Keratins regulate protein biosynthesis through localization of GLUT1 and -3 upstream of AMP kinase and Raptor

Embryonic development is a fine-tuned interplay of rapid cell growth and differentiation. It is governed by signaling processes that are coordinated in a spatiotemporal manner through interactions with cytoskeletal and scaffold proteins such as keratins in epithelia. However, the function of keratins in spatiotemporal scaffolding and signaling control is unclear. K7, -8, -18, and -19 represent the first keratins during mouse development and begin to form a primary cytoskeleton at nascent desmosomes in the trophectoderm (Jackson et al., 1980). From then on, these keratins are present in all embryonic and extraembryonic epithelia. Owing to their redundancy, it has not been possible to assign and discriminate their mechanical and signaling functions during embryo development and in tissue homeostasis (Hesse et al., 2000; Tamai et al., 2000; Jaquemar et al., 2003). The former is highlighted by previous gene knockout (KO) studies, which have arrived at contradictory results (Baribault et al., 1993; Magin et al., 1998; Hesse et al., 2000; Tamai et al., 2000; Jaquemar et al., 2003). Deletion of K8 caused an embryonic lethal phenotype at embryonic day (E) 12.5, which is associated with placental malfunctions caused by maternal TNF-induced apoptosis (Baribault et al., 1993; Jaquemar et al., 2003). Deletion of K18 permitted normal development because of the presence of K19, illustrating functional redundancy, at least for these two keratins (Magin et al., 1998). The combined deletion of K18/K19 and of K8/K19, which eliminated redundancy, caused fragility of giant trophoblast cells followed by extensive hemorrhages, which led to death at ∼E10 (Hesse et al., 2000; Tamai et al., 2000). This was interpreted to indicate a primary mechanical function of keratins, which is analogous to that seen in skin epidermis (Fuchs and Cleveland, 1998; Hesse et al., 2000; Tamai et al., 2000; Kim and Coulombe, 2007; Magin et al., 2007).

To systematically analyze keratin functions during embryo development, we exploited the genomic organization of keratin genes. The mouse type I and II keratin families are clustered on two contigs, which are located on chromosomes 11 and 15, respectively (Hesse et al., 2001, 2004; Schweizer et al., 2006). In this study, we describe mice lacking the type II gene cluster. Given that the assembly of keratin filaments from heterodimers requires one member from each family and that keratins are rapidly degraded in the absence of a dimerization partner, mice lacking the type II gene cluster should be devoid of the entire keratin multiprotein family.

To test current hypotheses on keratin function in mouse development, we used the Cre-loxP system (Ramírez-Solis et al., 1995) to flox the type II keratin gene cluster spanning 0.68 Mb of the genome in mouse embryonic stem (ES) cells (Fig. 1 A; Hesse et al., 2004). Targeting constructs from the Mutagenic Insertion and Chromosome Engineering Resource (MICER; Adams et al., 2004) were engineered with gaps to aid in insertional targeting (Fig. 1, B and C; and Fig. S1 A). Southern blotting confirmed correct targeting at a frequency of 8% (Fig. 1, F and G; and Fig. S1 A). Empty 3′ and 5′ hprt vectors labeled for in situ hybridization against spread chromosomes from double-targeted ES cell clones identified double-targeted clones in cis (Fig. S1 B). The floxed gene cluster contained all type II keratins and the type I keratin Krt18, which with K8 forms the first keratin pair during embryonic development (Fig. 1 A; Lu et al., 2005), but no other known genes, including microRNA genes.

Cre-mediated deletion of the keratin type II cluster (Fig. 1, D and E) did not affect ES cell pluripotency, and mice with a constitutive deletion of the keratin type II cluster (KtyII^−/−) were generated. Deletion of all 27 keratin genes was confirmed by RT-PCR and immunofluorescence microscopy (Fig. 1, H and I; and Fig. 2). Consistent with the proteolytic sensitivity of type I keratins in the absence of their type II keratin binding partners, the sole embryonic type I keratin K19 was expressed at the mRNA but not at the protein level (Fig. 1 H and Fig. S1 D; Magin et al., 2007). Therefore, our mice lack all 54 mammalian keratins (Fig. 1 H). The type III intermediate filament protein vimentin, which is frequently up-regulated during epithelial-mesenchymal transition after loss of keratin expression (Thiery, 2002; Yang and Weinberg, 2008), was not up-regulated at the transcript or protein level, indicating that deletion of the KtyII^−/− cluster did not grossly perturb epithelial cell morphology or function (Fig. S2). In support, the expression of the constitutive chaperone Hsc70, which can bind keratins (Liao et al., 1997; Betz et al., 2006), was unaltered. The stress-inducible Hsp70 was not detectable (see Fig. 4 A). Furthermore, activity of MAPK, as examined by Western blotting of candidate target proteins, appeared largely unchanged (see Fig. 4 C).

All KtyII^−/− mice died at ∼E9.5 (Fig. S1 E). Because keratins maintain tissue integrity by interacting with desmosomes to provide intercellular adhesion, we investigated the gross appearance and histology of E9.5 embryos. Unlike previous single or double keratin KO mice, which suffered from cytolysis and hemorrhages (Baribault et al., 1993; Hesse et al., 2000; Tamai et al., 2000), KtyII^−/− embryos had intact embryonic and extraembryonic epithelia (Fig. 3, A–F). These findings suggest that keratins have no essential mechanical function until this stage of mouse development and that the phenotype of previous keratin KOs may result from dominant-negative effects. Yet, the desmosomes in KtyII^−/− embryos were smaller and partially mislocalized (Figs. 1 I and 2, A and B), which is consistent with the involvement of keratins in desmosome assembly (Godsel et al., 2005). During epidermal differentiation, desmoplakin was reported to regulate microtubule organization through ninein (Lechler and Fuchs, 2005). Staining for ninein revealed a prominent localization along the plasma membrane of yolk sac tissue (not depicted) and at centrosomes in embryonic epithelia (Fig. 2 C). Ninein-positive centrosomes retained their apical position in KtyII^−/− embryos (Fig. 2 C′), which is in agreement with unaltered γ-tubulin staining (not depicted). This is in contrast to a previous study (Ameen et al., 2001), indicating other and possibly compensatory mechanisms involved in centrosome positioning. The localization of the adherens junction protein epithelial cadherin (E-cadherin) and the tight junction–associated proteins ZO-1 and occludin were highly similar in both genotypes of embryos, indicating that actin-associated adhesion complexes and the actin cytoskeleton (not depicted) maintain epithelial integrity and polarity (Fig. 2, D–F).

KtyII^−/− embryos exhibited a striking growth retardation, which started at ∼E8.5 and was fully apparent 1 d later (Fig. 3, G–J). Before the onset of placenta formation at E9.5, the embryo is fully dependent on the yolk sac for nutrient supply. The mammalian target of rapamycin (mTOR) C1 complex regulates protein synthesis by integrating growth factor signals and nutrients. Stimulation of the mTORC1 complex up-regulates protein synthesis by phosphorylation of its downstream targets ribosomal protein S6 kinase (S6K) and eukaryotic initiation factor 4E-binding protein 1 (4E-BP1; Wullschleger et al., 2006). We hypothesized that the embryonic mortality was caused by defective energy metabolism in the yolk sac.

In a metabolic labeling experiment, analysis of ³⁵S-labeled Met/Cys incorporation showed that protein biosynthesis was reduced by ∼48% in the yolk sac and by ∼45% in the embryonic tissue of KtyII^−/− embryos (Fig. 3 K). Moreover, phosphorylation of the mTORC1 targets S6K and 4E-BP1 was reduced (Fig. 4 D), and eIF2-α phosphorylation was increased (Fig. 4 B). mTORC1 activity is regulated by several mechanisms, among them sequestration through 14-3-3 proteins. Previously, the skin keratin K17 was found to positively regulate protein biosynthesis and keratinocyte growth through 14-3-3–σ-mediated mTOR sequestration, suggesting a distinct role of certain keratins in wound repair (Kim et al., 2006). Although we detected sufficient 14-3-3 protein by Western blotting, the small size of embryos prevented a more detailed analysis. Given that limited glucose supply is known to severely restrict embryo growth and to increase apoptosis (Schmidt et al., 2009) and that mTORC1 is nutrient sensitive (Shaw and Cantley, 2006), we were prompted to investigate upstream regulators of mTORC1 that might depend on keratins.

Limited nutrition represses growth and protein biosynthesis, and early mouse embryos predominantly rely on glycolysis (Pantaleon and Kaye, 1998). Therefore, we analyzed the AMP kinase (AMPK), the cellular energy sensor which is phosphorylated when AMP levels are elevated (Hardie, 2007). Phosphorylated (P) AMPK inhibits the mTORC1 complex through its binding partner Raptor (Gwinn et al., 2008). Using phosphospecific antibodies, we found increases of 20% in P-AMPK and 30% in P-Raptor (Fig. 4, E and F). Furthermore, we detected no change in phosphorylation of Akt, which is a positive regulator of mTOR signaling (Fig. 4 E). These findings confirmed our hypothesis that nutrient shortage, which can signal through AMPK, reduced protein synthesis in KtyII^−/− embryos.

Next, we performed a biochemical rescue experiment in which glucose was added to isolated KtyII^−/− and wild-type (WT) embryos in ex vivo culture; the metabolic inhibitor deoxyglucose served as a negative control. In WT embryos, P-AMPK levels were decreased by 10% in 5 mM glucose and increased by 52% in deoxyglucose medium (Fig. 4 G). However, KtyII^−/− embryos failed to respond to either treatment, indicating an unsuccessful rescue. Their P-AMPK levels were similar to those of deoxyglucose-treated WT embryos, and their total AMPK levels were unchanged (Fig. 4 G). These findings strongly suggest that keratins participate in the regulation of cellular glucose uptake.

Glucose transport is mainly performed by energy-independent facilitative transporters of the GLUT family, which are expressed during blastocyst formation when embryos switch from pyruvate to glucose as the major energy source (Barbehenn et al., 1974). In early embryos, GLUT1 and -3 are the main transporters that regulate glucose distribution, and their KO causes severe growth retardation with increased apoptosis during embryo development (Wang et al., 2006; Ganguly et al., 2007; Schmidt et al., 2009). FACS sorting of trypsinized KtyII^−/− embryos confirmed a 10-fold increase in apoptotic cells, indicating a nutrition defect (Fig. 3 L). To analyze this further, double immunofluorescence analysis with GLUT1 and -3 and K8/K18 antibodies were performed. This showed that GLUT1 and -3 were predominantly localized to the apical plasma membrane of the yolk sac. In WT yolk sacs, K8/K18 keratins were confined in the subapical region (Figs. 1 I and 5, A and B). Of note, subapical intermediate filament organization is evolutionarily conserved down to Caenorhabditis elegans, where it is required for gut epithelial organization (Hüsken et al., 2008). In KtyII^−/− embryos, GLUT1 and -3 levels at the apical plasma membrane of the yolk sac were markedly reduced, causing redistribution of the transporters throughout endodermal cells (Fig. 5, A′, A″, B′, and B″). In line with AMPK activation, their total transcript and protein levels were unaltered (GLUT1) or slightly elevated (GLUT3; Fig. 5, C and D). Our data lead to a model in which full activation of mTORC1 depends on the correct localization of GLUT1 and -3 by subapical keratins in the mouse embryo (Fig. 5 E).

These findings reveal a novel regulatory mechanism by which keratins coordinate cell growth and protein synthesis at the level of GLUT transporters. The highly regulated expression and subcellular organization of keratins strongly suggest their involvement in growth regulation and protein targeting beyond embryo development, as demonstrated for K17 in skin wound healing and K8 in colonic epithelia (Toivola et al., 2004; Kim et al., 2006). Although the molecular mechanisms are not yet known, it is well established that the correct localization of cell adhesion proteins depends on keratins and vimentin (Godsel et al., 2005; Toivola et al., 2005; Nieminen et al., 2006). Possibly, keratins and vimentin orchestrate the local interaction of 14-3-3 proteins with their multiple binding partners during organelle transport, cell polarity, and signaling. Furthermore, our data have far-ranging implications for the analysis of other large mammalian gene families as they suggest that some of the previous single and double keratin KOs may represent gain-of-toxic-function phenotypes (Hesse et al., 2000; Tamai et al., 2000; Jaquemar et al., 2003; Magin et al., 2004). This implicates that the pathomechanisms underlying skin keratin disorders not only result from mechanical fragility but from disturbed regulation of cell growth and signaling, opening new therapeutic opportunities (Kim et al., 2006; Kerns et al., 2007; Roth et al., 2009).

The 5′ hprt vector clone MHPN117k13 (Adams et al., 2004) was used to target the 5′ end of the type II keratin cluster. This contained an insert of 6.0 kb spanning 101,202,450–101,208,417 bp on chromosome 15. A gap of 1.5 kb was generated with unique restriction sites, NheI and Bsp119I, to yield 2.0- and 2.4-kb arms of homology (Figs. 1 A and S1 A). NotI linkers were introduced at the cut sites to obtain a NotI restriction site. The plasmid was linearized with NotI before targeting and transfected (200 µg) into AB2.2 cells (gift from A. Bradley, Wellcome Trust Sanger Institute, Cambridge, England, UK) at 3 µF and 800 V. G418 selection was initiated 24 h after targeting at 350 µg/ml. Neomycin-resistant colonies were screened for homologous recombination; the plasmid without the gap served as a positive control. PCR primers spanned the gap region and vector backbone (Table S1). Eight clones that were PCR positive for the homologous recombination event were further confirmed by Southern blotting with a probe specific to the gap region (Table S1). Five were positive for the homologous recombination at the 5′ end of the keratin type II cluster. Clone 2 was used to target the 3′ end of the keratin type II cluster.

The insert from the MICER 5′ hprt clone MHPN322c09 (Adams et al., 2004) was excised from the vector at the AscI sites flanking the insert, cloned into an empty 3′ hprt vector, and named MHPP322c09. This contained a 6.9-kb insert spanning 101,876,016–101,882,964 bp on chromosome 15. A gap of 1.4 kb was generated with SacI restriction sites to yield 3.0- and 2.4-kb arms of homology (Fig. 1 B and Fig. S1 A). SacI was used to linearize the plasmid to target into the aforementioned clone 2. ES cells were targeted as described in the previous section and selected in 3 µg/ml puromycin 24 h after electroporation. 8 of 96 picked clones showed homologous recombination, as determined by PCR with primers spanning the gap region and vector backbone (Table S1). These clones were confirmed by Southern blot analysis with a 487-bp probe designed within the gap region (Table S1). All PCR-positive clones were correctly recombined at the 3′ end of the keratin type II cluster, as confirmed by Southern blot analysis.

Metaphase chromosome spreads on slides were performed as previously described (Henegariu et al., 2001). Empty 3′ and 5′ hprt vectors were labeled with biotin and digoxigenin, respectively, by nick translation and used for chromosomal in situ hybridization against spread chromosomes from double-targeted ES cell clones according to a standard protocol (Wrehlke et al., 1999). Red and green signals (or a yellow overlap) on a single chromosome confirmed the double targeting in cis (Fig. S1 B). One of the three double-targeted clones tested by fluorescence in situ hybridization was confirmed to have the targeted constructs in cis. This clone was subjected to Cre expression.

The double-targeted clone in cis was transiently transfected with 200 µg of CrePac vector (Taniguchi et al., 1998) using the identical conditions described in Targeting the 5′ end of the keratin type II cluster. Selection with 1× hypoxanthine and thymine was initiated 24 h after transfection for 10 d. 96 colonies were screened for deletion of the keratin cluster with primers specific to the 3′ and 5′ hprt regions. This was further confirmed by Southern blot analysis with probes specific to the 3′ and 5′ hprt regions. Two independent clones positive for the deleted cluster were used to generate male chimeras by blastocyst injections (gift from R. Maniu, Universität Bonn, Bonn, Germany). Male chimeras were outbred to C57BL/6 WT females to generate mice heterozygous for the keratin type II deletion. Heterozygous progeny were inbred to generate KtyII^−/− mice.

Total RNA was isolated from E9.5 embryos and yolk sacs with RNeasy Micro kits (QIAGEN). Reverse transcription was performed with RevertAid First Strand cDNA Synthesis kits (Fermentas). The primers and PCR conditions have been described previously (Tonack et al., 2004; Lu et al., 2005). For vimentin cDNA synthesis, 2 µg of total RNA was reverse transcribed in a volume of 20 µl. PCR with Platinum Taq polymerase (Invitrogen) was performed in 25-µl reactions containing 0.1 µl of template cDNA, according to the manufacturer's protocol. PCR reactions were performed as follows: 35 cycles at 94°C for 30 s, 65°C for 30 s, and 72°C for 20 s. The sequences of the primer pairs are outlined in Table S1.

Western blotting was performed as follows. Total protein was extracted in SDS-PAGE sample buffer (50 mM Na phosphate, pH 6.8, 5% SDS, 40 mM DTT, 5 mM EDTA, 5 mM EGTA, and 15% glycerol). The samples were heated for 5 min at 95°C, sonicated three times for 30 s and, in between intervals, kept for 30 s at 95°C; the procedure was repeated, and after an additional 10 min at 95°C, the insoluble material was removed by centrifugation. Total protein was determined by the protein quantification kit (Bio-Rad Laboratories), and extracts were loaded equally. Separation of total protein extracts was performed by standard procedures (8 and 10% SDS-PAGE). Proteins were electrotransferred to 0.1-µm nitrocellulose membranes (GE Healthcare) by wet blotting in Towbin buffer (25 mM Tris-HCl, pH 8.8, 192 mM glycin, 0.1% SDS, and 10% methanol). Membranes were stained with 0.5% Ponceau S. Secondary antibodies were used with Super Signal (Thermo Fisher Scientific) as a substrate (Reichelt and Magin, 2002). Antibodies and dilutions are listed in Table S2.

Mouse concepti at different gestational ages were prepared. For light microscopy and immunofluorescence, tissues were snap frozen in isopentane precooled at −80°C and cryosectioned (10 µm). Immunofluorescence analysis was performed as follows: tissues were snap frozen in isopentane precooled at −80°C and stored at the same temperature. Frozen sections (8–12 µm thick) were fixed in acetone at −20°C for 10 min and dried for a few hours before further processing. All antibodies were diluted in TBS containing 1% BSA, as stated in Table S2. Primary mouse monoclonal antibodies were detected with subclass-specific secondary antibodies to minimize background. Slides were mounted in ProLong Gold antifade reagent (Invitrogen; Reichelt and Magin, 2002). Antibodies and dilutions are listed in Table S2. For routine histology, tissues were fixed overnight at 4°C in 4% formalin, sequentially incubated at 4°C overnight in 15% and 30% sucrose, embedded in paraffin, and sectioned (5 µm). Sections were placed on Superfrost Plus slides (Menzel-Gläser) and dried. After deparaffination, sections were stained with hematoxylin and eosin (H&E; Reichelt and Magin, 2002).

The images of the H&E-stained paraffin sections were acquired using a fluorescence microscope (Axioplan 2; Carl Zeiss, Inc.) with a Plan-Neofluar 10× 0.30 NA objective and a Plan-Apochromat 63× 1.4 NA oil immersion objective at RT using a camera (AxioCam MR; Carl Zeiss, Inc.). Image analysis and processing were performed using AxioVision 4.6 (Carl Zeiss, Inc.) and Photoshop 6.0 (Adobe) software. Whole mount images were taken using a stereoscopic zoom microscope (model 1500; Nikon) with an HR Plan-Apochromat 1–3× zoom objective at RT using a camera (DS-U2/L2; Nikon). Image analysis and processing were performed with Photoshop software. Immunofluorescent samples were analyzed with a fluorescence laser-scanning confocal microscope (LSM 710; Carl Zeiss, Inc.). For immunofluorescence microscopy and data processing, image stacks were collected with confocal microscope (TCS SP5; Leica) or a laser-scanning microscope (LSM 710) with a Plan-Apochromat 63× 1.4 NA oil immersion objective at RT, and projection views of stacks were produced with the aid of Amira software (Visage Imaging). LUT (lookup table; brightness and gamma) was adjusted using Photoshop. In some instances, single focal planes are shown.

E9.5 embryos were trypsinized, and single-cell suspensions were fixed in 2% PFA/PBS for 10 min at 4°C, washed with PBS, and permeabilized with 0.25% Triton X-100/PBS for 5 min at 4°C. The cells were washed and then stained for 30 min at 4°C with E-cadherin (Invitrogen) in 0.5% BSA/TBS and, after washing, were labeled with goat anti–rat Alexa Fluor 488 (Dianova) for 15 min at 4°C. Cells were washed again and stained for 30 min at RT with cleaved poly(ADP-ribose) polymerase–phycoerythrin (BD) in 0.5% BSA/TBS. Cells were sorted on a flow fluorocytometer (FACSCanto II or LSRII; BD), and data were analyzed using FlowJo software (Tree Star, Inc.).

We used antibodies against biotinylated antiavidin (Vector Laboratories), antidigoxigenin (Boehringer Ingelheim), K8/K18 (Progen), K19 (Troma-3; American Type Culture Collection), desmoplakin (gift from D. Garrod, University of Manchester, Manchester, England, UK), vimentin (T.M. Magin laboratory), ZO-1 (Zytomed Systems GmbH), occludin (Invitrogen), E-cadherin (Invitrogen), ninein (gift from J.B. Rattner, University of Calgary, Calgary, Alberta, Canada), 4E-BP1, P–4E-BP1 (Thr37/46), p70 S6K, P–p70 S6K (Thr389), eIF2-α, P–eIF2-α (Ser51), AMPK, P–AMPK-α (Thr172), Raptor, P-Raptor (Ser792), Akt, P-Akt (Ser473), P (Thr)-MAPK/Cdk substrates (all from Cell Signaling Technology), GLUT1 (Millipore), GLUT3 (Alpha Diagnostic International, Inc.), Hsc70/Hsp70 (Stressgen), 14-3-3–β (Santa Cruz Biotechnology, Inc.), and secondary antibodies (Dianova). Dilutions are listed in Table S2.

E9.5 embryos and yolk sacs were dissected from the uterus, and the head was retained for genotyping. Embryos and yolk sac were incubated separately in Met-free Dulbecco's modified Eagle's medium (Invitrogen) containing 25 mM Hepes buffer, 10% dialyzed fetal calf serum, 1% nonessential amino acids, 1× Na pyruvate, and 1% Glutamax (all Invitrogen) at 37°C for 15 min to remove endogenous Met. The medium was discarded, and tissues were labeled with 100 µl of ³⁵S-labeled Met/Cys (1,000 Ci/mmol; 0.1 mCi/ml) in Met-free Dulbecco's modified Eagle's medium for 1 h at 37°C. The medium was aspirated, and tissues were washed in ice-cold PBS. Proteins were precipitated with 10% trichloroacetic acid, and incorporated radioactivity was measured by liquid scintillation. The rate of ³⁵S-labeled Met/Cys incorporation per minute per milligram of protein was calculated using the Bradford reagent (Bradford, 1976).

E9.5 embryos and yolk sacs were dissected from the uterus. Embryos and yolk sac together were incubated in M2⁻ medium (94.59 mM NaCl, 4.78 mM KCl, 1.19 mM KH₂PO₄, 1.19 MgSO₄, 1.71 mM CaCl₂, 4.0 mM NaHCO₃, 21 mM Hepes, and 4 g/liter albumin bovine fraction V; all from Sigma-Aldrich) without glucose or its metabolites, with 5.5 mM glucose (Sigma-Aldrich), or with 5.5 mM deoxyglucose (Sigma-Aldrich) for 10 min at 37°C. Subsequently, embryos were lysed in boiling Laemmli buffer and analyzed by Western blotting for total and P-AMPK levels.

Fig. S1 shows additional characterizations of the genetic engineering at the keratin type II gene locus. Fig. S2 demonstrates that the deletion of keratins does not induce vimentin. Tables S1 and S2 provide the details of the primers and the antibodies, respectively, used in this study.

We thank Prof. A. Bradley for providing the AB2.2 ES cell line, snlp feeder cell line, and MICER clones for targeting and Mrs. R. Maniu for the blastocyst injections. We gratefully thank Prof. J.B. Rattner for his gift of ninein antiserum and E. Endl, A. Dolf, and P. Wurst from the Flow Cytometry Core Facility at the Institute for Molecular Medicine and Experimental Immunology, University of Bonn. We thank our former colleague Dr. M. Hesse for stimulating ideas and input in the experimental design. We are grateful to our colleagues M. Hatzfeld and M. Hoch for critical input. Also, we thank colleagues who provided antibodies.

This project was funded by Deutsche Forschungsgemeinschaft (German Research Council MA 1316/7) grant SFB832, Bonfor (Cytoskeletal Research Group), and the Bonner Forum Biomedizin.