Cell Type- and Tissue Context-dependent Nuclear Distribution of Human Ago2*

Argonaute-2 protein (Ago2), a major component of RNA-induced silencing complex (RISC), has been viewed as a cytoplasmic protein. In this study, we demonstrated by immunofluorescence confocal microscopy that Ago2 is distributed mainly as a nuclear protein in primary human foreskin keratinocytes in monolayer cultures and their derived organotypic (raft) cultures, although it exhibits only a minimal level of nuclear distribution in continuous cell lines such as HeLa and HaCaT cells. Oncogenic human papillomavirus type 16 (HPV16) or type 18 (HPV18) infection of the keratinocytes does not affect the nuclear Ago2 distribution. Examination of human tissues reveals that Ago2 exhibits primarily as a nuclear protein in skin, normal cervix, and cervical cancer tissues, but not in larynx. Together, our data provide the first convincing evidence that the subcellular distribution of Ago2 occurs in a cell type- and tissue context-dependent manner and may correlate with its various functions in regulation of gene expression.


Argonaute-2 protein (Ago2), a major component of RNA-induced silencing complex (RISC), has been viewed as a cytoplasmic protein. In this study, we demonstrated by immunofluorescence confocal microscopy that Ago2 is distributed mainly as a nuclear protein in primary human foreskin keratinocytes in monolayer cultures and their derived organotypic (raft) cultures, although it exhibits only a minimal level of nuclear distribution in continuous cell lines such as HeLa and HaCaT cells.
Oncogenic human papillomavirus type 16 (HPV16) or type 18 (HPV18) infection of the keratinocytes does not affect the nuclear Ago2 distribution. Examination of human tissues reveals that Ago2 exhibits primarily as a nuclear protein in skin, normal cervix, and cervical cancer tissues, but not in larynx. Together, our data provide the first convincing evidence that the subcellular distribution of Ago2 occurs in a cell type-and tissue context-dependent manner and may correlate with its various functions in regulation of gene expression.
Argonaute 2 (Ago2), 3 also known as EIF2C2 (eukaryotic translation initiation factor 2C 2), is an endonuclease and a major component of the RNA-induced silencing complex (RISC). Ago2 binds microRNAs, siRNAs, and Piwi-interacting RNAs and mediates the loading of these small noncoding RNAs onto RISC to recognize specific targets through base-pairing, ultimately leading to mRNA translation inhibition or degradation (1,2). Ago2 has been viewed exclusively as a cytoplasmic protein because it is ectopically expressed in the cytoplasm (3) and is distributed in the cytoplasmic RNA granules, including P-bodies and stress granules (4). However, recent data indicate that Ago2 might be a nuclear-cytoplasmic shuttling protein in cells (3,(5)(6)(7)(8) and may be involved in various nuclear processes including nuclear RNAi (5), pre-mRNA alternative splicing (9,10), and transcription (11,12). In support of these observations, Ago2 has been found partially in the nuclei of HeLa (3), A549 (5), and T47D (5) cells and senescent WI-38 fibroblast cells (12). To date, the observed nuclear Ago2 and its function remain uncertain and an important subject of scientific debate. To address the uncertainty of whether nuclear Ago2 is significant, we examined, by a comprehensive microscopic analysis, Ago2 distribution in several cell types in monolayer or organotypic cultures, as well as in human tissues from cervix, skin, and larynx. From examining such a range of cell types in different growth environments, we found the nuclear distribution of human Ago2 to be context-dependent physiologically.

Experimental Procedures
Cells and Tissues-HeLa, HEK293, and HaCaT cells were cultivated in DMEM supplemented with 10% FBS(HyClone, GE Healthcare). Primary human foreskin keratinocytes (HFKs) were cultured as described previously (13,14) in EpiLife calcium-free medium (Life Technologies, Thermo Scientific) supplemented with 5% fetal bovine serum (HyClone) and human keratinocyte growth supplement (Life Technologies, Thermo Scientific) in the presence of mitomycin C-treated J2 feeder cells. Raft cultures derived from primary HFKs with or without human papillomavirus type 16 (HPV16) or type 18 (HPV18) infection were prepared as described (13,14). Formaldehydefixed paraffin-embedded tissue sections of human normal cervix, cervical cancer, skin, and larynx were obtained from US Biomax, Inc. (Rockville, MD).
Subcellular Fractionation and Western Blotting Analysis-Nuclear and cytoplasmic fractionation of HeLa cells and primary keratinocytes were obtained using a nuclei isolation kit from Sigma (catalog number NUC101) according to the protocol provided by the company, with slight modification. Briefly, cells growing in a 35-mm dish were scraped in 1ϫ PBS and lysed using 350 l of Nuclei EZ lysis buffer. Nuclei were pelleted by centrifugation (500 ϫ g) for 5 mins at 4°C. The supernatant was collected as the cytoplasmic fraction, and the obtained nuclear pellet was resuspended/washed gently one time with 1ϫ PBS and then resuspended in 350 l of lysis buffer. Equal volumes of nuclear and cytoplasmic extracts or total protein samples were examined by a standard Western blotting protocol after solubilizing in Laemmli SDS buffer supplemented with 5% (v/v) 2-mercaptoethanol.
Pre-absorption of Anti-Ago2 Antibody with Recombinant HA-Ago2 Protein-The recombinant Ago2 protein was immunopurified from HEK293 cells transfected with an HA-tagged hAgo2 expression vector (Addgene). Briefly, the whole cell extract prepared at 48 h after transfection was incubated with anti-HA antibody-conjugated agarose beads. The bound HA-Ago2 on the beads was then eluted by incubation with HA peptides three times to obtain three elution fractions (E1, E2, and E3) (15), and its purity in each fraction was confirmed by a Silver Quest silver staining kit (Life Technologies, Thermo Fisher Scientific) after SDS-PAGE. To block the specific activity of Ago2 antibody, the purified HA-Ago2 (100 ng) from the mixed fractions 2-3 was incubated with 100 l of anti-Ago2 antibody diluted (1:50) in blocking buffer (2% BSA in PBST (PBS with 0.02% Tween 20)) for 20 min at room temperature, and the immunocomplexes were removed by centrifugation at 13,000 rpm for 10 min at 4°C. The supernatant was used as the Ago2-adsorbed antibody for immunostaining.
Immunofluorescence Analysis (IFA) and Confocal Microscopy-Indirect immunofluorescence antibody staining was performed as described previously (16,17). Briefly HeLa and HaCaT cells grown on poly-D-lysine-treated glass coverslips were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature. The formalin-fixed paraffin-embedded tissue sections were dewaxed by heating at 60°C for 1 h followed by a 10-min xylene wash and then rehydrated in a decreasing ethanol series. After dewaxing, the slides were boiled for 15 min in 1ϫ Antigen Retrieval Citra Plus buffer (catalog number HK086, BioGeneX, Fremont, CA) using a microwave. After an extensive wash in deionized water, the slides were immersed in 1ϫ PBS. Before staining, both cell and tissue sections were permeabilized for 15 min by 0.5% Triton X-100 in 1ϫ PBS for 15 min at room temperature followed by a 2-h incubation with blocking buffer (2% BSA in 0.05% Tween PBS (TPBS)) at 4°C. The primary Ab at 1:50 dilution in blocking buffer was incubated overnight at 4°C. After three washes with TPBS, the samples were incubated with Alexa Fluor-conjugated secondary antibodies diluted at 1:500 in blocking buffer for 2 h at 37°C followed by extensive wash with TPBS. Before mounting, the nuclei were counterstained with Hoechst dye 33258 (Sigma). Confocal fluorescence images were collected with a Zeiss LSM780 laserscanning microscope (Carl Zeiss, Inc., Thornwood, NY) equipped with 20ϫ Plan-Apochromat (numerical aperture, 0.8) and 63ϫ Plan-Apochromat (numerical aperture, 1.4) objective lenses. The x-y pixel sizes of 0.4 and 0.07 m and optical slice thicknesses of 1.5 and 0.9 m were used to acquire confocal images with the 20ϫ and 63ϫ objectives lenses, respectively. Maximum intensity projections and isosurface volume renderings (both of which are three-dimensional reconstructions of confocal z-stacks) were generated using the Imaris (version 8.0.2) image processing software (Bitplane, Inc., Concord, MA). The profile module of the ImageJ software (http:// rsb.info.nih.gov/ij/) was used to determine signal intensities for each channel.
Induction of Stress Granule Formation by Arsenite Treatment-HeLa cells or primary keratinocytes grown on glass coverslips were treated with 0.5 mM sodium arsenite for 30 min, followed by PBS wash and fixation.

Results and Discussion
Minimal Nuclear Distribution of Ago2 in HeLa and HaCaT Cells, Two Continuous Keratinocyte Cell Lines-To detect cell compartmental distribution of endogenous Ago2 protein by confocal microscopy, we first confirmed the specificity of a commercially available anti-Ago2 antibody by Western blotting using the total cell extract from HeLa, a cervical cancer cell line infected by HPV18, with or without Ago2 siRNA (si-Ago2) treatment. The selected antibody was able to detect a single, prominent protein band with the expected size of ϳ100 kDa corresponding to endogenous Ago2 (Fig. 1A). The treatment of HeLa cells with Ago2-specific siRNA led to reduction of the Ago2-specific protein, confirming the antibody specificity toward Ago2 (Fig. 1A). The selected antibody was then used in Ago2 IFA on HeLa cells. In agreement with previous studies (3, Nuclei were counterstained with Hoechst dye for DNA. Images were captured by confocal microscopy. Scale bar ϭ 10 m in A and 20 m in B. C, subcellular ago2 was examined in nuclear (N) and cytoplasmic (C) fractions of HeLa and primary HFKs by Western blotting using a mouse anti-Ago2 antibody. Cytoplasmic tubulin and nuclear hnRNP C1/C2 were used to provide fractionation efficiency. The relative amounts of Ago2 in the nuclear fraction (being set to 1) over the cytoplasmic fraction are shown at the bottom of the Ago2 blot after normalizing with nuclear hnRNP C1/C2 and cytoplasmic tubulin, respectively. D, purified HA-Ago2 protein for antibody absorption assays. Recombinant HA-Ago2 protein was purified from the cellular extract of HEK293 cells by an anti-HA antibody affinity purification. The purity of HA-Ago2 (arrow) in elution fractions (E1-3) was determined by SDS-PAGE with silver staining. E and F, subcellular Ago2 distribution in HFK-derived raft cultures without (E, HFK) or with HPV16 (F, HFK16) infection. Endogenous Ago2 was detected by immunostaining with an anti-Ago2 antibody (green). Ago2-absorbed anti-Ago2 antibody in F and secondary antibody only (2 o Ab only, F) served as a negative control staining. Nuclei were counterstained with Hoechst dye. DIC, differential interference contrast image. Images were captured by confocal microscopy. Scale bar ϭ 20 m in E and F. Dashed boxes represent zoomed areas. G, quantification of the cells with nuclear Ago2 staining in HFK raft cultures without or with HPV16 (HFK16) or HPV18 (HFK18) infection. Mean ϩ S.D. from 150 cells in each condition from three independent experiments. 1 o , mouse anti-Ago2 antibody; 2 o , rabbit anti-mouse IgG antibody. 5, 6), a small portion of Ago2 was found to be nuclear (Fig. 1B). To eliminate the possible effect of HPV viral oncogene expression on Ago2 distribution in HeLa cells, HaCaT, an immortal skin keratinocyte cell line lacking HPV infection, was further used for the study, and we found that a small portion of Ago2 in HaCaT cells is also nuclear, whereas the majority of Ago2 are cytoplasmic (Fig. 1C).
However, nuclear and cytoplasmic Ago2 are distinguishable for their responses to arsenite, a chemical inducer of cellular stress. Upon cellular stress induced by arsenite, both Ago2 and TIA-1, a stress granule (SG)-specific marker, are recruited to SGs involved in translation inhibition (Fig. 1D) (4,18,19). By IFA staining of Ago2 and TIA-1 in arsenite-treated HeLa cells, we found that both cytoplasmic Ago2 and TIA-1, but not nuclear Ago2 and TIA-1, are co-stained in numerous cytoplasmic SGs (Fig. 1E). In contrast, there is no SG formation in the cell nuclei where Ago2 and TIA-1 remain generally dispersed throughout the nucleus.

Ago2 Is Predominantly a Nuclear Protein in Primary Keratinocytes Both in Monolayer and in Organotypic (Raft) Culture
Conditions-Considering that the physiological state of continuous cell lines growing in monolayer cultures differs significantly from cells in tissues, we performed IFA immunostaining of Ago2 in primary HFKs grown on coverslips. To our surprise, we found a predominant nuclear distribution of Ago2 in primary HFKs ( Fig. 2A) in monolayer culture, which is in sharp contrast to what we saw in HeLa and HaCaT cell lines (Fig. 1, B and C). The nuclear Ago2 distribution became significantly reduced in the HFKs when Ago2 was knocked down (Fig. 2B  and supplemental Fig. S1A). Further nuclear and cytoplasmic fractionation and Western blotting analyses confirmed more Ago2 in the cytoplasm than in the nucleus in HeLa cells, but in contrast, more Ago2 in the nucleus than in the cytoplasm in HFKs (Fig. 2C). In addition, we found that, when compared with HeLa cells (supplemental Fig. 1B), HFKs did not form a visible amount of stress granules in response to arsenite (supplemental Fig. 1C).
Next we examined Ago2 distribution in HFK-derived organotypic cultures that mimic epithelial tissues with characteristic differentiation of keratinocytes from the undifferentiated . Tissue-context nuclear distribution of Ago2. A, immunostaining of endogenous Ago2 in normal cervix, skin, and larynx by anti-Ago2 antibody. B and C, distinguishable subcellular distribution of endogenous Ago2 and PABPC1. Normal or cancerous cervical tissues were double-stained with anti-Ago2 (green) and anti-PABPC1 (red) antibodies (B). Nuclei were counterstained with Hoechst dye (A and B). All images were captured by confocal microscopy. Scale bars ϭ 20 m. Dashed boxes represent zoomed areas. The levels of Ago2 and PABPC1 in the selected cells from normal cervix were measured by a line crossing over the stained cells. Nuclear DNA counterstained with Hoechst dye served to define the cell nuclear and cytoplasmic border (C). The intensity plot of individual measurements in arbitrary units (AU) was shown after normalization to maximum intensities. basal layer to the terminally differentiated cornified layer. To exclude any possible nonspecific staining of the anti-Ago2 antibody, we applied the purified HA-tagged Ago2 (HA-Ago2, Fig.  2D) to specifically absorb anti-Ago2 antibody before its Ago2 staining. IFA staining of HFKs in the organotypic tissues revealed that Ago2 is predominantly localized in the nuclei of basal and parabasal keratinocytes (Fig. 2E). The infection of HFKs in the organotypic tissues with HPV16 or HPV18 does not affect the nuclear distribution of Ago2 ( Fig. 2F and supplemental Fig. S2 A), although both HPVs promote cell proliferation of undifferentiated keratinocytes (20).
The nuclear specific IFA staining of Ago2 in the keratinocytes was further confirmed by using an Ago2-absorbed anti-Ago2 antibody. In this assay, the anti-Ago2 antibody at 1:50 dilution was pre-absorbed by the affinity-purified HA-Ago2 expressed from HEK293 cells (Fig. 2D). When compared with the anti-Ago2 antibody labeling without Ago2 pre-absorption, the pre-absorbed anti-Ago2 antibody exhibited a significant reduction in nuclear Ago2 staining, with the signal intensity nearly close to the background signal from the negative control (staining by the secondary antibody only) (Fig. 2F). These data rule out the possibility of any cross-reactivity of the anti-Ago2 antibody to other nuclear proteins. By counting ϳ150 HFKs in organotypic cultures with or without HPV16 (HFK16) or HPV18 (HFK18) infection, we found that the nuclear Ago2 in all three conditions could be found in nearly ϳ95% of keratinocytes by anti-Ago2 antibody, but only in ϳ3% of the cells by the Ago-2 absorbed anti-Ago2 antibody and in none in the cells by the secondary antibody only (Fig. 2G). Together, we conclude that Ago2 is expressed primarily as a nuclear protein in HFKs.
Intranuclear Detection of Ago2 in HFK-derived Organotypic Tissues Defined by Lamin A/C Staining-Given that Ago2 is mainly a nuclear protein in HFKs, we wished to define whether Ago2 could also be distributed to the nuclear membrane as suggested in a previous study (21). To confirm Ago2 staining inside the nuclear membrane, we used an anti-lamin A/C antibody that recognizes both lamin A and lamin C (due to alternative RNA splicing) in the inner membrane of the nucleus. Double immunostaining of the cells revealed that Ago2 in the nucleus is clearly overlapped with DNA and surrounded by lamin A/C of the nuclear membrane (Fig. 3A). Nuclear distribution of Ago2 in HFKs was found to be independent of HPV 16 or HPV18 infection ( Fig. 3A and supplemental Fig. S2B). We also acquired confocal z-stacks of these samples to examine Ago2 distribution in the different focal planes of the nucleus (Fig. 3B for line-guided orthogonal view of the nuclei and supplemental Fig. S3 for various focus planes up (ϩ) or down (Ϫ) from the selected section). As shown in Fig. 3B and supplemental Fig. S3, a three-dimensional orthogonal view and the focus plane views up or down from the selected section indeed showed more Ago2 distribution inside of the nucleus, but separate from the nuclear membrane. This intranuclear distribution of Ago2, overlapping with the nuclear DNA but not with the nuclear membrane lamin A/C, is measurable by an arbitrary line drawn across three separate cells (Fig. 3B, arrows) and can be seen clearly in the videos generated from confocal z-stacks (supplemental Movies 1 and 2).

Nuclear Distribution of Ago2 Varies from Different Human
Tissues-We next investigated Ago2 distribution in human tissues including cervix, skin, and larynx by immunostaining with the same mouse anti-Ago2 antibody used in Figs. 1-3. We showed that Ago2 protein is expressed primarily as a nuclear protein in normal cervix and skin, but expressed as a major cytoplasmic protein in larynx (Fig. 4A). Normal cervix staining with a rabbit monoclonal anti-Ago2 antibody also gave the same result, showing nuclear Ago2 distribution in the cervix (supplemental Fig. S4). To exclude any possibility of fewer Ago2 staining signals in the cytoplasm than in the nucleus in human cervix being an artifact of tissue fixation and processing, we performed double staining of Ago2 along with cytoplasmic PABPC1 protein both in the normal and in the cancerous cervical tissues. The double immunostaining of the cervical tissues showed an exclusive cytoplasmic distribution of PABPC1, confirming the good preservation of cytoplasmic antigens during tissue fixation and processing (Fig. 4B). Notably, Ago2 was highly enriched in the nucleus when compared with the cytoplasm in the fixed tissues (Fig. 4B, supplemental Movies 3 and 4). A signal intensity plot across a randomly selected cell showed the distinct distribution of nuclear Ago2 from cytoplasmic PABPC1 in the cell compartments (Fig. 4C). Altogether, these data indicate that Ago2 can be either a nuclear or a cytoplasmic protein in a tissue-context dependent manner.
In summary, immunostaining and Western blotting in this study have revealed the minimal presence of Ago2 in the nuclei of HeLa and HaCaT cells, but the predominant presence of Ago2 in the nuclei of primary HFKs and their derived organotypic cultures. Ago2 is also distributed primarily as a nuclear protein in skin, normal cervix, and cervical cancer tissues, but not in larynx. By measuring Ago2 signal intensity in the cells or tissues, the level of Ago2 in the nucleus was found to be much higher than that measured in the cytoplasm in HFKs and in the tissues of skin and cervix (supplemental Fig. S5). Therefore, the predominant cytoplasmic distribution of Ago2 observed in continuous cell lines HeLa or HaCaT cells does not reflect the accurate distribution of Ago2 in primary tissues such as cervix or skin.