Signals downstream of Akt can either favor or oppose stem cell (SC) maintenance, but how this dual role can be achieved is still undefined. Using human limbal keratinocyte stem cells (LKSCs), a SC type used in transplantation therapies for corneal regeneration, we show that Akt signaling is prominent in SC populations both in vivo and in vitro, and that Akt1 promotes while Akt2 opposes SC self-renewal. Noteworthy, loss of Akt2 signaling enhances LKSC maintenance ex vivo, whereas Akt1 depletion anticipates SC exhaustion. Mechanistically, the antagonistic functions of Akt1 and Akt2 in SC control are mainly dictated by their differential subcellular distribution, being nuclear Akt2 selectively implicated in FOXO inhibition. Akt2 downregulation favors LKSC maintenance as a result of a gain of FOXO functions, which attenuates the mechanistic target of rapamycin complex one signaling via tuberous sclerosis one gene induction, and promotes growth factor signaling through Akt1. Consistently, Akt2 deficiency also enhances limbal SCs in vivo. Thus, our findings reveal distinct roles for nuclear versus cytosolic Akt signaling in normal epithelial SC control and suggest that the selective Akt2 inhibition may provide novel pharmacological strategies for human LKSC expansion in therapeutic settings and mechanistic research. Stem Cells2014;32:754–769
The cornea is the most important light refracting structure of the eye, and the integrity of its stratified epithelium is key for visual acuity. In adult humans, the continuous renewal of the corneal epithelium relies on keratinocyte stem cells (SCs) located in the basal layer of the limbus, the narrow transitional zone separating the cornea from the bulbar conjunctiva [1-7]. The similarities between limbal and epidermal keratinocytes allowed the establishment of methods to propagate human limbal keratinocyte stem cells (LKSCs) ex vivo for therapeutic transplants to treat limbal SC deficiencies, a group of diverse pathological conditions leading to severe visual impairment [2, 4, 8-12]. Culture of human limbal keratinocytes (LKs) originate three types of colonies, namely, holoclones, meroclones, and paraclones [9, 10], being holoclones the exclusive cell populations retaining SC features such as long-term self-renewal and the capacity to regenerate tissue in vivo [13, 14]. Meroclones and paraclones, instead, lack long-term self-renewal ability and raise progenies of transit-amplifying cells. Limbal holoclones are characterized by elevated ΔNp63α protein expression [15, 16], the functions of which are essential for the maintenance of stratified epithelial SCs proliferative potential . Since LKSCs are among the few cultured human SC types currently used in regenerative medicine, and the majority of failures of their therapeutic applications are due to a paucity of SCs [2, 14], the identification of signaling pathways that could be pharmacologically targeted to enhance SC self-renewal ex vivo is of key clinical significance. However, the signaling pathways underlying LKSC self-renewal are still largely undefined.
The phosphoinositide 3-kinase (PI3K)/Akt signaling network  has emerged as central player in SC biology. While this signaling module promotes self-renewal and pluripotency in embryonic stem (ES) cells [19-22], activation of the Akt pathway following phosphatase and tensin homolog (PTEN) loss in adult tissues can either lead to SC depletion [23, 24] or enhancement [25, 26]. Akt signaling also favors the execution of epidermal keratinocyte terminal differentiation program [27-29], but its role in epithelial SC decisions is poorly defined. Among the three Akt isoforms , the majority of tissues expresses predominantly Akt1 and Akt2, which play both redundant and specific functions, with Akt1 being primarily involved in cell growth and proliferation and Akt2 in glucose metabolism control [31-36]. It is currently unknown whether the dual role of Akt signaling in SC regulation relies in part on selective Akt isoform functions. Interestingly, during tumor development, Akt1 and Akt2 often act in a complementary opposing manner [37, 38]. Specific gain of function of Akt2 promotes migration and invasion in breast cancer epithelial cells . Moreover, Akt2 induces epithelial to mesenchymal transition [40, 41], a process involved in de novo formation of breast cancer SCs . Thus, an outstanding question is whether Akt1 and Akt2 differentially regulate normal epithelial SCs.
Candidate molecules implicated in SC regulation downstream of Akt are the mechanistic target of rapamycin complex one (mTORC1) signaling complex, positively regulated by Akt signaling , and the FOXO family of transcription factors, which are directly inhibited by Akt-dependent phosphorylation [44-47]. A persistently elevated mTORC1 activity causes an initial expansion followed by a later decline of several adult SC types [48, 49], whereas individual or combined ablation of FOXO genes in mice leads to SC depletion in various lineages [50-52]. In both fly and mammalian cells, FOXOs can restrain mTORC1 activity [53-56], but the role of this regulatory network in SC maintenance has not been defined.
Here we show that in contrast to immortalized epithelial cells, Akt2 restrains LKSC maintenance. Mechanistically, loss of Akt2 disrupts a nuclear signaling network integrating FOXO functions with growth factor signaling, thus favoring cytosolic Akt1 activity that is required for LKSC propagation. We also provide evidence that Akt2−/− mice display an expansion of stem/progenitor cells in stratified epithelia. These findings may have translational implications as they suggest that the development of Akt2-specific inhibitors could improve epithelial SC expansion for the therapy of human corneal disease.
Materials and Methods
Human and Mouse Specimens
Human corneas taken from adult organ donors (after written informed consent was obtained) and considered unsuitable for transplantation were examined with a slit lamp immediately after retrieval and classified as homeostatic corneas, which did not show epithelial defects, dehydration, edema, or inflammation and were taken 3.93 ± 5.15-hour from death. Human corneas were provided by D. Ponzin and A. Ruzza (The Veneto Eye Bank Foundation, Venice, Italy) and F. Genzano (Piedmont Cornea Bank, San Giovanni Battista Hospital, Turin, Italy), and studies on human tissues were approved by the Institutional Ethical Review Committees for research only. Akt2−/− mice were from Jackson Laboratories (Bar Harbor, ME, http://www.jax.org). Mouse tissues were obtained from Akt2−/−  and wild-type mice in C57BL/6 background. Procedures involving care and experimental manipulation of mice were approved by the University of Turin Ethics Committee, Italy.
Cell Culture and Treatments
LKs were isolated from the limbal rings of corneal specimens by trypsinization and cultivated as described in  on a feeder layer of lethally irradiated 3T3-J2 . Briefly, we isolate an average of approximately 1.8 × 104 viable, cells per millimeter square of tissue. Cell viability was assessed by Trypan Blue (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com) dye exclusion.
Cells were plated (1.5 × 104 per centimeter square) and cultured in 5% CO2 and humidified atmosphere in Dulbecco's modified Eagle's medium (DMEM) and Ham's F12 media (2:1) containing FCS (10%), insulin (5 µg/mL), adenine (0.18 mM), hydrocortisone (0.4 µg/mL), cholera toxin (0.1 nM), triiodothyronine (2 nM), glutamine (4 mM) and penicillin-streptomycin (50 IU/mL). Epidermal growth factor (10 ng/mL) was added beginning at the first feeding, 3 days after plating (LKSC media).
Cultures were then fed every other day. For serial propagation, cells were passaged at the stage of subconfluence (harvested, counted, and plated in triplicate samples at each passage for determination of cell recovery), until cultures reached senescence [10, 57]. For colony forming efficiency (CFE) assays, cells (100–2,000) were plated onto 3T3 feeder layers in 100-mm dishes and cultured as above. Colonies were fixed 10–12 days later, stained with Rhodamine-B, and scored under a dissecting microscope. Total colonies were calculated as a percentage of total plated cells (number of colonies × 100/number of cells plated). The aborted colonies were counted as described [9, 10, 57]. Calculation of the number of cell generation and population doublings were performed using the following formula: x = 3.322log N/No., where N is the total number of cells harvested at subculture and No. is the number of clonogenic cells. The number of clonogenic cells, which was determined separately in parallel dishes at the time of cell passage, was calculated from CFE data. Clonal analysis was performed as described by [9, 10]. Briefly, single cells were inoculated onto multiwell plates containing a feeder layer of 3T3 cells. After 7 days of culture, clones were identified with an inverted microscope and transferred to replicate dishes. One dish (1/4 of the clone) was fixed 9–12 days later and stained with Rhodamine-B for clonal type classification, determined by the percentage of terminal colonies formed by the progeny of founding cell. When less than 5% of colonies were terminal, the clone was scored as holoclone. When more than 95% of the colonies were terminal, the clone was classified as paraclone. When more than 5% but less than 95% of colonies were terminal, the clone was classified as meroclone. The second dish was used for further propagation and analysis. In selected experiments, individual colonies were examined by immunofluorescence (IF) analysis with p63α-specific IgG. Morphological classification of colonies was performed as follows: colonies were scored as holoclone-like when they were large, round with regular borders, and formed entirely by small cells with an high nuclear/cytoplasmic ratio, and p63α is uniformly expressed; meroclone-like colonies were medium-large colonies with irregular borders and formed by both small and large cells, and p63α is expressed only at cell borders; paraclone-like colonies were small colonies with irregular borders and formed exclusively by large cells negative for p63α [15, 16]. LK analysis was typically performed on cells between passages 1 and 8, or as indicated. For inhibition of signaling pathways, cells were fed with fully supplemented media containing the following compounds: 20 µM LY294002 (PI3K-Akt), 5 µM Akt inhibitor VIII (Akt1,-2 inhibitor), and 20 nM rapamycin (mTORC1) (all from Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com or Calbiochem, San Diego, CA, http://www.emdbiosciences.com). For long-term incubations with inhibitors, cells were fed every day with medium supplemented with the compounds for the duration of the experiment. LK differentiation was induced by exposing cells to 12-O-tetradecanoylphorbol 13-acetate (TPA; 100ng/ml). For starvation-stimulation experiments, cells were starved by 24-hour incubation with medium without cytokines and serum supplements and then stimulated with fully supplemented medium as indicated.
IF, Immunohistochemistry, Microscopy, and Image Analysis
Cultured LK colonies grown on feeder layer were washed in phosphate-buffered saline (PBS), fixed in 4% paraformaldehyde (PFA) for 15 minutes at room temperature, then permeabilized in 0.3% Triton X-100 in tris-buffered saline (TBS). After washing in PBS, cells were blocked using 0.5% bovine serum albumin (BSA) and 5% goat serum before incubation with primary antibodies overnight at 4°C. Human and mouse corneas were fixed in 4% PFA overnight at 4°C and then incubated in 7.5% and 15% sucrose for 20 minutes followed by 30% sucrose for 2 days at 4°C for cryoprotection before embedding in OCT (Tissue Tek) in molds on dry ice. Cryosections (5–7 µm) were cut, adhered to slides, and dried quickly at room temperature before blocking and immunostaining as described above. Primary antibodies for IF and immunohistochemistry were as follows: 4A4 pan-p63, 5-Bromo-2-deoxyuridine (BrdU) mAb (BD Biosciences, San Diego, CA, http://www.bdbiosciences.com) and a p63α-specific antiserum ; anti-phospho Akt (S473), (Cell Signaling and Technology), Ki67 mAbs (Novocastra, Newcastle upon Tyne, U.K., http://www.novocastra.co.uk), Connexin-43 (Sigma), and Akt2. Tetramethylrhodamine-5-(and 6)-isothiocyanate (TRITC)- and fluorescein isothiocyanate (FITC)-labeled secondary antibodies were from Jackson ImmunoResearch (West Grove, PA, http://www.jacksonimmuno.com). For IF and immunohistochemistry, samples were counterstained with 4′,6′-diamidino-2-phenylindole (DAPI) and mounted in Vectashield aqueous medium (Vector Labs, Burlingame, CA, http://www.vectorlabs.com), then analyzed using a Zeiss LSM 510 Meta and Leica TSCII SP5X confocal microscope. Multitrack analysis was used for image acquisition. Confocal acquisitions were analyzed for quantification of colocalization with ImageJ Software (NIH).
Western Blot, Immunoprecipitation, and Protein Analysis
Mass or clonal cultures of LK were extracted on ice with RIPA buffer containing 50 mM Tris pH 7.4, 1% Triton X-100, 1% Na-deoxycholate, 150 mM NaCl, 0.1% SDS supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM Na3VO4, 10 mM NaF, and 40 mg/mL protease inhibitor cocktail (Sigma-Aldrich). After centrifugation to remove debris, protein concentration of lysates was measured using Protein Assay solution (Bio-Rad, Hercules, CA, http://www.bio-rad. com). Nuclear and cytoplasmic protein extraction was performed using the NE-PER Nuclear and Cytoplasmic Extraction kit (Pierce Chemical Co., Rockford, IL, http://www.piercenet.com) following conditions supplied by the manufacturer. Immunoprecipitations (IPs) with pAkt antibodies were performed as follows: we predetermined that 2 µL of biotinylated mAb against phospho-(Ser 473) Akt precipitated comparable amounts of phosphorylated Akt proteins from 2.5 mg of RIPA extracts from early and late passage cells. After boiling in Laemmli sample buffer, equal amounts of protein (5–15 µg) were elecrophoresed on 7.5% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 5% nonfat dry milk in Tris-buffered saline, 0.1% Tween, and incubated overnight with the indicated antibodies following the manufacturer's instructions. Antibodies for Western blot were as follows: anti-p27Kip and GAPDH mAbs (BD Bioscience), anti-Bmi1 mAb (Upstate Biotechnology), involucrin mAb (Novocastra), anti-C/EBPδ, p16INK4A, Lamin A/C, catalase, tubulin, Hsp90, and actin (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, http://www.scbt.com); phospho-Akt (Ser473), phospho-Akt (Thr308), phospho-p70S6K (Thr389), phospho-Foxo1 (Ser256), phospho-Foxo3 (Ser253), phospho-PRAS40 (Thr246), phospho-TSC2 (Thr1462), phospho-RPS6 (Ser235/236), phospho-4E-binding protein one (p4EBP1) (Thr37/46), cleaved caspase-3, Akt1, Akt2, Akt3, FOXO1, FOXO3, p70S6K, PRAS40, tuberous sclerosis two (TSC2), TSC1, Cyclin D1, and hemagglutinin (HA) epitope tag were from Cell Signaling and Technologies. Sestrin-3 antibody was from Abcam (Cambridge, U.K., http://www.abcam.com). Horseradish peroxidase-conjugated secondary antibodies were from Cell Signaling and Technologies and GE Healthcare. Immunoblots were acquired with the molecular imager ChemiDoc XRS, and densitometric analysis was performed with Quantity One software (Bio-Rad).
Lentiviral, Adenoviral Vectors, and Viral Transduction
Lentiviral vectors (pLKO.1-EGFP) containing short hairpin RNA (shRNA) constructs against Akt1 (5′-CGAGTTTGAGTACCTGAAGCT-3′) and Akt2 (5′-CAAGGTACTTCGATGATGAAT-3′ and 5′-CATGA ATGAGGTGTCTGTCAT-3′, sequence a and b, respectively) were generated [58, 59]. A lentiviral vector pLVTHM containing a second shRNA construct against Akt1 (5′-GGACTACCTGCACTCGGAGAA-3′) was also generated and used . A pLKO.1-EGFP containing shRNA ineffective control sequence was purchased from Sigma-Aldrich. Lentiviral vectors (pLKO.1 puro) containing shRNA constructs against human TSC1 (clone ID TRCN0000039734 and TRCN0000039737 [shRNA a and b, respectively]) were purchased from Open Biosystem. pLVTHM-Akt1-mCherry (sh Akt1-Ch) and pLVTHM-mCherry (sh Ctrl-Ch) vectors were generated by replacing green fluorescent protein (GFP) with mCherry cDNA from pLVTHM-Akt1 and pLVTHM vectors, respectively. Lentiviral constructs for the expression of Akt isoforms were generated as follows: the GFP cDNA sequence was excided from the pCCLsin.PPT.hPGK.GFP vector and replaced with wild-type HA-Akt2 and HA-Akt1 cDNAs from Addgene plasmids #16000 and #9004, respectively, to generate the pCCLsin.Akt2wt and pCCLsin.Akt1wt (LV Akt1) vectors constructs. A control vector carrying a mCherry cDNA sequence was similarly generated (pCCLsin.mCherry; LV Ctrl-Ch) and used along with pCCLsin.PPT.hPGK.GFP (LV Ctrl). A lentiviral construct for the expression of a nuclear localization signal (NLS) Akt1 was obtained by generating a synthetic NLS sequence by primer annealing (forward primer 5′-GGAGA TCTCACCATGGGCCCAAAAAAGAAAAGAAA AGTTCGGC-3′ and reverse primer 5′-CGAACTTTTCTTTTCTTTT TTGGGCCC ATGGTGAGATCTCCGC-3′), inserting the NLS sequence in-frame at the 5′ end of the HA-Akt1wt fragment, and the entire cassette was subsequently cloned into the pCCLsin.PPT. hPGK.pre. vector. A lentiviral vector expressing a nonsilenceable Akt2 wild-type construct (Akt2*) was generated by introducing seven silent point mutations in the Akt2 cDNA coding sequence, converting the 5′-CAAGGTACTTCGATGATGAAT-3′ shRNA target sequence into: 5′-CGAAGATATTTTGACGACGAGT-3′ (pCCLsin.Akt2*, Akt2*). This latter construct was further modified by introducing a single point mutation (K181M) to generate a vector expressing a nonsilenceable Akt2* kinase dead mutant (pCCLsin.Akt2*KD, Akt2*KD). Mutagenesis was performed by using the Quick Change Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, http://www.stratagene.com) following manufacturer's instructions.
For generation of lentiviruses, 293T cells were transfected with a combination of pLKO.1 or pLVTHM or pCCLsin vectors and packaging plasmids pCMV-dR8.74 and pMD2.G using the Effectene reagent (Qiagen, Valencia, CA, http://www1.qiagen.com). Viral supernatants were harvested for more than 36–60 hours, filtrated (0.22-µm pore), and used after concentration by ultracentrifugation (50,000g for 2 hours). Viral titers were determined by transduction of HeLa or TS cells with serial dilutions of the vector stocks and ranged from 107 to 108 TU per milliliter. Transduction efficiency was evaluated (after 72 hours) by scoring GFP+ or HA+ cells by flow cytometry or by scoring vital cells after puromycin selection (24 hours), in the case of PLKO.1 puro vectors.
A replication-defective recombinant adenovirus-expressing simultaneously a GFP protein along with a HA-tagged truncated version of human FOXO3 cDNA wild-type (AdFOXO), devoid of the transactivation domain (D256) (AdFDN) under the control of the CMV promoter were purchased from Vector Biolabs (Philadelphia, PA). A recombinant adenovirus-expressing the GFP (AdGFP) driven by a CMV promoter was used as a control . Adenoviruses were amplified, purified, and used as previously described . Infection efficiency was determined based on percentages of GFP+ normal LKs measured by flow cytometry and/or fluorescence miscroscopy. Infection efficiency was approximately 95% in all experiments.
Subconfluent primary or clonal limbal cultures were trypsinized, and 4 × 104 cells were resuspended in 1 mL of culture medium containing 8 µg/mL polybrene (Sigma) and transduced with lentiviral vectors overnight at 37°C. Gene transfer efficiency was assessed 5 days after transduction by scoring GFP+ cells either by confocal fluorescence miscroscopy analysis or by flow cytometry. Transduction efficiency was approximately 90% in all experiments.
Recombinant Adenovirus Infection
Infection of 30% subconfluent shAkt2 cells, grown on a feeder layer was performed in LK medium without supplements for 1 hour, followed by a further overnight incubation in complete medium.
Cell Cycle Analysis
Triplicate samples of subconfluent LKs kept under basal growth condition (24-hour medium change) were washed off from feeder layer and trypsinized. Cells were fixed and permeabilized in ice-cold 70% ethanol, washed in PBS and incubated 40 minutes at 37°C in propidium iodide (PI) plus RNase solution (BD Bioscience) and processed for flow cytometry on a FACSCalibur (Becton & Dickinson, Franklin Lakes, NJ, http://www.bd.com).
Terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL) assay was performed using the in situ cell death detection kit from Roche (Basel, Switzerland, http://www.roche-applied-science.com), following the manufacturer's instructions. Briefly, cells were fixed with 4% PFA and permeabilized in 0.2% Triton X-100. The TUNEL assay was carried out by incubating the fixed cells with the TUNEL reagent containing TMR red labeled nucleotides at 37°C for 1 hour. The samples were washed in 1× PBS, mounted in mounting media, and fluorescent images were captured using a Leica TSCII SP5X confocal microscope at 10× magnification.
Quantitative Real-Time Polymerase Chain Reaction
RNA was harvested from cells using Triazol reagent (Invitrogen) according to manufacturer's instructions. Cultured cells were gently washed off their 3T3J2 feeders before resuspension in Triazol. RNA was treated with Turbo DNase (Ambion, Austin, TX, http://www.ambion.com) to remove any DNA contaminants and then used for cDNA synthesis using M-MLV ReverseTranscriptase (Promega, Madison, WI, http://www.promega.com) with random primers (Promega). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using Platinum SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen) and run on an ABI 7900HT Fast Real Time PCR System (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). Quantitative PCR was performed on a minimum of two independent biological replicates, and for each cDNA sample PCR reactions were performed in triplicate. Relative mRNA levels of the target gene were normalized for 18S expression, and normalized standard deviations were calculated.
Primer Sequences for RT-PCR
Human sestrin-3: Fwd, 5′-gtacatgctgtggtcctcc-3′; Rev 5′-gcaagatcacaaacgcagaaat-3′. Human Akt1, Akt2, TSC1, TSC2, and 18S primers were purchased from Qiagen (QuantiTect Primer Assay; Hilden, Germany, http://www1.qiagen.com).
Soft Agar Assay
Lentivirus-transduced LKs and the MDA231 breast adenocarcinoma cell line were seeded at a density of 2 × 104 cells per six-multiwell plate and cultured in 1.5% soft agar over a 3% soft agar layer in LK complete media or DMEM plus 10% FBS, respectively at 37°C for 14 days. Colonies raised were analyzed under a microscope and the total numbers of colonies in each dish were counted.
Data obtained from densitometric analysis of immunoblots, growth assay, CFE, and IF analysis were plotted as mean ± SD. Results were assessed for statistical significance by a standard two-tailed Student's t test as indicated. p values are indicated in figure legends.
LK Progenitors Display High Akt Activation Levels Both In Vivo and In Vitro
To identify intracellular signals active in the human limbal epithelial SC niche in vivo, we analyzed the activity of several intracellular pathways in homeostatic corneal tissues. We found that the staining for active Akt (pAkt) nearly overlaps with that of the LKSC marker p63 (Fig. 1A, 1B), which is confined to the basal layer of the limbal epithelium and absent in the cornea [15, 16, 62]. The majority of pAkt+ cells coincide with p63bright cells of the basal layer, but pAkt staining was also detected in few p63dim cells reminiscent of postmitotic cells in transit toward the suprabasal layers (Fig. 1B). Ki67, a marker of actively cycling cells, was detected in few basal limbal cells, as previously reported [63, 64]. We found that pAkt+ cells were either Ki67+ or Ki67− (Supporting Information Fig. S1A), indicating that Akt activation does not reflect the cell proliferative status.
In the attempt to determine whether Akt signaling activity is regulated during LK clonal propagation, we found that attenuation of Akt activity (Fig. 1E) parallels with the drop in clonogenic capacity (Fig. 1C) and proliferative potential (Fig. 1D) that occurs upon LK subcultivation. Since during this process pAkt and ΔNp63α levels decreased, while the level of the involucrin differentiation marker increased, it is likely that the attenuation of Akt activity reflects the commitment of cells toward differentiation. Consistently, analysis of pAkt and involucrin levels revealed that colonies with holoclone morphology have higher Akt activity and low involucrin expression as compared to meroclone colonies (Fig. 1F, upper panels), in spite having similar percentages of Ki67+ cells (Fig. 1F, lower panels), indicating that attenuation of Akt phosphorylation does not merely reflect a decrease in cell proliferation per se. Analysis of clonal cultures derived either from individual holoclones or meroclones also confirmed an increased Akt signaling in uncommitted cell populations (Supporting Information Fig. S1B). These findings suggest that elevated levels of Akt activity are typical of LKSCs and their earliest cellular progenies, both in vivo and in vitro.
To determine if a transient PI3K/Akt inhibition affects LK clonogenic capacity, we treated primary LKs with PI3K inhibitor Ly294002 (Ly) and with an Akt1–2 inhibitor (Akt-i). Upon PI3K/Akt inhibition, we found a mild decrease in cell proliferation without detectable signs of apoptosis (Supporting Information Fig. S1C), coupled to a strong decrease of ΔNp63α protein level and an increase in involucrin expression (Fig. 1G). Notably, the progenies of inhibitor-treated cells showed a approximately 50% drop in clonogenic ability as compared to control cells (Fig. 1H) but displayed similar levels of BrdU incorporation and cell recovery short term (Fig. 1I; Supporting Information Fig. S1D). Thus, Akt activity is prominent in SC-enriched LK populations both in vivo and in vitro, and a transient blockade of PI3K/Akt signaling impairs the LKSC clonogenic potential.
Opposite Roles of Akt1 and Akt2 in LKSC Maintenance
Among the three Akt isoforms, LKs express predominantly Akt1 and Akt2 (Fig. 2A). By comparing Akt isoform expression levels during LK propagation, we found that whereas Akt1 protein remains constant, Akt2 level increases approximately twofold in late passage committed cells (Fig. 2A and data not shown). Consistently, by comparing holoclone- and meroclone-derived cultures, we found an increase of Akt2 to Akt1 ratio in the latter cell populations. (Supporting Information Fig. S1B). Since Akt activity drops and the Akt isoform ratio changes during clonal propagation (Figs. 1D, 2A), we aimed to determine Akt isoforms activation by comparing similar inputs of active Akt proteins in IP experiments with limiting amounts of pAkt antibodies from early and late passage cells. We found that whereas Akt1 activation is prominent in early passages, Akt2 activation is relatively increased in late passages (Fig. 2A, bottom panels, IP), suggesting that Akt1 activity may promote maintenance of LKSCs, and Akt2 exit from the SC compartment. To test this hypothesis, we used a lentiviral-based approach to stably downregulate Akt1 or −2 in LKSC-enriched cultures (Supporting Information Fig. S2A). Primary LKs, either untransduced (−) or infected with viral vectors encoding either a control- or Akt1- and Akt2-specific shRNAs (shCtrl, shAkt1, shAkt2), were serially subcultivated until cells reached replicative senescence. Akt isoforms' downregulation was stable and resulted in approximately 80% protein knockdown with two independent shRNA to Akt1 and Akt2 (Fig. 2B; Supporting Information S2B, S2C, S2E). Analysis of cell proliferation of transduced cells and their early progenies revealed that cultures grew at comparable levels except for shAkt1 LKs, which displayed a modest decrease in cell proliferation, whereas the amounts of apoptotic cells (sub-G0-G1 fraction) was negligible in all cultures (Supporting Information Fig. S2F–S2H). While untransduced and shCtrl LKs displayed a comparable clonal evolution profile (Fig. 2B–2G), shAkt2 LKs maintained long-term an undifferentiated morphology (Fig. 2C) and showed elevated expression of the holoclone markers ΔNp63α, Bmi1, and C/EBPδ [15, 16], (Fig. 2B). ShAkt2 LKs displayed a sustained clonogenic capacity (Fig. 2D; Supporting Information Fig. S2C, S2D) mirrored by a delay in the formation of aborted colonies (Fig. 2F) and generated more than 120 doublings (Fig. 2E) prior to the onset of replicative senescence. Morphological, functional, and biochemical analysis confirmed that these cells were preserved in a state removed from differentiation and senescence (Fig. 2B, 2G), but retained differentiation ability (Supporting Information Fig. S2I), and were not transformed (Supporting Information Fig. S2J). In contrast, shAkt1 cells displayed morphological features suggestive of premature differentiation and/or senescence (Fig. 2C, 2G), and their clonal progression profile was anticipated as they displayed increased tendency to raise aborted colonies (Fig. 2F). Analysis of both clonogenic potential and proliferative capacity confirmed that shAkt1 LKs undergo a premature proliferative rundown (Fig. 2D–2F). Thus, Akt1 and Akt2 play distinct roles in LKs, as Akt1 de novo loss hampers LK self-renewal by anticipating the onset of replicative senescence, whereas Akt2 downregulation favors the maintenance of cells displaying the expected properties of cultured LKSCs.
Decreased Akt2 Signaling Activity Enhances LKSC Maintenance in an Akt1-Dependent Manner
In the attempt to determine whether the LKSC phenotype induced by Akt2 downregulation could be rescued by the expression of a nonsilenceable Akt2 mutant (Akt2*), to assess for both kinase dependence and isoform specificity, we also generated lentiviral constructs expressing a kinase-dead Akt2 (Akt2*KD), Akt1 wild-type (Akt1), or an unrelated mCherry protein (Ctrl-Ch). Analysis of transduced cells revealed that the undifferentiated morphology and sustained clonogenicity of shAkt2 LKs was specifically reverted upon expression of Akt2*, but not Akt2*KD, wild-type Akt1 or Ctrl-Ch (Fig. 3A, 3B). In particular, Akt2* expression induced morphological changes of premature stratification, coupled to a approximately 50% reduction in clonogenicity with a nearly complete ablation of holoclone-like colonies (Fig. 3C). Consistently, Akt2* expression was paralleled by a decrease of ΔNp63α and upregulation of involucrin (Fig. 3D). To determine the role of Akt1 in the clonogenic potential of shAkt2 LK, we subjected shAkt2 LKs to Akt1 knockdown and found a approximately 50% decrease in cell clonogenic ability and loss of holoclone-like colonies (Fig. 3E–3G). Collectively, these data indicate that the SC phenotype of shAkt2 cells depends selectively on the suppression of Akt2 signaling activity and requires a functional Akt1 protein.
Akt2-Deficient Cells Display Increased FOXO Functions Coupled with Attenuation of mTORC1 Signaling and Sustained Akt Activity
To determine the molecular bases of the distinct roles of Akt isoforms in LKSC regulation, we first analyzed the kinetics of Akt activation during starvation stimulation experiments in both Akt1 and Akt2 knockdown cells (Fig. 4A). Here, we did not detect either a decrease in the amplitude of total Akt activity or reduced phosphorylation of cytosolic Akt substrates such as glycogen synthase kinase 3α and -β (data not shown), suggesting that the effects of individual Akt isoforms knockdown depend on the impairment of specific downstream events rather than on major drops in Akt signaling outputs. Surprisingly, however, shAkt2 cells, showed a more sustained Akt activity at late times of stimulation (48 hours), suggesting the involvement of Akt2 in a negative feedback loop that limits the duration of Akt activity upon growth factor stimulation.
Among direct Akt substrates, FOXO transcription factors have been implicated in SC maintenance in several cell lineages [51, 65], but their role in stratified epithelial SCs is still unknown. In normal LKs, we found that the expression of FOXO1 and FOXO3 parallels with that of the LKSC marker ΔNp63α and decreases with the onset of replicative senescence (Fig. 4B and data not shown). Therefore, we hypothesized that the effects of Akt2 deficiency on LKSC maintenance may be due, at least in part, to a defective ability of these cells to antagonize FOXO functions. Indeed, analysis of FOXO1 expression revealed higher protein level in shAkt2 LKs compared to Ctrl- and shAkt1 LKs, and this paralleled with decreased phosphorylation at the Ser 256 Akt target site (Fig. 4A). Increased FOXO1 and-3 levels and decreased phosphorylation of respective Akt substrate sites were also observed under steady-state conditions (Fig. 4C). Consistently, shAkt2 cells displayed higher levels of protein products of FOXO target genes such as p27 and catalase  (Fig. 4D) that paralleled with elevated expression of nuclear and cytosolic FOXOs (Fig. 4E) and a triad of LKSC markers such as ΔNp63α, Bmi1, and C/EBPδ (Supporting Information Fig. S3A) . Consistent with a role of Akt1 in promoting LKSC maintenance, shAkt1 cells displayed instead a decrease in FOXO1 and FOXO3 proteins (Fig. 4A, 4C, 4E) that was reminiscent of the reduced FOXO levels of late passage LKs (see Fig. 4B), and which may reflect a premature onset of cell senescence, in agreement with the data of Figure 2C, 2G. Altogether, these data indicate that Akt2, but not Akt1, limits FOXO activity in cultured LKSCs.
One crucial function of FOXO proteins conserved between flies and mammals is attenuation of mTORC1 signaling . Having found a higher mTORC1 activity in normal SC-depleted LK cultures as assessed by phosphorylation of mTOR downstream targets S6 ribosomal protein and p4EBP1 (Fig. 4B), we hypothesized that in shAkt2 cells, the gain in FOXO activity may be coupled to attenuation of mTORC1 signaling. In fact, we found that these cells display a reduced S6K phosphorylation in response to growth factors (Fig. 4F). Since mTORC1 inhibition can potentiate Akt activity for the loss of a negative feedback to growth factor receptor signaling , mTORC1 attenuation may contribute to the sustained Akt activation of shAkt2 cells. This possibility is supported by the observation that mTORC1 inhibition by rapamycin also enhances Akt activity in LKs (Supporting Information Fig. S3B). Interestingly, Akt2 deficiency did not parallel with defective phosphorylation of cytosolic Akt substrates involved in mTORC1 regulation such as TSC2 and PRAS40 (Fig. 4F), suggesting the existence of an alternative mechanism of mTORC1 regulation downstream of Akt2. Thus, Akt2 knockdown promotes signaling features such as gain of FOXO functions, reduced mTORC1 signaling and prolonged Akt activity that are also found in normal LKSCs populations.
The SC Phenotype of Akt2-Deficient Cells Depends on Endogenous FOXO Transcriptional Activity
To determine the effects of suppression of FOXO transcriptional activity on the shAkt2 LKSC phenotype, we expressed in these cells a FOXO mutant protein deleted in its transactivation domain (FDN) via an adenoviral vector [67, 68] (Supporting Information Fig. S4A). Upon FDN expression, shAkt2 LKs showed reduced levels of p27 and catalase and a approximately 50% drop in their clonogenic ability, which was virtually set to zero within the third passage (Fig. 5A, 5B), even if the expression of the HA-tagged transgene was no longer detected after the first passage, indicating that even a transient inhibition of FOXO transcriptional activity is sufficient to induce long-term phenotypic changes in these cells. Accordingly, we detected reduced expression of limbal SC markers such as ΔNp63α and C/EBPδ that were paralleled by morphological signs of cell death (Supporting Information Fig. S4B). Apoptosis was revealed by Western and IF analysis for cleaved caspase-3 (Fig. 5A, 5C) and TUNEL assay (Supporting Information Fig. S4C) indicating that inhibition of FOXO transcriptional activity hampers the viability of clonogenic shAkt2 cells. Notably, we found that shAkt2 cells expressing the FDN mutant protein also show increased pS6K levels (Fig. 5A), indicating that their reduced mTORC1 signaling depend, at least in part, on endogenous FOXO activity.
To determine the mechanisms underlying FOXO-dependent mTORC1 inhibition in the context of Akt2 knockdown, we analyzed the expression of FOXO target genes potentially involved in mTORC1 inhibition, such as sestrin-3 and TSC1 [54, 69]. We found that both genes were upregulated in shAkt2 cells (Fig. 5D, 5E); interestingly, also TSC2 mRNA and protein levels were induced in these cells in a FOXO-dependent manner (Fig. 5F). Upon TSC1 knockdown, shAkt2 cells displayed increased S6K phosphorylation (Fig. 5G), acquired a flat, differentiated morphology (Fig. 6H, top panels), lost their clonogenic capacity (Fig. 5I, bottom panels), and upregulated involucrin expression (Fig. 5I). Collectively, these data indicate that the FOXO-TSC1 axis in shAkt2 LKs is key for attenuating mTORC1 signaling and to preserve SC maintenance.
Differential Nuclear Localization of Akt Isoforms Dictates Their Selective Activity Toward FOXO
We then aimed to determine the basis of Akt2 versus Akt1 specificity in FOXO regulation. Since both isoforms are highly homologous in their kinase domains, we reasoned that the greater ability of Akt2 versus Akt1 to downregulate FOXOs might rely on a differential subcellular distribution. We found that in normal LKs both Akt isoforms are present in the cytosol, but Akt2 displays a preferential nuclear localization both under basal growth conditions and at various times after starvation-stimulation (Fig. 6A and data not shown). Moreover, fractionation analysis of shAkt2 LKs expressing exogenous Akt1 and Akt2 proteins (Fig. 6B), in which the SC phenotype is selectively reverted by Akt2 expression (Fig. 4A–4E), indicated that in the presence of comparable levels of ectopic Akt proteins in the cytosol, Akt2 is approximately 10-fold more abundant than Akt1 in the nucleus, and this parallels with an increased phosphorylation of nuclear FOXO1 (Fig. 6B). The preferential nuclear distribution of ectopically expressed Akt2 protein was also confirmed by IF analysis (Fig. 6C). Notably, the progenies of shAkt2 cells expressing a wild-type Akt1 protein fused with a nuclear localization signal peptide (NLSAkt1), compared to those expressing either a control vector or Akt1 wild-type, selectively lost their undifferentiated morphology and clonogenicity and upregulated involucrin expression (Fig. 6D–6F). Analysis of the subcellular distribution of the mutant protein revealed that a significant fraction of NLSAkt1 was localized to the nucleus, with a parallel increase in the levels of both nuclear pAkt and phospho-FOXO1 (Fig. 6G). Thus, expression of a nuclear Akt1 protein effectively rescues the shAkt2 SC phenotype, and this function parallels with its ability to phosphorylate FOXO within the nuclear compartment.
Akt2 Deficiency Parallels with an Expansion of Putative Corneal Stem/Progenitor Cells In Vivo
To verify if Akt2 deficiency favors the expansion of LK stem/progenitor cells in vivo, we first analyzed the limbal-corneal epithelium of Akt2-deficient mice. Differently from the human tissue, both the limbal and the corneal murine epithelia have been recently shown to contain SC populations displaying elevated p63 levels  . Compared to wild-type control mice, Akt2−/− mice display a higher fraction of p63bright cells in both limbal and corneal epithelia coupled with a reduced expression of the corneal differentiation marker connexin 43 in the central cornea (Fig. 7A). These data indicate that Akt2 deficiency is coupled with an expansion of putative stem/progenitor cells in stratified epithelia in vivo.
Although the PI3K/Akt signaling axis has emerged as an important player in SC biology, how this signaling module regulate stratified epithelial SCs is still rather unexplored, especially in respect to the functions of individual Akt isoforms. By using human LKs as a model system, we reveal here that Akt activity is prominent in keratinocyte populations enriched for SCs and declines in differentiated cell populations both in vivo and in vitro. Moreover, a transient inhibition of the global PI3K/Akt signaling impairs LK clonogenic ability and lowers the level of the LKSC marker p63 indicating important functions in the balance between SC self-renewal and differentiation (Fig. 1). Cultured LKs express Akt1 and Akt2, with Akt1 being predominantly phosphorylated in early passages, switching to Akt2 in later passages. Since we found an increased ratio of Akt2 versus Akt1 protein expression in cell differentiation, this may account in part for the observation that in more differentiated or senescent cell populations, the activity of Akt2 prevails over that of Akt1, in spite of a drop of global Akt activity (Fig. 2A). Moreover, these data are not in contrast with the possibility that a differentiation-specific mechanism of Akt2 activation may contribute to increase the balance between Akt2 and Akt1 signaling, challenging the opinion that all Akt isoforms share common stimulatory mechanisms as suggested in neutrofils . These changes in the balance between Akt1 and Akt2 signaling have likely important roles in LKSC maintenance, as Akt1 knockdown leads to a premature exhaustion of SC proliferative potential, whereas Akt2 downregulation prolongs LK expansion ex vivo by counteracting differentiation or senescence programs, fostering cell populations with morphological, functional and molecular features typical of bona fide SCs (Fig. 2).
Our study suggests that Akt2 acts at the decision of LKSCs to self-renew or exit the SC compartment rather than being an essential component of the cell differentiation machinery, as Akt2 knockdown cells retain differentiation ability (Supporting Information Fig. S2I). In support of an in vivo role for Akt2 deficiency in preserving SC maintenance, we reveal that knockout of Akt2 leads to an expansion of putative stem/progenitor cell pools in the murine cornea.
Notably, the different effects of Akt1 or Akt2 deficiency on LKSC fate decisions were found mostly independent of short-term effects on cell survival and proliferation, and were not associated with drops in overall Akt signaling outputs, findings that are consistent with the notion that Akt isoforms also share partly redundant functions .
Mechanistically, we find that both Akt1 and Akt2 are present in the cytosol of LKs, but Akt2 is the predominant Akt isoform in the nucleus phosphorylating FOXOs in this cell compartment (Fig. 6). The preferential nuclear localization of Akt2 is key for its ability to limit epithelial SC self-renewal since knockdown of Akt2, but not Akt1, enhances FOXO functions, attenuates mTORC1 activity, and promotes growth factor signaling to Akt1, thereby preventing LK commitment toward differentiation (see schematic diagram in Fig. 6H). The decreased phosphorylation of FOXO1 and FOXO3 at Akt target sites induced by Akt2 knockdown suggests that this is likely one primary determinant of the enhanced FOXO transcriptional activity , although we cannot rule out additional mechanisms, due to the complexity of the regulatory inputs on these molecules [66, 73]. We found that one critical event in preventing LK differentiation following Akt2 knockdown is the induction of the TSC1 FOXO target gene , which contributes to mTORC1 attenuation (Fig. 5). Thus, our study reveals that via Akt2, the FOXO/mTORC1 axis is involved in epithelial SC regulation. It would be interesting to investigate whether a similar hierarchy of signals takes place in the hematopoietic system, in which either loss of FOXO functions  or increased mTORC1 activity induced by TSC1 mutations  impair SCs maintenance.
Downstream of Akt2, the roles of FOXOs in LKSC control likely extend beyond mTORC1 signaling inhibition, since we found that although prolonged rapamycin exposure of LKs causes a transient increase in clonogenic ability, it fails to increase FOXO levels and enhance cell lifespan (data not shown), recapitulating only some aspects of the shAkt2 cell phenotype. Moreover, interference with FOXO activity has severe consequences on the survival of clonogenic LKs, whereas the sustained mTORC1 activity induced by TSC1 loss triggers cell differentiation rather than death. It is likely that FOXOs preserves viability of clonogenic cells via direct target genes implicated in stress protection such as catalase and sestrin-3 [66, 75], which are also upregulated upon Akt2 knockdown. Additionally, as FOXOs cross-talk with keratinocyte fate-regulatory genes [66, 76], our data suggest that they may participate to the transcriptional regulatory network of LKSCs, either directly or indirectly.
Interestingly, it was recently reported that prolonged rapamycin treatment enhances cultured oral keratinocyte lifespan protecting SCs from replicative senescence . These effects of rapamycin, however, were found uncoupled from Akt activation and/or disruption of mTORC1 negative feedback to growth factor signaling , which occur instead in our experimental model. Whether this discrepancy reflects cell-type-specific differences or different culture conditions deserve further investigation.
The differential localization of Akt isoforms is of great functional significance, since the effects of Akt2 knockdown on LKSC self-renewal cannot be compensated by expression of Akt1, unless its expression is enforced into the nucleus (Fig. 6). Thus, we propose a model in which unbalances in the nuclear versus cytosolic Akt signaling determine opposite effects on LKSC maintenance, whereby a predominant cytosolic signaling would promote SC maintenance, while an increase in the nuclear activity would promote SC exhaustion. The mechanisms dictating Akt isoforms nuclear localization are currently unknown. Neither Akt1 nor Akt2 bear canonical nuclear localization signals; conversely, all Akt isoforms possess one conserved putative nuclear export signal, the functions of which are likely cell-type-specific and/or context dependent [78, 79]. Since Akt proteins do not diffuse passively across the nuclear envelope , it seems likely that Akt2 nuclear localization may require interaction with specific cargo proteins, the identity of which needs to be determined.
Our findings have potential translational implications since the effects of Akt2 knockdown are rescued by expression of a kinase active but not a kinase dead Akt2 protein, suggesting that the development of Akt2-selective inhibitors may facilitate epithelial SC expansion from patients in which the SC pools are compromised by traumas, aging or disease. This possibility is further supported by our observation that attenuation of Akt2 signaling causes ΔNp63α upregulation, a factor that correlates with the success of cultured LKSC grafts in therapeutic settings [2, 14].
Importantly, our data indicating that Akt2 opposes normal epithelial SC functions are apparently at odds with the notion that in a human immortalized breast cell line, Akt2 favors de novo formation of cancer SCs [40-42] . Therefore, our work indicates that caution should be taken when interpreting the biological function of Akt signaling in normal epithelial cells or in immortalized and/or transformed cell lines that already evaded fail-safe mechanisms of tumor suppression. Collectively, our findings challenge the current view of how distinct Akt isoforms regulate normal versus cancer SCs, suggesting more complex roles for these molecules in physiological and pathological conditions.
This study reveals for the first time that individual Akt isoforms play opposite roles in normal epithelial SCs expansion, and in particular, that Akt2 nuclear signaling opposes LKSC maintenance by selective regulation of a FOXO-mTORC1 signaling pathway. Based on these findings, our work suggests that pharmacological strategies aimed to attenuate Akt2 signaling may facilitate the expansion of LKSCs from patients in which the endogenous SC pools are compromised by traumas, aging or disease.
We thank Michele De Luca and Graziella Pellegrini for their helpful advice and support on the human limbal epithelial model system; Diego Ponzin and the Veneto Eye Bank Foundation for providing this study with the human specimens for the in vivo studies, and Federico Genzano for providing additional corneal specimens; Paolo Dotto, Stefano Gustincich, Francesca Persichetti, Sandro Goruppi and Caterina Missero for critical reading of the manuscript; Annalisa Camporeale for help with FACS analysis; Paolo Provero for advice in statistical analysis. This manuscript is dedicated to the loving memory of Maria Luisa Vinay Calautti. The financial support of Telethon, Italy (TCP 06001) (E.C.), the Piemonte Region (PISTEM and Converging Technologies Grants) (E.C.), the Italian Association for Cancer Research (AIRC), IG-11346 (S.C.) and IG-8675 (R.P.). are gratefully acknowledged. F.A. is currently affiliated with Inserm U1091, IBV, Diabetes Genetics Team, Université de Nice-Sophia Antipolis, Nice, France; G.T. is currently affiliated with Cardiff School of Biosciences, Cardiff, Wales, U.K.; E.D.L. is currently affiliated with Center for Bio-Molecular Nanotechnology, Italian Institute of Technology, Arnesano (LE), Italy.
S.S.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; B.T., E.D., C.J., F.A., and G.T.: collection and assembly of data and data analysis and interpretation; E.D.L.: data analysis; R.P.: provision of study material; S.C., E.T., and P.P.P.: data analysis and interpretation; E.C.: conception and design, assembly of data, data analysis and interpretation, manuscript writing and final approval of manuscript.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflict of interests.