cancers Article Pivotal Role of AKT2 during Dynamic Phenotypic Change of Breast Cancer Stem Cells Petra Gener 1, *,† , Diana Rafael 1,2,† , Joaquin Seras-Franzoso 1 , Anna Perez 1 , Luis Alamo Pindado 1 , Glòria Casas 1 , Diego Arango 3 , Yolanda Fernández 2,4 , Zamira V. Díaz-Riascos 1,4 , Ibane Abasolo 1,2,4 and Simó Schwartz Jr. 1,2,4, * 1 Drug Delivery and Targeting Group, Molecular Biology and Biochemistry Research Centre for Nanomedicine (CIBBIM-Nanomedicine), Vall d’Hebron Institut de Recerca, Universitat Autònoma de Barcelona, 08035 Barcelona, Spain 2 Networking Research Centre for Bioengineering, Biomaterials, and Nanomedicine (CIBER-BBN), Instituto de Salud Carlos III, 28029 Madrid, Spain 3 Biomedical Research in Digestive Tract Tumors, CIBBIM-Nanomedicine, Vall d’Hebron Institut de Recerca, Universitat Autònoma de Barcelona, 08035 Barcelona, Spain 4 Functional Validation & Preclinical Research (FVPR), CIBBIM-Nanomedicine, Vall d’Hebron Institut de Recerca, Universitat Autònoma de Barcelona, 08035 Barcelona, Spain * Correspondence: petra.gener@vhir.org (P.G.); simo.schwartz@vhir.org (S.S.J.) † These authors contributed equally to this work.  Received: 15 June 2019; Accepted: 22 July 2019; Published: 26 July 2019  Abstract: Therapeutic resistance seen in aggressive forms of breast cancer remains challenging for current treatments. More than half of the patients suffer from a disease relapse, most of them with distant metastases. Cancer maintenance, resistance to therapy, and metastatic disease seem to be sustained by the presence of cancer stem cells (CSC) within a tumor. The difficulty in targeting this subpopulation derives from their dynamic interconversion process, where CSC can differentiate to non-CSC, which in turn de-differentiate into cells with CSC properties. Using fluorescent CSC models driven by the expression of ALDH1A 1(aldehyde dehydrogenase 1A1), we confirmed this dynamic phenotypic change in MDA-MB-231 breast cancer cells and to identify Serine/Threonine Kinase 2 (AKT2) as an important player in the process. To confirm the central role of AKT2, we silenced AKT2 expression via small interfering RNA and using a chemical inhibitor (CCT128930), in both CSC and non-CSC from different cancer cell lines. Our results revealed that AKT2 inhibition effectively prevents non-CSC reversion through mesenchymal to epithelial transition, reducing invasion and colony formation ability of both, non-CSC and CSC. Further, AKT2 inhibition reduced CSC survival in low attachment conditions. Interestingly, in orthotopic tumor mouse models, high expression levels of AKT2 were detected in circulating tumor cells (CTC). These findings suggest AKT2 as a promising target for future anti-cancer therapies at three important levels: (i) Epithelial-to-mesenchymal transition (EMT) reversion and maintenance of CSC subpopulation in primary tumors, (ii) reduction of CTC and the likelihood of metastatic spread, and (iii) prevention of tumor recurrence through inhibition of CSC tumorigenic and metastatic potential. Keywords: cancer stem cells (CSC); dynamic phenotype; epithelial-to-mesenchymal transition (EMT); AKT2 targeting 1. Introduction Drug resistance in breast cancer cell populations is acquired through phenotypic changes mediated by either epithelial-to-mesenchymal transition (EMT), or tumor cell drift towards a cancer stem cell (CSC) phenotype [1,2]. As a consequence of phenotypic dynamism and clonal evolution during the Cancers 2019, 11, 1058; doi:10.3390/cancers11081058 www.mdpi.com/journal/cancers Cancers 2019, 11, 1058 2 of 18 cancer treatment, tumors contain a heterogeneous mixture of various cell types, with different stem-like and mesenchymal phenotypes. Moreover, in this intricate network made of diverse cell populations, the characteristics of mesenchymal and stem-like phenotype are often intertwined, becoming difficult to distinguish from one another [2–5]. EMT cells, as well as CSC, are described as responsible for metastatic spread, tumor progression, and resistance. CSC also possesses a tumor initiating capacity and cell renewal capacity, even though some mesenchymal cells have been shown to be able to initiate tumor growth. Moreover, the EMT conversion stimulates expression of CSC genes and vice-versa, thus CSC often expresses genes characteristic for mesenchymal cells [1,2,6]. Due to the high level of complexity, the investigation of dynamic phenotypic changes during cancer development and/or treatment has proven technically challenging, and the use of in vitro or in vivo models for this purpose are fairly limited. Permanently tagged CSC are able to mimic, at least to some extent, the complexity of phenotypic dynamism in laboratory conditions. Previously, we described CSC tagging based on an expression vector containing the tdTomato fluorescent protein under the control of ALDH1A1 (aldehyde dehydrogenase 1A1) promoter [7]. Since ALDH1A1 is expressed in CSC, this approach enables the identification and isolation of CSC, and has proved successful for the MCF7 breast cancer cell line and HCT116 colon cancer cell line [7]. Here, we use the same strategy to identify and trace CSC from the highly aggressive triple-negative breast cancer (TNBC) cell line, MDA-MB-231. TNBC presents the overall worst prognosis, greater metastatic potential, and higher relapse rate compared to other breast cancer types [8]. Sequential single-agent chemotherapy remains the standard of care for patients with metastatic TNBC, since targeted therapies have failed due to the lack of commonly targetable receptors (estrogen receptor (ER), progesterone receptor (PR), and HER2/neu). Consequently, overall survival among patients with this form of breast cancer has not changed over the past 20 years; this highlights the need for advances in therapeutic options for these patients [9,10]. Two PARP (poly ADP ribose polymerase) inhibitors (Olaparib, Talazoparib) for patients with BRCA-mutated tumors, and immunotherapy for patients with PD-L1 (programmed death-ligand 1)—positive TNBC have been recently approved, but longer follow-ups are needed to determine how these agents improve overall survival among patients with TNBC [11]. In this study, we explored the importance of the AKT2/TWIST and AKT2/mTOR axis for CSCsurvival and malignancy. Even though AKT2 has previously been identified as a promising target for cancer treatment, its role in determining CSC phenotype has not been addressed [12–14]. AKT2 was chosen for this study as the main subtype of the protein kinases B family. AKT2 has been previously shown to have a direct relationship with malignant transformation and tumor dissemination [15–17]. In contrast, cells transfected with AKT1 and AKT3 exhibit only minor alterations in cell invasiveness and these isoforms do not appear to play a role during EMT [18,19]. Early works have also demonstrated that siRNA (small interfering RNA) depletion of AKT2 results in reduced cell growth and invasiveness, with MET only occurring in cells overexpressing AKT2, confirming its potential role in tumorigenesis [20]. Even though AKT2 is not frequently over-amplified in breast cancer, its expression is linked with poor prognosis. In contrast, no correlation has been detected with AKT1 or AKT3 in terms of patients’ survival [21,22]. AKT2 is a major downstream effector of the canonical PI3-K (phosphoinositide 3-kinase) pathway, which appears generally associated with acquisition of the malignant phenotype in cancer cells [17–19]. AKT/PKB is a key regulator of various cell processes and its signaling outcome depends on cellular background and context. Similarly, all cancer-related molecular mechanisms are highly dependent on cell type and phenotype. AKT2 may act via different signaling mechanisms; TWIST and mTOR being the main downstream effectors of AKT2 [15,20,21]. Within the context of CSC, TWIST (twist family bHLH transcription factor 1) arose as a very promising candidate due to its important and well-known role in tumor invasion, migration, dissemination, and drug resistance [23,24]. Since TWIST-mediated effects are regulated by AKT2, silencing of this oncogene could be explored as a potential strategy to reduce TWIST-mediated EMT through decrease of E-CADHERIN expression [20,25]. Furthermore, mTOR (mechanistic target of rapamycin kinase) signaling activation, as a result of increased activity Cancers 2019, 11, 1058 3 of 18 of PI3K/AKT, significantly contributes to the initiation and development of tumors due to their involvement in cell growth, proliferation, motility, invasion, and survival. In this context, mTOR activity Cancers 2019, 11, x FOR PEER REVIEW 3 of 19 is found deregulated in many types of cancer including breast, prostate, lung, melanoma, bladder, brain, and development renal carcinomas of tumors[26,27]. Moreover, due to their involvementrecent in celldata also growth, suggest that proliferation, theinvasion, motility, PI3K/AKT2/mTOR and survival. In signaling pathway this context, strongly mTOR activity modulates is found deregulated CSC biology [14,28–30].inTargeting many typesthe of cancer including axis via EMT signaling breast, prostate, lung, melanoma, bladder, brain, and renal carcinomas [26,27]. Moreover, recent data AKT2/TWIST and PI3K/AKT2/mTOR in order to revert EMT and restore the epithelial phenotype also suggest that the PI3K/AKT2/mTOR signaling pathway strongly modulates CSC biology [14,28– appears to 30]. be aTargeting promising strategy the EMT in cancer signaling axis viatherapy. AKT2/TWIST and PI3K/AKT2/mTOR in order to revert EMT and restore the epithelial phenotype appears to be a promising strategy in cancer therapy. 2. Results 2. Results 2.1. Characterization and Dynamism of MDA-MB-231 CSC-Like Cell Model 2.1. Characterization and Dynamism of MDA-MB-231 CSC-Like Cell Model In order toIntag TNBC order to tagcells TNBC with cellsawith CSCa phenotype, CSC phenotype, MDA-MB-231 MDA-MB-231 cellscellswere werestably stably transfected with transfected ALDH1A1-tdTomato reporter vector, with ALDH1A1-tdTomato and reporter flow vector, andsorted based flow sorted on tdTomato based on tdTomatofluorescence (Figure 1A,B). fluorescence (Figure Consecutive1A,B). Consecutiveofcell cell passages passages+ofMDA-MB-231 tdTomato tdTomato + MDA-MB-231 cells led cells to anledinitial to an decline initial decline and and thenthen stabilization + stabilization of the tdTomato+ subpopulation (Figure 1C), which accounted for 14.7% ± 2.8% of the tdTomato subpopulation (Figure 1C), which accounted for 14.7% ± 2.8% of tdTomato cells + of within tdTomato+ cells within the MDA-MB-231-ALDH1A1/tdTomato cell line (Figure 1B). The stem like + cells was the MDA-MB-231-ALDH1A1/tdTomato cell line (Figure 1B). The stem like nature of tdTomato nature of tdTomato+ cells was confirmed by increased expression of stem cell markers compared to confirmed by increased − cells expression − tdTomato (Figure 1D). of stem cell markers compared to tdTomato cells (Figure 1D). (A) (B) (C) (D) Figure 1. MDA-MB-231-ALDH1A1/tdTomato Figure 1. MDA-MB-231-ALDH1A1/tdTomatocancer cancer stem cell stem cell (CSC) (CSC) model. model. After After transfection transfection with with reporter vector ALDH1A1/tdTomato, reporter vector ALDH1A1/tdTomato,CSC-like cells CSC-like express cells expressfluorescent reporter fluorescent reporter (tdTomato) (tdTomato) underunder the CSC-specific thepromoter CSC-specific promoter (ALDH1A1). (ALDH1A1). Scale bar Scale bar represents represents 20 µm 20 µm (A). This(A). This allows allows the CSCthe quantification CSC + quantification and sorting by fluorescence-activated cell sorting (FACS) (B). Sorted tdTomato+ cell and sorting by fluorescence-activated cell sorting (FACS) (B). Sorted tdTomato cell population dropped and stabilized over passages (C). We confirmed by qPCR, that tdTomato+ cells (CSC) express stem cell markers. Results are expressed as NRQ (relative normalized quantities) mean ± SEM (n ≥ 3); * p < 0.05; ** p < 0.01, *** p < 0.001 (D). Cancers 2019, 11, x FOR PEER REVIEW 4 of 19 population dropped and stabilized over passages (C). We confirmed by qPCR, that tdTomato+ cells (CSC) express stem cell markers. Results are expressed as NRQ (relative normalized quantities) mean ± SEM Cancers 2019, (n ≥ 3); * p < 0.05; ** p < 0.01, ***p < 0.001 (D). 11, 1058 4 of 18 Briefly, the relative increase of ALDH1A1 mRNA in tdTomato+ CSC was 4.46 ± 0.53 fold (p = Briefly,Other 0.025). the relative increase of CSC markers, ALDH1A1 such NOTCH4, mRNA in tdTomato+ALOX5, POU5F1/OCT4, 4.46 ± 0.53 CSC wasCMKLRI, fold (p = ABCG2, 0.025). and CXCR1 Other CSC markers, such NOTCH4, POU5F1/OCT4, ALOX5, CMKLRI, ABCG2, and CXCR1 were were also found over-expressed (1.78 ± 0.18, p = 0.01; 1.78 ± 0.09, p = 0.017; 2.42 ± 0.06, p = 0.04; 1.89 ±also found over-expressed 0.12, 0.08,±p 0.18, p = 0.02, 1.99 ±(1.78 p= = 0.04; 3.10.01; 1.78p ± ± 0.78, = 0.034; = 0.017; 0.09, pfold, respectively). = 0.04; 1.89 Nopdifference 2.42 ± 0.06, in ± 0.12,and CD44 p = 0.02, CD241.99 mRNA± 0.08, p = 0.04; levels were3.1 ± 0.78, pin= MDA-MB-231 observed 0.034; fold, respectively). tdTomato+ No difference cells compared in CD44 and CD24 to tdTomato − (non- mRNA levels were observed in MDA-MB-231 tdTomato + cells compared to tdTomato− (non-CSC; CSC; Figure S1A). Of note, overall expression of CD44 was high, and CD24 was low in MDA-MB-231 Figure S1A). cells Of note, according to overall expression of CD44 their mesenchymal-stem was like high,characteristics (MSL) and CD24 was(Figure low inS1B) MDA-MB-231 [31]. cells according to their mesenchymal-stem like (MSL) characteristics (Figure S1B) [31]. As expected, tdTomato MDA-MB-231 CSC-like cells were able to grow as mammospheres in + As expected, tdTomato + MDA-MB-231 CSC-like cells were able to grow as mammospheres in low low attachment cell culture conditions, when seeded in a medium without serum (Figure 2). attachment cell culture conditions, when seeded in a medium without serum (Figure 2). (A) (B) Figure 2. Mammosphere culture of tdTomato+ cells. tdTomato+ CSC were able to form mammospheres in low attachment plates, in serum free media. Scale bar represents 100 µm (A) tdTomato+ cells formed Figure 2. significantly Mammosphere more mammospheres culture than tdTomato − cells+ (B). of tdTomato cells. < 0.001 + CSC were able to form *** ptdTomato mammospheres in low attachment plates, in serum free media. Scale bar represents 100 µm (A) Nevertheless, the tdTomato − (non-CSC) cells were also able to grow −in these conditions but tdTomato+ cells formed significantly more mammospheres than tdTomato cells (B). ***p < 0.001 formed significantly fewer mammospheres than equivalent amounts of tdTomato+ cells (Figure 2, Table 1). Nevertheless, While tdTomato + CSC-like the tdTomato − (non-CSC) cells cells start to form thewere also able towhen mammospheres growjustin 10 these conditions cells/well were but formed seeded significantlyincidence (mammosphere = 6/12; 50 (n/n, (%))), fewer mammospheres than it equivalent amounts was necessary of at to seed tdTomato cells (Figure least 100+ cells/well of 2, Table −1). tdTomato While tdTomato non-CSC + CSC-like cells to induce cells startformation mammosphere to form the mammospheres (mammosphere incidence = 1/8; when just 1012cells/well (n/n, (%)).were seeded 12/12; In contrast, (mammosphere incidence 100 (n/n, (%)) = 6/12; of incidence was50observed (n/n, (%))), it was when necessarycells 100 CSC-like to seed wereatseeded least 100 cells/well per well (Tableof1). tdTomato− non-CSC cells to induce mammosphere formation (mammosphere incidence = 1/8; 12 (n/n, (%)). In contrast, 12/12; 100 (n/n, (%)) of incidence was observed when 100 CSC-like cells were seeded per well Table 1. Incidence (Table 1). of mammosphere formation. Initial Number of Cells Stemness 1 3 10 30 100 300 1000 Per Well Type CSC-like − − + + + + + Mammosphere formation Non-CSC − − − − + + + Incidence of mammosphere CSC-like N/A N/A 6/12 8/12 12/12 12/12 12/12 formation (n/n) Non-CSC N/A N/A N/A N/A 1/8 4/10 7/12 Incidence of mammosphere CSC-like N/A N/A 50% 67% 100% 100% 100% formation (%) Non-CSC N/A N/A N/A N/A 12% 50% 58% n/n—number of wells with mammospheres per number of wells in total; N/A—not applicable. Interestingly, when only the tdTomato− cell population was cultured in non-attachment conditions; tdTomato+ cells appeared from day 7 within the cell culture (Figure 3). This was not due to a fast proliferation of some residual CSC; since it was possible to observe a sudden switch to tdTomato+ Cancers 2019, 11, 1058 5 of 18 phenotype of the whole mammosphere (Figure S2). These “new” tdTomato+ cells expressed a significantly higher amount of stem cell markers (ALDH1A1, NOTCH4, POU5F1/OCT4, ALOX5, and CMKLR1) and had increased invasion potential, compared to sorted tdTomato− cells (Figure 3D,E). Interestingly, the increase of tdTomato and stem cell gene expression in originally tdTomato− cells, was just transient, since when “new” tdTomato+ CSC were cultured back in attachment conditions, in complete medium (four days), the expression of tdTomato and stem cell markers decreased back to normal and just few tdTomato+ cells (5.6%) were detected by flow cytometry (Figure 3). Analogously, we observed de novo tdTomato expression and stem cell gene expression, as well as increased invasion capacity, when using tdTomato− cells sorted from an aggressive HCT8 colon cancer cell line (Figure S3). OfCancers 2019, 11, x FOR note, tdTomato PEER REVIEW − MCF7 epithelial cells did not survive in mammosphere culture conditions. 6 of 19 (A) (B) (C) (D) (E) Figure 3. MDA-MB-231 reversion process. Schematic representation of the experiment. Briefly, tdTomato− (non-CSC) were placed in low attachment plates in FBS (Fetal Bovine Serum)-free culture medium and were cultured for 7–17 days. During this time cells acquired the CSC phenotype. Afterwards, tdTomato+ Figure 3. MDA-MB-231 reversion process. Schematic representation of the experiment. Briefly, cells were plated back into adherent plates in complete culture medium and were allowed to grow for tdTomato− (non-CSC) were placed in low attachment plates in FBS (Fetal Bovine Serum)-free culture another four days (A). The cells were monitored over the time by fluorescent microscopy (B) and flow medium and were cultured for 7–17 days. During this time cells acquired the CSC phenotype. cytometry (C). The stem cell properties of “new” tdTomato+ cells were confirmed by a change in stem Afterwards, tdTomato+ cells were plated back into adherent plates in complete culture medium and cell gene expression profile, measured by qPCR (D) and by an increase of invasive ability (E). Results were allowed to grow for another four days (A). The cells were monitored over the time by fluorescent are expressed as NRQ (relative normalized quantities) mean ± SEM (n ≥ 3); (* p < 0.05; ** p < 0.01, microscopy (B) and flow cytometry (C). The stem cell properties of “new” tdTomato+ cells were *** p < 0.001). confirmed by a change in stem cell gene expression profile, measured by qPCR (D) and by an increase In of invasive order ability (E). Results to determine the tumorare expressed initiatingascapacity NRQ (relative normalized quantities) of MDA-MB-231 tdTomato + cells, mean ± SEM (n sorted ≥ + 3); (* p < 0.05; ** p < 0.01, − *** p < 0.001). tdTomato , and tdTomato cells were xenotransplanted into mice. Two groups (eight animals each) Cancers 2019, 11, 1058 6 of 18 were injected with 1000 cells into their flanks, and tumors were allowed to grow for 71 days. Tumors were detected in 8/8 animals in the group of tdTomato+ CSC, as expected, while just in 6/8 animals that were xenotransplanted. In both cases, obtained tumor volumes were similar (Figure 4C). Ex-vivo analysis of tumors showed expression of tdTomato (59.8% ± 5.3%) in all tumors of the CSC-like group. However, the presence of tdTomato+ (17.79%) cells was confirmed by immunohistochemistry and flow Cancers 2019, cytometry 11, in just x FOR 1/6 PEER in animals injected with tdTomato− non-CSC cells (Figure 4). REVIEW tumors 7 of 19 (A) (B) (C) Figure4.4.MDA-MB-231 Figure MDA-MB-231ininvivo vivomodel. model.Tumor Tumorgrowth growthandandKaplan–Meier Kaplan–Meiersurvival survivalcurves curvesofofmice mice inoculated inoculated inoculated inoculated orthotopically orthotopically into into the the right right mammary mammary fad pad fad pad (IMFP) (IMFP) with MDA-MB- with MDA-MB-231dTomato − 231dTomato + non-CSC (n = 6), and MDA-MB-231dTomato CSC (n = 8) cell variants was monitored over time. − non-CSC (n = 6), and MDA-MB-231dTomato + CSC (n = 8) cell variants was monitored Noover time. Nosignificant statistically statistically significant differences weredifferences detected (A).were detectedof (A). Summary Summary information of information regarding tumor regarding incidence, tumor tumor incidence, growth, tumor growth, and tdTomato and tdTomato and TWIST expressionand TWIST within expression the tumors within theoftumors (B). Examples IHC (B). Examples of IHC (immunohistochemistry) (immunohistochemistry) (HE/td Tomato) (HE/td Tomato) in representative tumors ofinanimals representative inoculatedtumors with of animals CSC-like inoculated cells, with CSC-like and non-CSC (tdTomato cells, and non-CSC positive as well as(tdTomato the tdTomatopositive as well negative as the Magnification example). tdTomato negative bar example).100 represents Magnification µm (C). bar represents 100 μm (C). 2.2. 2.2.Involvement InvolvementofofAKT2 AKT2ininDynamism DynamismofofEMT EMTand andCSC-Like CSC-LikePhenotype Phenotype The − Theobservation observationofofrelatively relativelyhigh highincidence incidenceofof tumors tumors derived derived from fromMDA-MB-231 MDA-MB-231 tdTomato tdTomato− cells, + cells,with lowlow with incidence of “new” incidence tdTomato of “new” tdTomatoCSC-like cells, made + CSC-like cells, us wonder made whether whether us wonder there is another there is type of phenotypic dynamism, for instance EMT, involved in the process of tumorigenesis. another type of phenotypic dynamism, for instance EMT, involved in the process of tumorigenesis. Notably, we observedwe Notably, more EMT genes observed more(low EMTE-CADHERIN, genes (lowhigh FIBRONECTIN, E-CADHERIN, highand high VIMENTIN) FIBRONECTIN, andin the high tumors derived from non-CSC compared to tumors derived from CSC-like tdTomato + cells (Figure 5A). VIMENTIN) in the tumors derived from non-CSC compared to tumors derived from CSC-like NOTCH4, tdTomatoan important + cells EMT (Figure 5A).and CSC marker, NOTCH4, was also EMT an important highlyandover-expressed CSC marker,inwasthe also tumors derived highly over- expressed in the tumors derived from MDA-MB-231 non-CSC. Besides, mTOR, an AKT2 downstream gene was also over-expressed in these tumors, compared to tumors derived from tdTomato+ cells (Figure 5A). Cancers 2019, 11, 1058 7 of 18 from MDA-MB-231 Cancers non-CSC. 2019, 11, x FOR Besides, mTOR, an AKT2 downstream gene was also over-expressed8inof 19 PEER REVIEW these tumors, compared to tumors derived from tdTomato+ cells (Figure 5A). (A) (B) (C) (D) Figure 5. AKT2/TWIST/mTOR axis in vivo and in vitro. Gene expression in the tumors derived from Figure non-CSC5. AKT2/TWIST/mTOR axis in from relative to tumors derived vivo and in vitro. CSC-like tdTomato + cells (A). in Gene expression the tumorscurve Dissociation derived from plot non-CSC confirmed thatrelative tumorsto tumors TWIST, expressed derivedevenfromthough CSC-like tdTomato the injected cellscells + did (A). Dissociation not express TWISTcurve per seplot (B). confirmed that tumors Gene expression expressed in circulating tumorTWIST, cells even though (CTCs) thefrom isolated injected cells did the blood not express of tumor bearingTWIST per se animals (B). to relative Gene geneexpression expression in circulating of tumors tumorexpression (C). Gene cells (CTCs) in lowisolated fromculture attachment the blood of tumor conditions, bearing relative animals to the relative same cells to gene cultured expression in monolayer of qPCR (D). tumors (C). are results Gene expression expressed as NRQin low attachment (relative culture normalized conditions, quantities) mean relative ± SEMto(nthe same ≥ 3); < 0.05; (* p cells ** p < 0.01). cultured in monolayer (D). qPCR results are expressed as NRQ (relative normalized quantities) mean ± SEM (n ≥ 3); (* p < 0.05; ** p < 0.01). TWIST, an AKT downstream gene, has been previously described as responsible for mesenchymal TWIST, to epithelia an Even switch. AKTthoughdownstream gene, has been injected MDA-MB-231 cellspreviously completelydescribed lack TWISTasexpression, responsible5/6 for mesenchymal (83%) tumors derived to epithelia switch.−Even from tdTomato though injected MDA-MB-231 MDA-MB-231 cells expressed TWISTcells completely (Figure lack 4B). In the TWIST group expression, of CSC-like 5/6 (83%) injected animals,tumors derived just 3/7 (42%)from expresstdTomato TWIST −inMDA-MB-231 cells expressed the resultant tumors (Figure 4,TWIST Figure(Figure 5B). AKT24B). In theexpression mRNA group of CSC-like level wasinjected howeveranimals, not changedjust 3/7 (42%) within the express differentTWIST groupsin the resultant (Figure tumors S4). Of note, we (Figure detected4,a Figure 5B). increase significant AKT2 mRNA of AKT2 expression mRNA expression (FC = 2.3 not level was however p = 0.01) changed ± 0.32, within the different in circulating groups tumor cells(Figure S4). Of note, (CTC) isolated fromwe thedetected blood ofatumorsignificant increase bearing animals.of AKT2 mRNACTCs In addition, expression derived (FC = 2.3 ± from 0.32, also tumors p = 0.01) in circulating expressed high levelstumorof mTOR (FC =isolated cells (CTC) 2.3 ± 0.26, = 0.02) fromp the bloodandofTWIST (FC = 2.6animals. tumor bearing ± 0.28, In p = addition, CTCs derived 0.03) compared to primary from tumors tumors also expressed (Figure 5C). This washighaccompanied levels of mTOR (FC = 2.3 ±decrease by significant 0.26, p = of 0.02) and TWIST expression E-CADHERIN (FC =p 0.07 (FC = 2.6 ± 0.28, = 0.03) p = 0.02) compared ± 0.13, to primary tumors (Figure 5C). This was accompanied (Figure 5C). byFurther significant decrease exploration of of theE-CADHERIN AKT2/TWIST/mTOR expressionaxis(FC = 0.07 ± 0.13, p =resembling in mammospheres, 0.02) (Figure 5C).in vitro, CTCs revealedFurther exploration MDA-MB-231 cells inoflow the attachment AKT2/TWIST/mTOR axis in mammospheres, expressed significantly resembling higher level of = 2.4 in AKT2 (FC CTCs vitro, ± 0.19, p =revealed 0.00003)MDA-MB-231 and mTOR (FC = 1.9 cells in ± low 0.05, p = 0.0003), attachment expressed compared significantly higher with the same level cells of AKT2 cultured (FC = 2.4 ± (Figure in monolayer 0.19, p =5D). 0.00003) and mTOR Moreover, these (FCgene= expression 1.9 ± 0.05, plevels = 0.0003), compared decreased whenwith the the same cells werecells cultured in monolayer (Figure 5D). Moreover, these gene expression levels decreased when the cells were cultured again in attachment conditions (Figure 5D). Expression of TWIST mRNA was not detected in these conditions. Cancers 2019, 11, 1058 8 of 18 cultured again in attachment conditions (Figure 5D). Expression of TWIST mRNA was not detected in these conditions. Interestingly, AKT2 expression was not altered in CSC fraction either in MDA-MB-231 cells or in MCF7 cells. In fact, none of the AKT isoforms (AKT1, AKT2, and AKT3) were differentially expressed in the CSC fraction compared with the non-CSC fraction (Figure S5, Text S2). 2.3. Consequences of AKT2 Inhibition for CSC-Like/EMT Phenotype The increased expression of EMT genes (AKT2, TWIST, and mTOR) in tdTomato− tumors, in CTCs and in mammospheres, highlights the importance of AKT2 in tumor growth and CSC malignancy. To pursue this line of investigation further, we explored the consequences of AKT2 silencing in three different cancer cell lines, and their CSC-like fractions (MDA-MB-231, HCT8, and MCF7). Even though the suitability of AKT2 targeting in general terms is well known, its role in CSC malignancy has not been elucidated yet [17–19,25,32]. Since AKT2 is known to play an important role in EMT, and is a likely candidate for driving the CSC phenotype, we first characterized mesenchymal gene expression levels in the investigated cells lines. CSC isolated from the mesenchymal cell lines MDA-MB-231 did not differ in EMT gene expression that was high in both “bulk” fraction as well as in CSC-like fraction (Figure 6A). In contrast, CSC isolated from epithelial cell lines (MCF7, HCT8) showed increased mesenchymal phenotype (low E-CADHERIN, high VIMENTIN, and FIBRONECTIN) compared to their respective “bulk” counterparts (Figure 6A). In addition, mesenchymal, stem-like cell lines SKBR3 and MDA-MB-468 were included in the analysis, as well. Specific inhibition of AKT2 by siRNA reduced AKT2 mRNA expression by 80% in all cell lines used, while no effect on AKT1 isoform expression was detected (Figure S6). Downregulation of AKT2 was also confirmed at the protein level by Western Blot (WB) (Figure 7). As expected, sorted CSC subpopulations showed significantly higher capacity for anchorage- independent growth compared to non-CSC fractions, further confirming the CSC nature of tdTomato+ cells (Figure 6B). Similarly, CSC from MDA-MB-231 and MCF7 showed significantly higher invasive potential. No difference in invasion was detected between CSC and non-CSC in HCT8 cell line (Figure 6C). For MDA-MB-231 and HCT8, AKT2 silencing significantly reduced the number of cells able to grow without anchorage, and the invasive potential of both CSC and non-CSC. In the case of MCF7, anchorage-independent growth capacity and invasive ability was strongly reduced after AKT2 silencing only in the CSC subpopulation (Figure 6B,C). Moreover, anchorage-independent growth and invasion were significantly reduced also in breast cancer cell lines with clear mesenchymal phenotype; with high expression of ALDH1A1, (MDA-MB-468 and SKBR3) after AKT2 silencing (Figure S7A,B). Furthermore, we observed a significant impairment of cell migration and proliferation of the mesenchymal cells in which AKT2 was silenced, in MDA-MB-231, MDA-MB-468, and SKBR3 cell lines (Figure S7C,D). In contrast, no difference in the migration and proliferation of “bulk” epithelial MCF7 cells was observed after AKT2 silencing (Figure S8B,C). To explain mechanistically the downstream effect of AKT2 silencing in different cell types, and in CSC-like cells derived from these cell lines, is however not straightforward. We observed different pathway involvement depending on the cellular background. In mesenchymal (MDA-MB-231) as well as epithelial (HCT8) cell lines that do not express TWIST in vitro, the predominant effect on AKT2 inhibition was de-regulation of mTOR signaling (p.53, GSK3b high) (Figure 7A), whereas in mesenchymal cell lines that do express TWIST (MDA-MB-468, SKBR3), and in mesenchymal CSC isolated from epithelial MCF7 cells, AKT2 inhibition caused the reversion of mesenchymal to epithelial phenotype (E-CADHERIN high, and N-CADHERIN, VIMENTIN, and FIBRONECTIN LOW; Figure 7B,C). No clear effects on EMT genes and mTOR pathway signaling were observed in MCF7 bulk cells, even though TWIST and mTOR mRNA were downregulated in these cells after AKT2 inhibition (Figure S8A). Cancers 2019, 11, 1058 9 of 18 Cancers 2019, 11, x FOR PEER REVIEW 10 of 19 (A) (B) (C) Figure EffectsofofAKT2 Figure 6.6. Effects AKT2 silencing silencing in MDA-MB-231, in MDA-MB-231, HCT-8, HCT-8, and MCF-7 and MCF-7 CSC. CSC. MDA-MB-231, MDA-MB-231, HCT- HCT-8, and MCF-7 CSC were separated from non-CSCs and the relative EMT gene 8, and MCF-7 CSC were separated from non-CSCs and the relative EMT gene expression profile was expression profile was quantified by qPCR. The presented values are relative to their non-CSC population quantified by qPCR. The presented values are relative to their non-CSC population (A). Anchorage- (A). Anchorage-independent growth assay: The graph represents the number of cells capable of independent growth assay: The graph represents the number of cells capable of anchorage- anchorage-independent growth, comparing CSCs and non-CSC transfected with siAKT2 and control independent growth, comparing CSCs and non-CSC transfected with siAKT2 and control siRNA siRNA (siC). Images are photographs of colonies formed after seven days of incubation with siAKT2 (siC). Images are photographs of colonies formed after seven days of incubation with siAKT2 and siC and siC (B). Invasion assay: The graph represents the number of cells that demonstrated invasive (B). Invasion assay: The graph represents the number of cells that demonstrated invasive capacity capacity comparing CSC and non-CSC transfected with siAKT2 and siC (C). Results are expressed as comparing CSC and non-CSC transfected with siAKT2 and siC (C). Results are expressed as NRQ NRQ (relative normalized quantities) mean ± SEM (n ≥ 3); (* p < 0.05; ** p < 0.01, *** p < 0.001). (relative normalized quantities) mean ± SEM (n ≥ 3); (* p < 0.05; ** p < 0.01, *** p < 0.001). Cancers 2019,11, Cancers2019, 11,1058 x FOR PEER REVIEW 10 of 11 of 18 19 (A) (B) (C) Figure 7. AKT2 silencing on gene expression. qPCR quantification of gene expression levels of key Figure 7. AKT2 regulators in the silencing on genepathway mTOR signaling expression. qPCR in cells quantification that do not express of TWIST gene expression levels HCT8; (MDA-MB-231, of key regulators in the mTOR signaling pathway in cells that do not express TWIST (MDA-MB-231, (A) and epithelial-to-mesenchymal transition (EMT) signaling pathway in mesenchymal cells that do HCT8; (A) and TWIST express epithelial-to-mesenchymal (MDA-MB-468, SKBR3; transition (B) and (EMT) signaling (MCF-7 CSC)pathway after thein mesenchymal treatment cells that(C). with siAKT2 do express The TWISTvalues presented (MDA-MB-468, SKBR3; are relative (B)transfected to cells and (MCF-7with CSC)siCafter andtheare treatment expressedwith assiAKT2 (C). The NRQ (relative presented values normalized are mean quantities) relative to cells ± SEM 3); (* p < 0.05; (n ≥transfected with < 0.01, ** psiC and*** p