDNA profiling [
40‐
43].
The
AKT1
E17K prevalence determined in our study was not correlated with age or menopausal stage. In line with findings described by Stemke-Hale et al. [
14], we did not find any tumors expressing HER2 at a level of IHC-Score 3+ also bearing the
AKT1
E17K mutation (0/37). Interestingly, we identified one HR-negative patient as having mutated
AKT1
E17K (1/10; 3/20 for HR-positive) within the group of relapsed cases. Thus, it is possible that
AKT1
E17K mutation may not be restricted to HR-positive breast cancer [
14]. However, any predominance in HR-positive breast cancer as described for
PIK3CA cannot be excluded because of the low number of cases in our HR-negative cohort [
26].
It has been described that
AKT1
E17K mutation could not be found in medullary and mucinous histotypes and is restricted to ductal and lobular histotypes [
13,
15]. In contrast to this, we found the mutation also in a relapsed patient with mucinous carcinoma (grade 1, ER/progesterone receptor unknown). Thus, our data support the assertion that large cohorts containing sufficient numbers in each clinical subgroup are required to reliably evaluate mutation association with histotype or other clinical parameters. Furthermore, in this cohort, prevalence of the
AKT1
E17K mutation was lower in patients with grade 3 disease compared with those with grade 1 or grade 2 disease. Comparable data were obtained for
PIK3CA mutations in breast cancer patients [
26].
This leads to the question as to why
AKT1
E17K prevalence is reduced in the more advanced disease setting. It has been hypothesized that
AKT activation confers a selective advantage during early HR-positive tumorigenesis but inhibits tumor dissemination during progression [
14]. This is supported by the observation that over-activation of
AKT drives initiation of tumorigenesis but inhibits invasion and metastasis in an ERBB-2-induced mammary tumor model [
44]. Accordingly,
AKT1
E17K-mutant clones could play a role during early tumorigenesis and – contradictory to the assumption of clonal stability – be overgrown during disease progression from grade 1 to 3 by other clones.
On the other hand, assuming that tumors bearing the AKT1
E17K mutation are rather more aggressive and rapidly growing, one could hypothesize that patients bearing this tumor phenotype could present earlier and hence be diagnosed at an early disease stage (e.g. UICC stage I/grade 1), as inferred in the present work. Consequently, patients bearing the AKT1
E17K mutation would already be undergoing therapy when their disease reached UICC stage IV/grade 3 and would therefore not be identifiable in a first-diagnosis cohort as used herein. However, they would be identifiable in a cohort not limited to first-diagnosis patients corresponding to our cohort B (3 AKT1
E17K-mutated patients out of 50 according to tissue analysis) or relapsed patients (5/45). Compared with patients with wild-type AKT1, patients with AKT1
E17K mutations would also therefore be expected to relapse sooner and/or sustain disease-related death. Long-term follow-up data from patients are required in order to confirm this hypothesis.
Although low sample numbers precluded comprehensive analysis, initial follow-up survival data indicate increased death rates in early-stage disease (UICC stages I–III) for patients with mutant
AKT1
E17K versus wild type. Thus,
AKT1
E17K could be a negative prognostic factor. However, in contrast to this, it has been reported that disease did not recur in six
AKT1
E17K-mutant patients [
14]. Further collection of follow-up data in collaboration with PATH Biobank is planned. This will facilitate confirmation of the impact of
AKT1
E17K (and also other mutations measured with the FoundationOne® panel) on disease progression or survival.
It has been described that
PIK3CA and
AKT1 mutations are not co-occurring in individual tumors [
5]. However, the sample size was insufficient to document statistical significance, and an exception (based on one case only) has been noted in another study where the patient bore both
PIK3CA and
AKT1 mutations [
13]. Our data clearly indicate that both mutations can co-exist within one tumor. However,
PIK3CA mutations are rather under-represented in
AKT1
E17K-mutant compared with wild-type samples. Moreover, in
AKT1-mutant samples, the
PIK3CA allele frequencies were lower than the
AKT1
E17K allele frequency. Thus, it is possible that the
AKT1
E17K mutation occurred earlier in development in these specific cancer samples. Of course, since our analysis was performed on the entire tumor tissue, the possibility that
AKT1 and
PIK3CA mutations existed in different cells within the tumor could not be excluded.
AKT1
E17K may represent a
bona fide oncogene in the context of human luminal breast cancer. Recently, knock-in of the mutation into a luminal breast cancer cell-line model against a
PIK3CA wild-type background was shown to restore pathway signaling, proliferation, and tumor growth
in vivo [
18]. Our data support the oncogenic role of
AKT1
E17K, as 16.2 % (6/37) of all
AKT1-mutant patients were identified as having no other mutation or genetic alterations known to drive their disease or which were likely involved in disease generation or progression. Thus,
AKT1
E17K in these patients was the most likely disease driver. To date, there are several clinical programs in place addressing
AKT as a potential therapeutic target [
45‐
47]. ATP-competitive inhibitors and allosteric inhibitors have shown promising efficacy in
AKT1-mutant cancers in preclinical studies [
48].
Acknowledgments
We thank Sibylle Walter and Teresa Lunt for technical assistance, Arndt Schmitz for providing Biobank capacities, and Michael Teufel, Henrik Seidel, and Eleni Lagkadinou for critical discussions. We also thank the breast cancer centers in Germany working for PATH Biobank as sample source sites for their help in allocating samples and data (Brust Zentrum am St.-Johannes-Hosptial in Dortmund; IBZ Brustzentrum am Klinikum Kassel; Brustzentrum Regio am Universitätsklinikum Marburg; Brustzentrum am Sana Klinikum Offenbach; Brustzentrum Bochum/Herne am St. Anna Hospital; Brustzentrum Johanniter-Krankenhaus-Bonn; Brustzentrum Regensburg am Caritas-Krankenhaus St. Josef; for details visit:
http://path-biobank.org/index.php/en/about-path/path-cooperative-clinics. We thank all patients who were willing to donate their samples—without their support the research work would not be possible.
The authors take full responsibility for the scope, direction, and content of the manuscript and have approved the submitted manuscript. They thank Kieran Davey, PhD, at Complete HealthVizion for his assistance in the revision of the manuscript, based on detailed discussion and feedback from the authors. Editorial assistance was funded by Bayer HealthCare Pharmaceuticals.