Occludin controls HIV transcription in brain pericytes via regulation of SIRT-1 activation

Cell cultures

Primary human brain capillary pericytes (ScienCell, Carlsbad, CA, USA) belonging to different batches and donors were cultured with 5% CO₂ at 37°C in serum- and growth factor–complemented growth medium (ScienCell) and used from passages 2 to 7. Pericyte cultures were characterized as positive for αSMA, NG2, HLA-DRIII, and cluster of differentiation (CD)13, -14, -34, -45, -73, -79, and -146, but negative for desmin and CD90 and -105. Human embryonic kidney (HEK)-293 cells were cultured up to 20 passages. U937 cells were differentiated with 20 nM phorbol-12-myristate-13-acetate for 24 h and cultured with RPMI enriched with 10% fetal bovine serum (Thermo Scientific–Life Technologies, Carlsbad, CA, USA). Human peripheral blood monocytes obtained from different donors were cultured at 37°C for 3 d in DMEM enriched with 10% pooled true human serum, 1% l-glutamine, 0.1% gentamicin, and 6 ng/ml macrophage colony-stimulating factor (M-CSF) to induce differentiation. Medium was later exchanged for a similar growth medium but without M-CSF, and cells were incubated for an additional 4 d before use in experiments.

HIV production, infection, and quantitation

HIV1-NL43 proviral DNA, also encoding interdomain green fluorescent protein (iGFP), was produced by transfection of HEK-293T/17 cells. The infection procedure was performed with cell-free virus (6). Transcription rates were quantified by measuring GFP fluorescence with excitation set at 390 nm, cutoff at 495 nm, and spectral detection in the 500–570 nm range. Representative spectra are shown in Supplemental Fig. S2.

Protein quantitation by spectrally resolved in-cell ELISA

Cells were cultured in 96-well plates, stained with DNA/RNA-binding anthraquinone (DRAQ)-5 (1:50; Cell Signaling Technology, Danvers, MA, USA), and fixed with ice-cold methanol. HIV-infected cells were fixed with ice-cold methanol and acetone (50:50 v/v). Nonspecific binding sites were blocked with 5% fat-free milk in Tris-buffered saline–Tween 20. The cells were immunolabeled for 2 h at room temperature with primary antibodies and then incubated with fluorescently labeled secondary antibodies. The following primary antibodies and dilutions (in PBS) were used: total and phospho-Ser⁴⁷ SIRT-1, mouse anti-NFκB-p65, and rabbit anti-NFκB-p65 AcK310 (all 1:100; Abcam, Cambridge, MA, USA) and mouse monoclonal anti-occludin (1:50; clone OC-3F10, Alexa-Fluor-594 conjugated; Thermo Scientific–Life Technologies). The following secondary antibodies were used at a 1:100 dilution in PBS: donkey anti-goat, anti-mouse (Dylight-680 and -800; Thermo Scientific–Pierce, Rockford, IL, USA) and anti-rabbit (Dylight-680, Dylight-800, or Alexa Fluor 405 and Pacific Blue) or goat anti-mouse (Alexa-Fluor-488; both from Thermo Scientific–Life Technologies). Readings were performed to detect and quantify the fluorescence spectral intensity of the bound antibodies. The spectrally resolved in-cell ELISA (SPRICE) data were normalized to DRAQ5 fluorescence. Representative spectra are shown in Supplemental Fig. S2.

Detection of NADH in living cells

Live-cell fluorescence spectroscopy was performed in pericytes. NADH was quantified by excitation at 350 nm, and its emission spectrum was recorded from 430 to 470 nm. Living HEK-293 cells were analyzed by confocal microscopy (510 LSM META equipped with nonlinear optics; Zeiss, Jena, Germany). NADH was excited with a titanium-sapphire biphotonic infrared laser locked at 720 nm, and its fluorescence was detected spectrally at 436–475 nm. Representative spectra are shown in Supplemental Fig. S3. Images were quantified in ImageJ [National Institutes of Health (NIH), Bethesda, MD, USA].

Quantitation of NADH oxidase activity

Treated cultures were scrapped on ice in extraction buffer containing 1% Triton X-100, 25 mM 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES)/NaOH (pH 7.4), 150 mM NaCl, 4 mM EDTA, and 1:500 protease inhibitor cocktail; transferred into Eppendorf (Hamburg, Germany) tubes; sonicated; and centrifuged. Protein concentration was determined by bicinchoninic acid (BCA) assay (Thermo Scientific-Pierce). Supernatants were diluted with aqueous solutions of NADH to achieve a final concentration of 3.5 mg/ml protein and a 5–500 µM range of NADH. Light absorbance at 340 and 260 nm was used to measure NADH and NAD⁺, respectively.

Occludin purification and determination of enzymatic activity

The CC domain of murine occludin (aa 264–521) was generated recombinantly in Escherichia coli and purified (14). The purified full C-terminal domain of human occludin was obtained commercially (TrueORFGold RC206468, codons 266–522; OriGene, Rockville, MD, USA). SDS-PAGE images of both constructs are shown in Supplemental Fig. S3.

To measure enzymatic activity, we diluted occludin fragments in elution buffer and added to them to a series of NADH aqueous solutions. Spectral absorbance was acquired for NAD⁺. Kinetic parameters were obtained by diluting the occludin/NADH solutions in reaction buffer and detecting NAD⁺ light absorbance every 30 s for 15 min.

Occludin depletion and overexpression

Occludin was silenced by transfection (Lipofectamine 2000; Thermo Scientific-Life Technologies) with 6.25 pM per well anti-occludin 27-mer human small interfering (si)RNA (Trilencer-2, UGCACCAAGCAAUGACAUAUAUGGT; OriGene) in RNA resuspension buffer (100 mM KAc and 30 mM HEPES; pH 7.0). Control siRNA [Trilencer-27 universal scrambled (SCR) negative control siRNA; OriGene] or custom made control (AAAGAGCGACUUUACSCACdTdT; Dharmacon, Lafayette, CO, USA) was used. Typical siRNA incubation lasted 16 h, as occludin’s half-life is 11 h.

For occludin overexpression experiments, cells grown to 75% confluence were transfected (Lipofectamine 2000) with N-terminally tagged occludin with yellow fluorescent protein (YFP) or C-terminally tagged occludin with DDK (TrueORFGold RC206468; Origene). Occludin levels were quantified by SPRICE. Representative spectral recordings and immunoblots of occludin are shown in Supplemental Figs. S2 and S3, respectively.

Immunofluorescence, confocal microscopy, and immunoblot analysis

Cells were fixed and processed for immunofluorescence (6). Occludin was labeled with an Alexa Fluor 594–conjugated monoclonal anti-occludin (C-terminal domain) antibody (1:100; Thermo–Scientific–Life Technologies). C-terminal-binding protein (CtBP)-1 was detected with an anti-CtBP1 antibody (1:200; Abcam) paired with a secondary anti-rabbit antibody conjugated with Pacific Blue (Thermo Scientific–Life Technologies). Cell nuclei were stained with DRAQ5 (1:200; Abcam). Live-cell imaging was performed on HEK-293 cells, with the cell membranes stained with 0.5% trypan blue, on an LSM 510 confocal microscope (Zeiss) equipped with a spectral detector. HEK-293 cells expressing YFP-tagged occludin were used as the control.

For immunoblot analysis, the cells were lysed, diluted in denaturing protein loading dye, heated, and loaded onto 5–20% SDS-Tris glycine extended (TGX) precast gels (Bio-Rad, Hercules, CA, USA) (15). Samples were transferred to PVDF membranes, immunolabeled, and detected by chemiluminescence. Representative immunoblots of NFκB-p65 and SIRT-1 are shown in Supplemental Fig. S2.

In silico analysis

Predictions of the tridimensional structure of the C-terminal domain of occludin (fragments aa 266–522, 266–415, and 323–522) were performed with I-TASSER (http://zhanglab.ccmb.med.umich.edu/I-TASSER/; University of Michigan, Ann Arbor, MI, USA). Prediction of the putative binding partners of the same fragments was performed with the protein ligand–binding site prediction server COACH (http://zhanglab.ccmb.med.umich.edu/COACH/; University of Michigan). Sequence similarity searches and homology were performed with BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi; National Center for Biotechnology Information, NIH). Docking of NADH in occludin was calculated with AutoDock Vina (Scripps Research Institute, San Diego, CA, USA). UCSF Chimera (UCSF Resource for Biocomputing, University of California, San Francisco, CA, USA) was used for molecular visualization.

Statistical analysis

Data were tested for normality with the D’Agostino-Pearson and Shapiro-Wilk tests with a confidence interval of 99%. For univariate datasets, a 1-way ANOVA, and for nonparametric data, the Kruskal-Wallis test, were performed to determine the probabilities. Groups to be compared were selected during experimental design, and the Fisher least-squares difference post hoc test was used to compare such groups. Multivariate datasets were analyzed with 2-way ANOVA. Binary datasets were compared by using the Welch-corrected t test or Mann-Whitney U test if the data were nonparametric. Significance was considered at P = 0.01 for all cases.

HIV replication inversely correlates with occludin levels in human BBB pericytes

Brain pericytes were infected with cell-free HIV-1, engineered to encode GFP as a transcriptional reporter. The maximum transcription rate occurred at 48 h, decreased thereafter to a minimum rate reached at 96 h, and remained stable after that time point (Fig. 1A). Viral replication rates measured by HIV-specific RT assay were consistent with the observed decrease in HIV replication rate (Fig. 1B). Levels of transcriptionally active (acetylated) NFκB-p65 peaked at 48 h after infection and decreased thereafter, correlating with the observed changes in HIV transcriptional efficiency (Fig. 1C). We also investigated the dynamics of occludin expression in HIV-infected pericytes. Occludin was depleted 48 h after infection, recovered at 96 h, and was overexpressed at 120 h (Fig. 1D, 1E), inversely mirroring the changes in HIV replication rates. Using specific protein kinase inhibitors, we found that the activity of p38-MAPK was needed for the initial elevation in HIV transcription and concurrent loss of occludin 48 h after infection, whereas ROCK activity was active in the subsequent occludin recovery and decreased HIV transcription (Supplemental Fig. S1A–D). Autocrine modulation of HIV transcription by interferons was excluded, as IFN-α, -β, and -γ levels remained unaffected at 48 h after infection (Supplemental Fig. S1E).

Occludin depletion causes a robust increase in viral transcription

To explore the possibility that it negatively influences HIV transcription, occludin was silenced, followed by HIV infection and evaluation of its transcription efficiency by GFP quantitation. Reduction of occludin levels induced an increase in HIV transcriptional efficiency 48 h after infection when compared to similarly infected wild-type (WT) and SCR negative control siRNA–treated pericytes (Fig. 1F, G). Culturing WT pericytes with the supernatants of HIV-infected/occludin-deficient pericytes resulted in a 3-fold increase in viral transcription, as compared to WT pericytes cultured with supernatants of HIV-infected/WT pericytes (Fig. 1H). Thus, relatively low variations in viral inoculum can result in much larger changes in the transcriptional rates of infected cells.

This possibility was further explored by infecting WT pericytes with different loads of virus originating from a single stock. Increasing the viral inoculum over a 200% range [100,000–350,000 scpm (scintillation counts per minute)] resulted in and elevated viral release into the supernatants by only 27% at 48 h after infection, whereas transcriptional efficiencies increased >200% (Fig. 1I). At lower inoculation titers (100,000–200,000 scpm) the rate of increase in viral release was more accentuated than that at higher titers, whereas elevated transcriptional efficiency was prominent only at higher inoculation titers when the rate of viral release stabilized (Fig. 1J). These findings indicate that relatively small differences in HIV release from occludin-depleted pericytes can cause a robust increase in viral transcription in newly infected cells.

Occludin levels regulate SIRT-1 expression and activation

Because HIV transcriptional efficiency is associated with NFκB activity (16), which, in turn, is modulated by the histone deacetylase SIRT-1 (17), the impact of occludin on the NFκB/SIRT pathway was evaluated. Both occludin depletion (Fig. 2A) and HIV infection (Fig. 2B) resulted in decreased SIRT-1 expression. The NADH-dependent nuclear translocation of CtBP1 has been associated to decreased SIRT-1 expression (18), the subcellular localization of CtBP1 was explored. Decreased cytoplasmic and increased nuclear localization of CtBP1 was found in the occludin-deficient pericytes (Fig. 2C, D) and in their HIV-infected counterparts (Fig. 2E, F).

We next evaluated the phosphorylation state of SIRT-1 as a function of occludin expression and HIV infection by measuring phosphorylation at Ser⁴⁷ which, together with phosphorylated Ser²⁷ and Ser⁵³⁰, results in enhanced deacetylase activity (19). A decrease in phosphorylated Ser⁴⁷ SIRT-1 was found in occludin-deficient (Fig. 3A) and HIV-infected (Fig. 3B) pericytes. To confirm that SIRT-1 levels are associated with occludin expression, total SIRT-1 and its Ser⁴⁷ phosphorylation were quantified in occludin-overexpressing pericytes. The initial effect of HIV on occludin levels was replicated by treating pericytes with occludin siRNA for 12 h and then allowing the cells to recover by exchanging siRNA-containing medium for normal growth medium. This procedure resulted in occludin overexpression 48 h after medium was exchanged (Fig. 3C). In addition, the process was enhanced when SIRT-1 activity was repressed with HR73 (20 µM) for 12 h, indicating that SIRT-1 modulates occludin recovery in pericytes by providing a feedback mechanism. The levels of both total SIRT-1 and its Ser⁴⁷ phosphorylated form were elevated in occludin-overexpressing pericytes (Fig. 3D, E).

To evaluate the effect of HIV infection on these events, we first induced pericytes to overexpress occludin and then infected them 24 h after siRNA recovery. Occludin levels were lower in pericytes infected for 48 h compared to those in the noninfected controls. However, in occludin-recovered/HIV-infected cells, occludin levels were still higher than in WT/noninfected pericytes (Fig. 3F). SIRT-1 expression and phosphorylation was then measured in the same recovered and infected pericytes. Under these conditions, SIRT-1 expression and the levels of its phosphorylated form were indistinguishable between infected and noninfected cells (Fig. 3G, 3H). To determine whether SIRT-1 recovery influences HIV transcription, we analyzed the same cultures spectrometrically to quantify GFP expression. Infection of occludin-recovered pericytes resulted in 3 times less HIV transcription than in WT cells (Fig. 3I).

Activation of SIRT-1 was also evaluated by measuring levels of acetylated NFκB-p65 (Lys³¹⁰), which is deacetylated by SIRT-1. Both occludin deficiency and HIV infection elevated acetyl NFκB-p65. In addition, blocking SIRT-1 activity with HR73 (20 µM, 24 h) enhanced levels of acetyl NFκB-p65 that resembled those in occludin-deficient cells (Supplemental Fig. S1F, G).

Occludin is an NADH oxidase

To this point, our observations indicated that occludin modulates HIV infection by controlling the expression and activation of SIRT-1. On the other hand, SIRT-1 activation depends on NAD⁺, and its transcriptional repressor CtBP1 requires NADH to translocate into the nuclei. To clarify possible mechanisms underlying these effects in the onset of occludin expression, we studied the structure of the C-terminal domain of occludin. The crystal structure of the molecule has only been resolved for the CC domain (20), which accounts for the most distal third of the C-terminal domain (Fig. 4A). Therefore, we performed in silico modeling to predict the structure of the full C-terminal domain. This putative model contains a flexible hinge region (Fig. 4B, C, yellow) that partially encircles the CC domain (Fig. 4B, C, white), connecting it with the rest of the molecule (Fig. 4B, dark green). The ZO-1-binding site is located away from this hinge, which also contains the occludin phosphorylation switch. A putative NADH-binding site was identified in a pocket formed by complementation of the CC domain (Fig. 4B, C, blue) with the hinge (Fig. 4B, C, green). The amino acid sequence corresponding to the modeled region is shown in Fig. 4D. Comparison of the predicted model with the known structures of several proteins indexed in structural databases indicated similarities with oxidoreductases that use NADH/NAD⁺ as a cosubstrate (Tables 1 and 2).

Similar modeling studies of intracellular domains of other TJ molecules revealed that the N-terminal domain of MARVEL D3 and the C-terminal domain of tricellulin, but not of claudin-5, share some structural similarities to other enzymes (lipases and DNA mismatch repair proteins, respectively; Supplemental Tables S1, S2), but not oxidoreductases. MARVEL D3 and tricellulin belong to the same protein family as occludin, implying a different evolutionary role for the members of this family of proteins. The intracellular domains of claudins are sufficiently small to prevent the formation of the structural foldings usually associated with enzymes.

To biochemically characterize occludin-mediated NADH and NAD⁺ interconversion, we incubated an NADH solution with the full C-terminal domain of occludin, originally expressed and purified from HEK-293 cells. Light-absorbance spectroscopy showed an increased NAD⁺:NADH ratio of the solution upon incubation with occludin, but not with occludin-free buffer or heat-denatured (100°C) occludin (Fig. 5A, J). In our experimental conditions (NADH 50–350 µM, 38°C, pH 7.3, with Ca⁺⁺ and Mg⁺⁺), occludin converted NADH into NAD⁺ at a maximum rate (V_max) of 499.6 nmol/s with a Michaelis constant (K_m) of 97 µM (Fig. 5B). Because our model suggested an NADH-binding pocket formed by complementation of the CC domain with the hinge, we explored whether the isolated CC domain of occludin also has effects on NADH/NAD⁺ conversion. Addition of the CC domain of murine occludin (99% similar in sequence to human) to a series of NADH solutions increased the conversion rate to a V_max of 30.91 µmol/s with a K_m of 229.5 µM (Fig. 5C, J). Thus, the full C-terminal domain binds NADH with higher affinity than the CC domain; however, the CC domain can convert NADH to NAD⁺ at high rates. Therefore, it is likely that the hinge, besides complementing the binding pocket and increasing the affinity for NADH, regulates the catalytic rate of the enzyme.

To explore the role of occludin as an oxidase in a cellular context, we added a lysate of HEK-293 cells to a NADH solution, resulting in an elevated NAD⁺:NADH ratio. This ratio increased further upon addition of the murine C-terminal domain, indicating enhanced NAD⁺ production (Fig. 5D). These findings were then validated in living cells. Biphotonic confocal microscopy evidenced a decrease in NADH content in living HEK-293 cells transiently overexpressing full-length occludin (Fig. 5E). Similarly, fluorescence spectroscopy showed a decrease in NADH content in living HEK-293 cells that overexpressed full-length occludin (Fig. 5F). These changes correlated functionally with nuclear translocation of CtBP1. Confocal microscopy of occludin-overexpressing HEK-293 cells showed increased cytosolic levels and less nuclear localization of CtBP1 than was observed in WT controls (Fig. 5G). Quantitation of occludin and CtBP1 evidenced a 3-fold increase in occludin expression that correlated with an ∼3-fold decrease in nuclear CtBP1 (Fig. 5H). To evaluate whether the occludin-induced changes in NADH levels found in HEK-293 cells also occur in pericytes, we measured NADH content in occludin-deficient pericytes. Occludin depletion resulted in enhanced cellular NADH levels when compared to WT cells (Fig. 5I). These observations indicated that the changes in subcellular localization of CtBP1 in pericytes (see Fig. 2C–F) were indeed caused by changes in cellular levels of NADH regulated by occludin levels. Variation in NADH levels correlated proportionally with variation in cellular flavin adenine dinucleotide FAD⁺ levels, indicating that the change in NADH/NAD⁺ ratio subsequent to alterations in occludin levels, did not alter the cellular redox ratio calculated as FAD⁺:NADH⁺FAD⁺ (Supplemental Fig. S3).

Correlating with the enzymatic role of occludin C-terminal domain, occludin immunoblot analysis in pericyte lysates indicated relatively low abundance of the canonical 65 kDa isoform 1 (typical epithelial occludin), but robust expression of isoforms 4 and 5 (31 and 23 kDa, respectively) (Supplemental Fig. S3H).

HIV replication correlates negatively with occludin levels in human macrophages

In the last series of experiments, we investigated to determine whether the observed regulatory effects of occludin on HIV transcription could be reproduced in HEK-293 cells and then in monocytes/macrophages, which share functional immunologic similarities with pericytes. Confocal microscopy and fluorescence spectroscopy of HEK-293 cells infected with HIV for 48 h revealed that occludin overexpression reduced HIV transcription to approximately half, compared with similarly infected WT controls (Fig. 6A, green signal; 6B). In addition, depletion of occludin resulted in a 3-fold increase in HIV transcription 48 h after infection of terminally differentiated U-937 monocytes (Fig. 6C), mimicking the results observed in HEK-292 cells and pericytes. We further explored these events in human primary macrophages differentiated from peripheral blood monocytes. Occludin depletion in noninfected macrophages resulted in augmented nuclear localization of CtBP1. HIV infection increased nuclear levels of CtBP1 48 h after infection in WT macrophages, an effect that was not further influenced by occludin depletion (Fig. 6D). Occludin depletion in human macrophages resulted in enhanced HIV transcription, as shown by confocal microscopy (Fig. 6E) and fluorescence spectroscopy (Fig. 6F), fully reproducing our observations in pericytes, HEK-293 cells, and U-937 monocytes.