Membrane-bound Trafficking Regulates Nuclear Transport of Integral Epidermal Growth Factor Receptor (EGFR) and ErbB-2*

Background: EGFR is translocated to the inner nuclear membrane through the INTERNET (integral trafficking from the ER to the nuclear envelope transport) pathway. Results: INTERNET regulates EGFR and ErbB-2 but not FGFR-1. Conclusion: At least two different pathways of nuclear transport exist for cell surface receptors. Significance: This provides a new direction for investigating the trafficking mechanisms of various nuclear RTKs. Nuclear localization of multiple receptor-tyrosine kinases (RTKs), such as EGF receptor (EGFR), ErbB-2, FGF receptor (FGFR), and many others, has been reported by several groups. We previously showed that cell surface EGFR is trafficked to the nucleus through a retrograde pathway from the Golgi to the endoplasmic reticulum (ER) and that EGFR is then translocated to the inner nuclear membrane (INM) through the INTERNET (integral trafficking from the ER to the nuclear envelope transport) pathway. However, the nuclear trafficking mechanisms of other membrane RTKs, apart from EGFR, remain unclear. The purpose of this study was to compare the nuclear transport of EGFR family proteins with that of FGFR-1. Interestingly, we found that digitonin permeabilization, which selectively releases soluble nuclear transporters from the cytoplasm and has been shown to inhibit nuclear transport of FGFR-1, had no effects on EGFR nuclear transport, raising the possibility that EGFR and FGFR-1 use different pathways to be translocated into the nucleus. Using the subnuclear fractionation assay, we further demonstrated that biotinylated cell surface ErbB-2, but not FGFR-1, is targeted to the INM, associating with Sec61β in the INM, similar to the nuclear trafficking of EGFR. Thus, ErbB-2, but not FGFR-1, shows a similar trafficking pathway to EGFR for translocation to the nucleus, indicating that at least two different pathways of nuclear transport exist for cell surface receptors. This finding provides a new direction for investigating the trafficking mechanisms of various nuclear RTKs.

We previously showed that the EGFR internalized from the cell surface after endocytosis is trafficked to the nucleus by a retrograde route from the Golgi to the endoplasmic reticulum (ER) (17) and is then translocated via the INTERNET (integral trafficking from the ER to the nuclear envelope transport) mechanism (18) to the inner nuclear membrane (INM) through the nuclear pore complexes (NPCs); this process is regulated by importin ␤ (19). Importin ␤ or importin ␣/␤ forms a complex with molecules carrying nuclear localization signals (NLSs), and importin ␤ directly associates with the nucleoporins and plays a critical role in the canonical model of nuclear transport through NPCs (20,21). Importin ␤ is also involved in the nuclear transport of ErbB-2 and FGF receptor-1 (FGFR-1) (22,23), in addition to EGFR, and importin ␤ co-localizes with ErbB-2 in the endosome (23), implying that the endocytic vesicle and/or endosome may function as a vehicle, using importin ␤ as a driver to carry cargo proteins in nuclear translocation.
In addition to the INTERNET mechanism for EGFR nuclear translocation (19), another pathway, named INFS (integrative nuclear FGFR-1 signaling) (24), has been proposed, in which FGFR-1 is released from the cytoplasmic membranes into the cytosol and then transported into the nucleus by indirect asso-ciation with importin ␤ (22) since FGFR-1 does not contain an NLS. Similarly, another study reported that EGFR is extracted from the ER membrane to the cytosol via the ER-associated degradation (ERAD) pathway, in which cytosolic EGFR is transported into the nucleus (25).
To date, the exact mechanisms by which RTKs embedded in the endosomal membrane translocate into the nucleus through NPCs remain controversial. The purpose of this study was to compare the nuclear transport of EGFR family proteins with that of FGFR-1 to improve our understanding of the trafficking mechanisms of various nuclear RTKs. In this study, we used the digitonin permeabilization system and cellular fractionation assays to investigate the difference between the two pathways of nuclear trafficking in EGFR/ErbB-2 and FGFR-1. We found that the membrane-bound INTERNET mechanism regulates the nuclear trafficking of EGFR/ErbB-2 but not that of FGFR-1. Thus, at least two different pathways exist to translocate cell surface receptors into the nucleus.
Cell Culture-A431 human epidermoid carcinoma cells, Her-5 cells (derived from Swiss 3T3 cells by stable transfection with the human EGFR expression vector), Swiss 3T3 mouse fibroblasts, and MDA-MB-468 and MDA-MB-453 human breast carcinoma cells were maintained in DMEM/F-12 supplemented with 10% FCS and 100 g/ml penicillin and streptomycin.
Permeabilization by Digitonin-Cells were washed in a transport buffer (20 mM HEPES pH 7.0, 110 mM KOAc, 5 mM NaOAc, 2 mM Mg[OAc] 2 , 1 mM egtazic acid, 2 mM dithiothreitol, and 1 mg/ml each aprotinin and leupeptin) and permeabilized with 75 M digitonin at 4°C for 6 min in a transport buffer. Cells were then treated with EGF at 37°C for 30 min in a transport buffer containing 1 mM ATP (Roche, San Francisco, CA).
Biotinylation of Cell Surface Proteins-Cell surface proteins in MDA-MB-453 cells were biotinylated using 1 mM Sulfo-NHS-LC-Biotin (Pierce) at room temperature for 30 min, and the cells were further incubated at 37°C for 15 min. Swiss 3T3 cells maintained in a medium containing 0.5% FCS for 48 h were biotinylated on the cell surface at 4°C for 30 min and then treated with or without FGF-2 (50 ng/ml) at 37°C for 60 min. The biotinylation reaction was quenched with PBS containing 100 mM glycine.
Cellular Fractionation, ER Purification, INM Purification, and Cytosolic/Membrane Fractionation-Cellular fractionation of non-nuclear/nuclear fractions (7) and ER purification (19) were performed as described previously. For INM purification (19), the isolated nuclei were extracted using cellular fractionation and suspended in buffer A (0.25 M sucrose, 50 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 1 mM dithiothreitol, and protease inhibitor mixture (Sigma)). After incubation with 1% (w/v) sodium citrate at 4°C for 30 min, the suspended nuclear pellet was centrifuged at 500 ϫ g for 15 min. The resulting pellet was digested with 250 mg/ml DNase I (Sigma) at 4°C for 14 h. After centrifugation at 10,000 ϫ g for 2 h, the supernatant was collected as a nucleoplasm (NP) portion. The digested pellet was submitted to recentrifugation at 100,000 ϫ g for 20 min on a sucrose gradient (INM-sucrose), and the purified INM fraction was then collected at the 0.25-1.60 M sucrose interface. In addition, another set of digested pellet was resuspended in NETN buffer (150 mM NaCl, 1 mM EDTA, 20 mM Tris, pH 8.0, 0.5% Nonidet P-40, and protease inhibitor mixture) and then sonicated (Sonics Vibra-Cell, amplitude 30; Sonics & Materials, Newtown, CT). After centrifugation at maximum speed, the INM portion was collected. Purification of the cytosolic, organelle membrane, and nucleic fractions was performed using a ProteoExtract Subcellular Proteome Extraction Kit (Calbiochem) with a slight modification. The organelle membraneenrich fraction was further isolated by centrifugation at 20,000 ϫ g for 30 min to remove mitochondria and other large organelles such as cell membrane portion (26,27) after sequential extraction according to the manufacturer's instructions.

EGF-dependent Nuclear Transport of EGFR Involves Membrane-bound
Trafficking-To compare the nuclear translocation of the INTERNET mechanism for EGFR (19) with that of the INFS pathway for FGFR-1 (24), we used the digitonin permeabilization assay, which is known to selectively wash away cytosolic nuclear transporters (28) such as soluble importin ␤ from the cytoplasm. Nuclear translocation of FGFR-1 is regulated by soluble importin ␤, as shown by digitonin pretreatment (22). By quantifying the percentage of cells with nuclear localization of FGFR-1, we confirmed that FGF-2-dependent nuclear transport of FGFR-1 (ab-10646, Abcam) was substantially blocked after digitonin permeabilization in Swiss 3T3 cells (Fig. 1, lower panel, 78% versus 7%), which express endogenous FGFR-1 (supplemental Fig. S1). To investigate whether EGFR uses the same mechanism, we then performed the digitonin permeabilization assay in MDA-MB-468 cells harboring endogenous EGFR but not ErbB-2 or FGFR-1 (supplemental Fig. S1). Interestingly, we found that EGFR can translocate to the nucleus through membrane-bound trafficking, which is different from the classic nuclear import pathways involving cytosolic proteins, such as the soluble form of importin ␤ (Fig. 2).
Digitonin permeabilizes the cholesterol-rich plasma membrane but leaves the NE and other intracellular membrane organelles structurally intact (28). Thus, in MDA-MB-468 cells, most of the soluble proteins in the non-nuclear fraction are selectively released from the cells during digitonin permeabili-zation ( Fig. 2A, lanes 5 and 6 versus lanes 1 and 2). To determine whether the ER membrane was intact after digitonin permeabilization, we further performed confocal immunofluorescence of calregulin, which resides in the ER lumen. The signal from calregulin was not detected in the digitonin-permeabilized cells but was readily detectable when the ER membrane was permeabilized with Triton X-100, suggesting that the structure of the ER remained intact during digitonin treatment, as expected (Fig. 2B). After validating the properties of digitonin-permeabilized cells, we were surprised to observe that, in contrast to FGFR-1 (Fig. 1), the nuclear transport of EGFR (sc-03, Santa Cruz Biotechnology) still occurred after EGF stimulation for 30 min following digitonin pretreatment in MDA-MB-468 cells (Fig. 2C), and we detected no marked difference in the percentage of cells with EGFR nuclear translocation with (66%) or without (58%) digitonin permeabilization (Fig. 2D). We obtained similar results from the experiments performed in EGFR stable transfectant Her-5 cells (29) (Fig. 2E, lower panel, 52% versus 51%; supplemental Fig. S1). We further used different antibodies to confirm the subnuclear localization of EGFR (Ab-13, Thermo Scientific) and FGFR-1 (sc-121, Santa Cruz Biotechnology) in the digitonin permeabilization systems by examining sequential photosections of nuclei in MDA-MB-468 (supplemental Fig. S2) and Swiss 3T3 (supplemental Fig. S3) cells in multiple planes. Consistently, the results of serial focal sections clearly showed that digitonin pretreatment did not affect the processes of nuclear transport of EGFR (supplemental Fig. S2, A versus B). However, in the case of FGFR-1, FGF-2 strongly induced nuclear FGFR-1 as evident from apparent white signals in the merged images representing localization of FGFR-1 (green) in the nucleus (blue) in the absence of digitonin (supplemental Fig. S3A) while no nuclear FGFR-1 was detectable in the presence of digitonin (supplemental Fig. S3B). In addition, we performed the digitonin permeabilization assay in HeLa cells transfected with an exogenous construct of either GFP-EGFR (supplemental Fig. S4, A and B) or GFP-FGFR-1 (supplemental Fig. S4, C and D) and examined sequential photosections of a nucleus of a HeLa cell in multiple planes by green fluorescence detection. Similar to the results of endogenous EGFR and FGFR-1 in response to digitonin pretreatment, EGFdependent nuclear transport of GFP-EGFR was still detectable following digitonin pretreatment, whereas FGF-2-induced nuclear transport of GFP-FGFR-1 was severely impaired after digitonin permeabilization. Together, these results suggest that membrane-associated proteins that cannot be washed away by digitonin pretreatment may be involved in the mechanism underlying EGFR transport into the nucleus, thus supporting the possibility that membrane-bound trafficking regulates the nuclear transport of EGFR (19), which is different from FGFR-1 nuclear translocation and from the classic nuclear import pathways involving soluble proteins.
ER Membrane-associated Importin ␤ Is Responsible for Membrane-bound Trafficking of EGFR Nuclear Transport-As a low concentration of importin ␤ can be detected around the nucleus in the digitonin-permeabilized cells (30) and importin ␤ is required for nuclear translocation of EGFR (31) and ErbB-2 (23), we asked whether importin ␤ could associate with intracellular membranes. Indeed, we detected a small pool of remaining importin ␤ in the non-nuclear extracts of the digitonin-permeabilized cells, which contained mainly membraneassociated proteins (Fig. 3A, lanes 1 and 2 versus lanes 3 and 4). In addition, importin ␤ in the nucleus was unchanged during digitonin treatment as digitonin was unable to permeabilize the nuclear membrane (Fig. 3A, lanes 5 and 6 versus lanes 7 and 8).
The purity of the non-nuclear and nuclear fractions was validated as evident from the absence of the ER marker calregulin in the nuclear fraction (Fig. 3A, lanes [5][6][7][8] and the nuclear marker lamin B in the non-nuclear fraction (Fig. 3A, lanes 1-4).
Saksena et al. (32) demonstrated that importin-␣-16, a membrane-associated isoform of Spodoptera frugiperda importin-␣, associates with the ER membranes and is involved in membrane protein trafficking. To determine whether the remaining importin ␤ also localizes with the ER membrane after digitonin permeabilization, we used an OptiPrep gradient technique, which is an established method to isolate the ER fraction as described under "Experimental Procedures." Immunoblotting analysis of the purified ER fractions with an anti-importin ␤ antibody showed that digitonin pretreatment did not affect the level of the ER-associated importin ␤ (Fig. 3B, lane 4 versus lane  3), whereas we observed an apparent decrease in importin ␤  . EGF-dependent nuclear transport of EGFR involves membrane-bound trafficking. A, most of the soluble proteins in the non-nuclear fraction were selectively released from the cells during the digitonin (75 M) permeabilization. MDA-MB-468 cells maintained in a serum-starved medium for 24 h were permeabilized with digitonin at different concentrations, followed by treatment with or without EGF for 30 min. Proteins in the non-nuclear and nuclear fractions obtained by cellular fractionation were then analyzed by Coomassie Blue staining. B, the structure of the ER remained intact after digitonin permeabilization. MDA-MB-468 cells maintained in a serum-starved medium for 24 h were treated with EGF for 30 min. After fixation, the cells were permeabilized with 0.5% Triton X-100 or 75 M digitonin, incubated with the indicated antibodies, and analyzed by confocal microscopy. Staining of CD44 antibody, a cell plasma membrane marker, is shown. All nuclear areas were confirmed by DAPI staining. Bar, 5 m. C and D, EGFR was transported to the nucleus after EGF stimulation in digitonin-permeabilized MDA-MB-468 cells. Cells maintained in a serum-starved medium for 24 h were permeabilized with digitonin and then treated with EGF as described under "Experimental Procedures." Cells were immunostained for EGFR and analyzed using confocal microscopy. Boxed areas are shown in detail in the insets. Nuclear transport of EGFR is shown in the inset panel in the yellow merged signals. Bar, 5 m (C) or 10 m (D). Bar diagram indicates percentage of cells with nuclear localization of EGFR, calculated from 50 cells. E, EGFR was transported to the nucleus after EGF treatment in digitoninpermeabilized Her-5 cells. Same as Fig. 2D but for MDA-MB-468 cells instead of Her-5 cells. Bar, 5 m. Nuclear transport of EGFR is shown in the inset panel in the yellow merged signals (white arrows). Bar diagram indicates that we detected no difference in the percentage of cells with EGFR nuclear translocation with (51%) or without (52%) digitonin permeabilization.
level in the non-nuclear/non-microsomal fraction containing cytosolic proteins (Fig. 3B, lane 2 versus lane 1). Therefore, the results suggest that soluble importin ␤ in the cytoplasm is released from the cells after digitonin treatment, while a por-tion of importin ␤ is associated with the ER membrane and unaffected by digitonin permeabilization, similar to the previous report (30). Consistently, co-localization (yellow signals in the merged images) of the remaining importin ␤ (red) and the FIGURE 3. Importin ␤ remains associated with the ER membrane in digitonin-permeabilized cells. A, a small amount of importin ␤ remained in the non-nuclear fraction after digitonin permeabilization in A431 cells. The purity of the non-nuclear and nuclear fractions was validated by calregulin (non-nuclear marker) and lamin B (nuclear marker). B, the remaining importin ␤ was associated with the ER membrane after digitonin permeabilization. The ER fractions in cells treated with or without digitonin were isolated using OptiPrep gradient purification and subjected to immunoblotting with the indicated antibodies. Similar results were obtained in two independent experiments (p value Ͻ0.05, calculated from Student's t test). C, co-localization of calnexin and the remaining importin ␤ in MDA-MB-468 cells after digitonin permeabilization was indicated by the yellow signals in the merged images (arrows). Boxed areas are shown in detail in the insets. Bar, 5 m. Bar diagram indicates the percentage of cells with co-localization of importin ␤ and the ER membrane protein calnexin, calculated from 35 cells. D, co-localization of calnexin and the remaining importin ␤ in Swiss 3T3 cells after digitonin permeabilization is indicated by the yellow signals in the merged images (arrows). Localization of importin ␤ in the cytoplasmic portion is represented by the red signals in the merged images (arrowheads). Boxed areas are shown in detail in the insets. Bar, 5 m. ER membrane protein calnexin (green) in MDA-MB-468 cells was detectable by confocal microscopy even after digitonin permeabilization (Fig. 3C, left panel, arrows), whereas importin ␤ signals in the cytoplasmic portion (Fig. 3C, inset 2 versus inset 1) but not the nuclear portion (Fig. 3C, inset 4 versus inset 3) were clearly diminished. Furthermore, quantification of the percentage of cells with co-localization showed that digitonin did not inhibit the co-localization of importin ␤ and calnexin (Fig. 3C, right panel, 91% versus 85%). Together with the previous findings that importin ␤ is involved in the nuclear transport of EGFR and ErbB-2 (23,31) and that importin ␤ further regulates INTERNET membrane trafficking of EGFR nuclear transport to the INM (19), these results suggest that the ER membrane-associated importin ␤ is responsible for membrane-bound trafficking of EGFR to the nucleus, favoring the membrane-bound INTERNET pathway but not the ERAD pathway involving cytosolic EGFR (19,25).
Interestingly, in addition to MDA-MB-468 cells used in EGFR experiments, importin ␤ signals in the cytoplasmic portion was also diminished but remained in the ER membrane in Swiss 3T3 cells we used for the FGFR-1 experiments (Fig. 3D). We found that yellow signals in the merged images representing co-localization of the remaining importin ␤ and calnexin were still detectable following digitonin pretreatment (Fig. 3D,  inset 3 versus insets 1 and 2, arrows); meanwhile, importin ␤ signals in the cytoplasmic portion were obviously decreased (Fig. 3D, inset 6 versus insets 4 and 5, arrowheads), similar to the results shown in Fig. 3C. Together with the previous findings indicating that digitonin permeabilization results in different outcomes of nuclear transport of EGFR ( Fig. 2 and supplemental Fig. S2) and FGFR-1 ( Fig. 1 and supplemental Fig. S3), our data suggest that the soluble importin ␤ regulates FGFR-1 nuclear transport via the INFS model and the ER membraneassociated importin ␤ mediates nuclear transport of EGFR through the membrane-bound INTERNET pathway.
Sec61␤ Regulates Nuclear Transport of ErbB-2 from the INM to the NP-Researchers have proposed that nuclear transport of ErbB-2 is similar to that of EGFR (23,31,33). Thus, we next asked whether ErbB-2 is also trafficked into the nucleus from the cell surface to the INM through the membrane-bound INTERNET mechanism. To this end, we analyzed ErbB-2 proteins in the INM using two cellular fractionation methods, as described under "Experimental Procedures," in MDA-MB-453 cells in which ErbB-2 is readily detectable but EGFR and FGFR-1 were undetectable, compared with those in other cells, including MDA-MB-468, Her-5, and Swiss 3T3 cells (supplemental Fig. S1). In brief, cell surface proteins were labeled with biotin, and then the biotinylated proteins were biochemically separated into various fractions, including the non-nuclear, outer nuclear membrane (ONM), NP, and INM pellet. The INM pellet was subjected to centrifugation on a sucrose gradient (INM-sucrose) and immunoblotted with emerin, an INM marker, indicating the recovery of INM in fraction 7 (Fig. 4A,  upper panel). In fraction 7, representing INM, we detected biotinylated cell surface ErbB-2 precipitated by streptavidinagarose beads in MDA-MB-453 cells (Fig. 4A, lower panel, lane  2), indicating that cell surface ErbB-2 was localized to the INM.
In addition, ultrastructural studies using immuno-EM further supported the localization of ErbB-2 inside the NE (Fig. 4B, arrows) and NP (Fig. 4B, arrowheads). We also performed con- focal microscopy to examine the subnuclear localization of ErbB-2 (Ab-3, Calbiochem) in the digitonin pretreatment assay in MDA-MB-453 cells and found that nuclear transport of ErbB-2 was still detected after digitonin permeabilization (supplemental Fig. S5). These results strongly suggest that cell surface ErbB-2 translocates to the nucleus through the membranebound mechanism, similar to that of EGFR (19). However, the distribution of FGFR-1 in the purified INM fractions was undetectable after precipitation using streptavidin-agarose beads (Fig. 4C, lower panel, lanes 1 and 2), indicating that nuclear transport of FGFR-1 is distinct from that of ErbB-2 and does not involve membrane-bound trafficking to the INM. In addition to the above-mentioned biochemical assays, we further demonstrated the INM localization of EGFR and ErbB-2, but not FGFR-1, using confocal microscopy (supplemental Fig. S6), further supporting the nuclear transport of EGFR and ErbB-2 is entirely membrane-bound. HeLa cells were transfected with an exogenous construct of GFP-EGFR (supplemental Fig. S6A), GFP-ErbB-2 (supplemental Fig. S6B), and GFP-FGFR-1 (supplemental Fig. S6C), and detected co-localization of the green fluorescence and the INM marker emerin. We observed the yellow signals representing co-localization of GFP-EGFR/ emerin (supplemental Fig. S6, A, insets 2 and 3, arrows) and GFP-ErbB-2/emerin (supplemental Fig. S6, B, insets 1 and 2,  arrows), but no obvious yellow signals were detected in GFP-FGFR-1-expressed HeLa cells after FGF-2 stimulation (supplemental Fig. S6C, inset 2), indicating both of EGFR and ErbB-2, but not FGFR-1, can be targeted to the INM.
Because the translocon Sec61␤ has been shown to be localized in the INM, where it associates with EGFR and regulates EGFR trafficking to the NP (19), we wanted to further study whether Sec61␤ is also involved in and interacts with the INMlocalized ErbB-2. To answer this question of ErbB-2-Sec61␤ association, we subjected another set of biotinylated lysates to subsequent subnuclear fractionation to obtain INM portions in NETN buffer because INM-sucrose fractions containing sucrose as a solvent were more challenging in a co-immunoprecipitation assay. First, we validated the purity of various portions by their specific markers (Fig. 5A). We did not detect any cross-contamination with the process of cellular fractionation in the INM portions (Fig. 5A, lanes 5 and 6) as evident from the presence of the INM marker LAP2 accompanying the absence of the ER markers calnexin and calregulin, the early endosome protein Rab5, the late endosome protein LAMP1, and the nuclear protein Sp1 in the INM portions. We then detected biotinylated cell surface ErbB-2 precipitated with streptavidinagarose beads both in the INM portion (Fig. 5B, lower panel,  lane 2), which was consistent results shown in Fig. 4A, and in the NP portion (Fig. 5B, lower panel, lane 4), as expected (23). Interestingly, we further showed the interaction of Sec61␤ with biotinylated ErbB-2 in the INM (Fig. 5B, lower panel, lane 2) but not in the NP (Fig. 5B, lower panel, lane 4) and detected no expression of Sec61␤ in the NP portions in MDA-MB-453 cells (Fig. 5B, upper panel). In addition, knocking down Sec61␤ expression by two individual small interfering RNAs (siRNA) targeting Sec61␤, siRNA-Sec61␤ (Fig. 5C) and siRNA-Sec61␤-2 (data not shown), increased ErbB-2 expression level in the INM (Fig. 5C, lane 2 versus lane 1) and decreased it in the NP (Fig. 5C, lane 4 versus lane 3), similar to previous results of EGFR in MDA-MB-468 cells (19). These results indicate that cell surface ErbB-2 is transported to the INM, where it associates with Sec61␤, through membrane-bound trafficking, which is similar to the INTERNET trafficking of EGFR, and further suggest that the INM-localized Sec61␤ also assists in the release of membrane-embedded ErbB-2 from the lipid bilayer of the INM to the nucleus.
Nuclear Transport of EGFR Is Distinct from that of FGFR-To strengthen the notion that nuclear translocation of the membrane-bound INTERNET mechanism for EGFR is different from that of the INFS pathway for FGFR-1, we further performed a cellular fractionation assay using ProteoExtract Subcellular Proteome Extraction kit (Fig. 6). Briefly, MDA-MB-468 (Fig. 6A) and Swiss 3T3 cells (Fig. 6B) were biochemically separated into various fractions, including the cytosolic, organelle membrane, and nucleic fractions upon their individual ligands treatment. We found that both of nuclear transport of EGFR (Fig. 6A, lanes 7-9) and FGFR-1 (Fig. 6B, lanes 7-9) occurred in a ligand-dependent manner; nevertheless, different expression ratios of EGFR and FGFR-1 were detected in the cytosolic and organelle membrane-enriched fractions. The results showed that EGFR was mainly expressed in the organelle membraneenriched fraction and increased in response to EGF (Fig. 6A, lanes [4][5][6], whereas level of EGFR in the cytosolic fraction was very low and had no significant induction after EGF stimulation (Fig. 6A, lanes 1-3), favoring the membrane-bound INTER-NET pathway for EGFR nuclear transport. In contrast, FGFR-1 was mainly detected in the cytosolic fraction and cytosolic FGFR-1 was increased upon FGF-2 treatment in a time-dependent manner (Fig. 6B, lanes 1-3), however, only a small amount of FGFR-1 was detected in the organelle membrane fraction and not to respond to ligand stimulation (Fig. 6B, lanes 4-6), supporting the INFS model proposed for FGFR-1 nuclear transport in which FGFR-1 is released from the cytoplasmic membrane into the cytosol and then transported into the nucleus.

DISCUSSION
Here, we built on the results of previous studies indicating that membrane-bound trafficking from the Golgi-ER to the INM is involved in the nuclear transport of EGFR (17)(18)(19) to propose a comprehensive model of the nuclear transport of EGFR and ErbB-2, but not FGFR-1, involving membranebound trafficking from the cell surface to the INM (Fig. 6C). In this model, the entire trafficking process of EGFR/ErbB-2 from the cell surface to the nucleus remains bound in the membrane (Fig. 6A) and is targeted to the INM through INTERNET trafficking, which is mediated by ER membrane-associated importin ␤ (Figs. 2 and 3). Similar to the nuclear transport of EGFR (19), ErbB-2 also translocates from the cell surface to the INM, where it interacts with the translocon Sec61␤ (Figs. 4 and 5). Unlike EGFR/ErbB-2, cell surface FGFR-1 cannot be detected in the INM (Fig. 4C). In addition, cytosolic FGFR-1 is increased upon ligand stimulation and accompanied by a small amount of FGFR-1 in the membrane fraction (Fig. 6B). We are aware of the size difference between FGFR-1 in the cytosolic and organelle membrane fractions. It is known that FGFR-1 has two alternative splice variants producing FGFR-1 with two different size, 145 and 120 kDa (22,34). Further systemic investigation is required to distinguish whether the two forms of FGFR-1 represent the two alternative splice variants. It has been reported that FGFR-1 serves as a transmembrane RTK as well as a soluble cytoplasmic protein because FGFR-1 has an atypical transmembrane domain (35). Thus, the INFS mechanism of nuclear transport of FGFR-1 via release of FGFR-1 to the cytosol (24) may be different from that of EGFR/ErbB-2, which is consistent with the different results that we found in digitonin-permeabilized cells ( Fig. 1  Although the exact mechanisms by which endosomal membrane-embedded EGFR/ErbB-2 is transported into the nucleus through NPCs remain unclear, we have here provided several new pieces of evidence here to show that, in addition to the ERAD pathway for cytosolic EGFR (25), the membrane-bound INTERNET mechanism regulates nuclear trafficking of EGFR and ErbB-2. Our digitonin permeabilization studies showed that EGF-induced nuclear translocation of EGFR still occurred while ER membrane-associated importin ␤ was present, even A, membrane-bound and nuclear EGFR was increased after EGF stimulation. MDA-MB-468 cells maintained in a serum-starved medium for 24 h were treated with EGF for a different period. Cells were then biochemically separated using ProteoExtractா Subcellular Proteome Extraction kit with a slight modification into various fractions, including cytosolic (lanes 1-3), organelle membrane (lanes 4-6), and nucleic fractions (lanes 7-9), and subjected to immunoblotting with the indicated antibodies. B, FGF-2 induced cytosolic and nuclear FGFR-1. Swiss 3T3 cells maintained in a medium containing 0.5% FCS for 24 h were treated with FGF-2 in a different period. Same as Fig. 6A but for MDA-MB-468 cells instead of Swiss 3T3 cells. C, comparison of nuclear trafficking mechanisms between EGFR/ErbB-2 and FGFR-1. Proposed model of nuclear transport of EGFR and ErbB-2 from the cell surface to the nucleus through INTERNET membrane-bound trafficking (left), which is different from that of FGFR-1 through the INFS pathway (right). The comprehensive trafficking pathways of EGFR and ErbB-2 from the cell surface to the nucleus remain in a membrane-bound environment. Cell surface EGFR and ErbB-2 are translocated to the nucleus from the ER membrane to the INM via INTERNET trafficking, which is mediated by ER membrane-associated importin ␤. The INM-localized EGFR and ErbB-2 are released from the lipid bilayer to the nucleus through the interaction of the translocon Sec61␤. On the other hand, the nuclear transport of FGFR-1 is through INFS mechanism (24), in which cytosolic FGFR-1 is transported into the nucleus by an indirect association with importin ␤, proposing FGFR1 is chaperoned to the nucleus by FGF-2 ligand harboring a NLS in an importin ␤-dependent manner (43,44). The scale of the diagram does not reflect the relative sizes of different molecules or subcellular structures. Imp␤, importin ␤; ER, endoplasmic reticulum; NPC, nuclear pore complex; ONM, outer nuclear membrane; INM, inner nuclear membrane.
without the soluble form of importin ␤ (Figs. 2, 3 and supplemental Figs. S2 and S4). These results strongly support the previous finding that the absence of importin ␤ expression significantly accumulated EGF-dependent EGFR translocation in the ONM/ER and inhibited that to the INM and the nucleus (19), indicating the ER membrane-bound importin ␤ is indeed responsible for the EGFR INTERNET membrane trafficking. In addition, subnuclear fractionation, sucrose gradient, and confocal immunofluorescence assays also support the translocation of EGFR (19) (supplemental Fig. S6) and ErbB-2 (Figs. 4, 5 and supplemental Fig. S6) from the cell surface to the INM via the INTERNET pathway. Moreover, previous investigators showed that Sec61 at the ER enables the release of EGFR into the cytoplasm; however, they mentioned that "attempts to detect cytosolic EGF receptor in vivo have not been successful" (25). Consistent with that statement, we also could not detect cytosolic EGFR expression upon EGF treatment in intact cells using an OptiPrep gradient technique (Fig. 3B, lane 1). Only a small amount of EGFR in the cytosolic fraction was found but with no ligand-dependent induction by another biochemical approach (Fig. 6A, lane 1-3).
One obvious difference between the INTERNET and ERAD models is the time period over which the experiments were performed. In this study, cells were treated with EGF for 30 min to demonstrate the INTERNET mechanism. However, in the ERAD system (25), both the in vitro ER retrotranslocation assay and the nuclear EGFR study were performed after 3-6 h of EGF treatment, which is different from the previous study in the same cell line showing that EGFR levels were elevated and reached a plateau at 30 -60 min after EGF treatment and then declined toward the baseline by 6 h after EGF treatment (36). In addition to EGF treatment, results for heparin-binding EGFlike growth factor (HB-EGF)-dependent EGFR nuclear localization also support this short-term elimination, within 2 h (37). This inconsistency in the timing of experiments could explain the discrepancy in results between in vivo and in vitro experiments.
In addition to EGFR and ErbB-2, EGFR family ligands, including EGF, pro-transforming growth factor-␣, and pro-HB-EGF, have also been found in the nucleus (38 -41). However, unlike the EGFR family, which contains NLSs (36), these ligands do not contain NLSs. Interestingly, pro-HB-EGF has been shown to be targeted to the INM, yet no interaction of pro-HB-EGF with importin proteins has been detected, probably because of a lack of NLSs (42). Whether these ligands translocate via active transport, as larger molecules do, or via passive diffusion, as smaller molecules do, is unclear. The present study raises an interesting possibility that these ligands associate with their cognate receptor EGFR and translocate together from the ER to the INM via the membrane-bound INTERNET mechanism. A previous report showing I 125 -EGF cross-linked to EGFR can be detected in the nucleus also favors this possibility (7). Further investigation is required to determine whether EGFR serves as an active transporter to translocate its ligands into the nucleus.
Endocytosis is involved in the nuclear transport of EGFR (31) and ErbB-2 (23), and the nuclear transport of cell surface EGFR and ErbB-2 depends on a membrane-bound environment (19) (Figs. [2][3][4][5][6]. We therefore speculated that EGFR and ErbB-2 in the nucleus are still associated with the components of the early secretory machinery from the cell surface membrane. We previously showed the co-localization of the membrane-bound endosomal marker EEA1 (early endosome antigen 1)/EGFR (31) and EEA1/ErbB-2 (23) near the NE, which also supports the notion that nuclear transport of EGFR and ErbB-2 involves membrane-bound trafficking. In summary, the current study comparing the nuclear transport of EGFR family proteins with that of FGFR-1 will lead to a better understanding of the trafficking mechanisms of various nuclear RTKs.