Granulin and Granulin Repeats Interact with the Tat·P-TEFb Complex and Inhibit Tat Transactivation*

The cellular positive transcription elongation factor b (P-TEFb), containing cyclin T1 and cyclin-dependent kinase 9 (CDK9), interacts with the human immunodeficiency virus, type 1 (HIV-1) regulatory protein Tat to enable viral transcription and replication. Cyclin T1 is an unusually long cyclin and is engaged by cellular regulatory proteins. Previous studies showed that the granulin/epithelin precursor (GEP) binds the histidine-rich region of cyclin T1 and inhibits P-TEFb function. GEP is composed of repeats that vary in sequence and properties. GEP also binds directly to Tat. Here we show that GEP and some of its constituent granulin repeats can inhibit HIV-1 transcription via Tat without directly binding to cyclin T1. The interactions of granulins with Tat and cyclin T1 differ with respect to their binding sites and divalent cation requirements, and we identified granulin repeats that bind differentially to Tat and cyclin T1. Granulins DE and E bind Tat but do not interact directly with cyclin T1. These granulins are present in complexes with Tat and P-TEFb in which Tat forms a bridge between the cellular proteins. Granulins DE and E repress transcription from the HIV-1 LTR and gene expression from the viral genome, raising the possibility of developing granulin-based inhibitors of viral infection.

stimulate the growth of these cells (3). Granulin mRNA is prominent in several human cells and tissues (4 -6). High levels of GEP expression are found in several human cancers, and its involvement in breast cancer has been examined in some detail (reviewed in Refs. [7][8][9]. Its expression is correlated with tumorigenicity, and the proliferation of malignant human breast carcinoma cells is inhibited by GEP antisense RNA or anti-GEP antibodies. GEP can substitute for estradiol in stimulating DNA synthesis in breast cancer cells, rendering them estrogen-independent and tamoxifen-resistant (10). GEP also stimulates the growth of several other cell lines, including epithelial cells and rat adrenal gland pheochromocytoma PC12 cells (4). In normal tissues full-length granulin and individual granulins mediate the wound response and coordinate host defense and wound repair (11,12).
GEP contains seven and a half repeats of a cysteine-rich motif (CX 5-6 CX 5 CCX 8 CCX 6 CCXDX 2 HCCPX 4 CX 5-6 C) in the order P-G-F-B-A-C-D-E, where A-G are full repeats and P is the half motif (Fig. 1A). The C-terminal region of the consensus sequence contains the conserved sequence CCXDX 2 HCCP and is suggested to have a metal binding site and to be involved in regulatory function (13). Several of the repeat units have been isolated from tissue culture medium, blood, and urine as ϳ6-kDa peptides known as granulins (or epithelins) and are individually biologically active. In some systems they stimulate cell proliferation, whereas in others they inhibit mitosis (7). The GEP repeats are joined by short intergranulin linkers of diverse sequence, containing the cleavage site for the serine proteinase elastase, which is secreted in large quantities by neutrophils during infection and inflammation (12).
The mode of action of GEP is not well understood. Granulin binding sites have been demonstrated, although cell surface receptors have not yet been characterized (14,15). Recently GEP was reported to interact with perlecan, a heparan sulfate proteoglycan found in basement membranes and cell surfaces. The perlecan-GEP interaction was suggested to modulate tumor growth (16). GEP leads to activation of the mitogen-activated protein kinase pathway and to stimulation of cyclin D1 protein expression. This can account for the cellular proliferation activity of granulin and its ability to replace estrogen in inducing the growth of breast cancer cells (17). More recently, we found that GEP can modulate transcription elongation by interacting with human cyclin T1, a component of positive transcription elongation factor b (P-TEFb) (18). P-TEFb is a general transcription factor required for the production of mRNA by RNA polymerase II (19). In addition to cyclin T1, P-TEFb contains the cyclin-dependent kinase CDK9, which phosphorylates the C-terminal domain of the large subunit of polymerase II, thereby facilitating the transition of polymerase II into a productive elongation mode (reviewed in Ref. 20). P-TEFb, and cyclin T1 in particular, is a target for cellular and viral transcription regulators. For example, cyclin T1 binds the HIV-1 transcription activator Tat (21) and allows the generation of full-length rather than prematurely terminated RNA chains (19,22,23). Cyclin T1 is an exceptionally large cyclin with unique sequences in its C-terminal region (Fig. 1B) that interact with regulatory proteins (24 -26). GEP binds specifically to the histidine-rich domain of cyclin T1 (18), which is also necessary for RNA polymerase II binding (27). GEP forms complexes with intact P-TEFb inside cells and represses transcription from the HIV-1 LTR, a promoter that is highly dependent on P-TEFb (18). The C-terminal region of granulin, GrnCDE, serves as a substrate for P-TEFb in kinase assays. Phosphorylation of a Cterminal domain peptide in vitro is inhibited when P-TEFb is bound to GST-GrnCDE, suggesting that granulin may act by repressing the phosphorylation of RNA polymerase II C-terminal domain by P-TEFb (18).
Remarkably, GEP also binds to the Tat proteins of HIV-1, HIV-2, and caprine arthritis encephalitis virus (18,28,29). HIV-1 Tat is synthesized as a protein of 72-101 amino acids (aa), depending on the viral strain and alternate splicing of its mRNA. The one-exon 72-aa form, which is sufficient to activate transcription from the HIV-1 promoter (30,31), is composed of five functional domains (Fig. 1C). GEP interacts with the cysteine-rich domain of Tat, which forms part of the its activation domain required for binding to P-TEFb and for transactivation. These data raised the possibility that granulins could inhibit HIV-1 transcription by binding the Tat activation domain.
Previous work indicated that granulins can inhibit the HIV-1 promoter via tethered cyclin T1 in the absence of Tat (18). In the present study, we aimed first to determine whether granulins can serve as specific inhibitors of Tat transactivation via direct interactions with Tat. We found that the interaction of granulin with Tat can be sufficient for transcriptional repression and that the interactions of GEP with cyclin T1 and Tat are separable. Granulins inhibited Tat transactivation mediated by truncated cyclin T1 that lacks the histidine-rich domain essential for binding granulin. The interactions of granulins with Tat and cyclin T1 are distinguishable with respect to their requirement for zinc ions, and distinct Tat sequences are involved in interactions with granulin and cyclin T1. Next, we assessed the contribution of specific granulins to Tat and cyclin T1 binding and to the inhibition of expression from the HIV-1 promoter. We found that granulins DE and E interact differentially with cyclin T1 and Tat: these granulins bind to Tat in cell extracts and in living cells but do not interact directly with cyclin T1. Granulins DE and E repressed transcription from the HIV-1 LTR and gene expression from the viral genome and are found in complexes with Tat and P-TEFb. Thus granulins DE and E are specific inhibitors of the Tat⅐P-TEFb complex and of viral gene expression.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-Adherent 293T, 3T3, and COS7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Jurkat cells were cultured in RPMI 1640 medium (Invitrogen) supplemented with 2 mM L-glutamine and 10% fetal bovine serum (Sigma-Aldrich).
Yeast Two-hybrid Interactions-Protein interactions were studied in yeast using the MATCHMAKER GAL4 system (Clontech) as described in Hoque et al. (18). The ␤-galactosidase liquid quantification assay was performed according to the manufacturer's instructions.
In Vitro Binding Assay-GrnCDE and GrnPGFBA were synthesized using the in vitro TNT quick-coupled transcription and translation system (Promega). GrnDE, GrnD, and GrnE were synthesized using the coupled wheat germ extract system (TNT, Promega). In vitro synthesized [ 35 S]Met-GrnCDE or [ 35 S]Met-PGFBA (10 l of reactions) were mixed with ϳ2 g of GST or GST fusion proteins, bound to glutathione-Sepharose beads, and incubated for 2 h at 4°C with gentle rocking. To test the effect of different metal ions, various concentrations of the specified salts were added to the binding buffer. The beads were washed extensively in EBCD plus 0.03% SDS. Bound proteins were released by heating to 100°C in Laemmli sample buffer and resolved in 12% polyacrylamide-SDS gels. [ 35 S]Met-labeled proteins were detected by autoradiography.
Preparation of Whole Cell Extracts-Cells were grown to 100% confluency in 100-mm tissue culture dishes, or in suspension to a density of 1 ϫ 10 6 cells/ml, and washed twice with 10 ml of phosphate-buffered saline (minus calcium and magnesium). Cells were then lysed in 1 ml of EBC buffer (50 mM Tris, pH 8.0, 120 mM NaCl, 0.5% Nonidet P-40) containing 1 mM dithiothreitol, 1 mM sodium vanadate, 0.1 mM phenylmethylsulfonyl fluoride, and 1 g/ml of leupeptin, pepstatin A, and antipain. After 30-min incubation on ice the cell lysate was sonicated for 10 s (medium frequency) to homogenize the suspension. Clear supernatants were collected after centrifugation in Eppendorf centrifuge (15 min, 13,000 rpm at 4°C) and stored at Ϫ80°C.
Co-immunoprecipitation-Human 293T cells (5 ϫ 10 6 cells) were transfected with 10 g of total plasmid DNA using Cytofectene (Bio-Rad) according to the manufacturer's instructions. Cells were harvested at 24 h post-transfection and lysed in 1 ml of immunoprecipitation lysis buffer (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 1%Triton X-100, 1 mM EDTA) containing 0.1 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin, 1 g/ml leupeptin, and 1 g/ml antipain. Extracts were cleared, and supernatants were collected after 15-min centrifugation at 4°C (Eppendorf, 13,000 rpm). Antibody (1 g) was bound to protein A-Sepharose or protein G-Sepharose by incubation at 4°C for 2 h with rocking. The conjugates were mixed with cell extract (500 g of total protein) and incubated overnight at 4°C with rocking. Unbound proteins were removed by extensive washing with EBCD buffer containing 0.03% SDS. Protein complexes bound to the conjugated antibody were separated by denaturing polyacrylamide gel electrophoresis and subjected to immunoblotting. Immunoblotting was performed as described previously (38). Anti-FLAG M2 monoclonal antibody was obtained from Sigma; anti-HA rabbit polyclonal antibody, anti-cyclin T1 goat polyclonal, and anti-CDK9 rabbit polyclonal were obtained from Santa Cruz Biotechnology. Detection was by enhanced chemiluminescence (ECL).
Co-localization of Granulins and Tat-Intracellular co-localization experiments were carried out in COS7 cells. Cells were grown on collagencoated cover slips in tissue culture plates to a confluency of 30 -40% and were transfected with the plasmids indicated using Cytofectene (Bio-Rad) according to the manufacturer's instructions. The expression, localization, and co-localization of the fluorescence-tagged proteins were directly analyzed using a Nikon TE-300 inverted fluorescence microscope. Digital images were processed using Adobe Photoshop.
Transfection and Luciferase Assays-3T3 cells (2 ϫ 10 5 cells) were transfected using Lipofectamine TM 2000 according to the manufacturer's instructions. After 24 h the cells were harvested, washed with phosphate-buffered saline, and lysed in 0.1 ml of 1ϫ Promega lysis buffer and assayed for luciferase activity. Jurkat cells (1 ϫ 10 7 cells) were suspended in 250 l of RPMI (without serum) and were transfected with 15 g of DNA by electroporation. Electroporation was carried out at 900 microfarads, 250 V, and 180 ohms. Cells were harvested 15 h post-transfection and assayed for luciferase activity. Luciferase assays were performed with the Promega Dual Luciferase Reporter System according to the manufacturer's instructions.

Granulin Binding to Tat Is Sufficient for Inhibition of Tat
Transactivation-Full-length granulin, as well as its N-terminal (GrnPGFBA) and C-terminal (GrnCDE) regions, inhibit Tat transactivation in mouse 3T3 cells supplemented with human cyclin T1 (18). Because granulin interacts with both cyclin T1 and Tat, the inhibition could be mediated by the binding of granulin to either or both of these transcription factors. Experiments with tethered cyclin T1 showed that granulin can inhibit expression from the HIV-1 promoter in the absence of Tat, i.e. via a granulin-cyclin T1 interaction (18). To determine whether the direct interaction between granulin and cyclin T1 is necessary for inhibition, we employed a truncated form of cyclin T1 (aa 1-333) that does not bind granulin in yeast two-hybrid or GST pull-down assays (18). The N-terminal region of human cyclin T1, containing its cyclin domain and TRM (Tat: TAR recognition motif), is sufficient to permit Tat transactivation in 3T3 cells (33,39).
As shown in Fig. 2A, Tat transactivation in 3T3 cells was greatly enhanced by either full-length human cyclin T1-(1-726) or human cyclin T1-(1-333). In both cases, GrnPGFBA and GrnCDE inhibited Tat transactivation by 3-to 6-fold. Similar results were obtained with another cyclin T1 truncation (1-397; data not shown). Thus, inhibition of Tat transactivation by granulin can occur in the absence of the granulin binding site on human cyclin T1. We note that the inhibition was weaker in the absence of the granulin binding site in the cyclin T1 C terminus. Similarly, inhibition was weaker in the absence of Tat, when cyclin T1 was tethered to the promoter (Ref. 18, and see Fig. 7C below), suggesting that granulin may preferentially target the Tat⅐P-TEFb complex.
To verify that the truncated cyclin T1 binds to CDK9 but not to GrnCDE, we conducted co-immunoprecipitation experiments using transfected 293T cell extracts (Fig. 2B). Both the endogenous cyclin T1 and HA-tagged truncated cyclin T1-(1-333) were co-precipitated with anti-CDK9 antibody (lanes 1 and 6), demonstrating that the truncated protein retains the ability to form P-TEFb complexes. As expected, endogenous cyclin T1 was co-precipitated with FLAG-tagged GrnCDE both in the absence (lane 4) (18)  of GrnCDE or GrnPGFBA with Tat can cause inhibition of transcription from the HIV-1 LTR.
Discrimination between Tat-granulin and Tat-cyclin T1 Interactions-It has been shown that GEP and GrnCDE interact with HIV-1 Tat in vitro and inside cells (18) and that GrnBA interacts with the highly conserved cysteine-rich domain of both HIV-1 Tat and HIV-2 Tat in yeast and in vitro (28). To define the requirements for the interaction of Tat with GrnPG-FBA and GrnCDE, we first conducted GST pull-down assays with a series of Tat deletions (Fig. 3, A and B). Strong binding was observed to full-length one-exon Tat (Tat1-72) and to the truncations lacking either the Tat acidic domain (Tat21-72) or the basic and auxiliary domains (Tat1-48). Strong binding was also observed to the two-exon form of Tat (Tat1-101) and to a deleted form truncated in the core region (Tat1-41) (data not shown). Weak binding was observed for the isolated cysteinerich region (Tat21-37), but binding was lost when the N-terminal half of the cysteine-rich region was removed (Tat30 -72) (Fig. 3B). In the yeast two-hybrid assay, GrnCDE binding was unaffected by point mutations in the acidic (P18A) and cysteine-rich (C22G and C31G) regions of Tat that abrogate its interaction with cyclin T1 (40). Two additional mutations in the cysteine-rich and core regions (C30G and K41A) that prevent cyclin T1 binding also do not interfere with GrnCDE binding in mammalian cells (see Fig. 8D below). Thus, the cysteine-rich domain of Tat is essential for granulin binding, but flanking regions in the Tat core and acidic regions may contribute to binding efficiency. These requirements are similar but not identical to those for Tat's interaction with cyclin T1, which requires the intact Tat activation domain (21).
The Tat-cyclin T1 interaction is dependent on zinc ions, probably because of the involvement of the cysteine-rich domain of Tat and the cys261 residue of cyclin T1 (39). Because granulin also contains multiple cysteine residues, we examined the zinc dependence of its interactions with Tat and cyclin T1. In pull-down assays, the binding of GrnPGFBA (and GrnCDE, data not shown) to Tat was unaffected by addition of Zn 2ϩ (Fig.  3C). On the other hand, the binding of GrnPGFBA and GrnCDE to cyclin T1 was greatly stimulated by addition of Zn 2ϩ (Fig. 3D). Similar results were obtained with the isolated histidine-rich domain of cyclin T1-(501-530), which contains its granulin binding site (Fig. 3E). Granulin binding to the histidine-rich domain of cyclin T1 was inhibited by EDTA (Fig. 3F). Calcium and magnesium ions also increased the binding of the cyclin T1 to GrnCDE in an EDTA-sensitive manner (data not shown). For CaCl 2 1 mM was enough to stimulate and for MgCl 2 5 mM was necessary, compared with ZnCl 2 , which had the strongest effect at 0.025 mM.
Some Granulin Repeats Interact Differentially with Cyclin T1 and Tat-Because the interactions of Tat with granulin and with cyclin T1 are not identical, although they involve overlapping Tat sequences, we reasoned that individual regions of granulin may bind differentially to the two transcription factors. All full-length granulin repeats contain a conserved symmetrical 12 cysteine-rich motif except for the G repeat, which contains only 10 cysteines (see the introduction). Otherwise, there is considerable sequence variation among the repeats as detailed under "Discussion." To determine the binding specificity of granulin repeats to cyclin T1 and Tat, we constructed two di-repeats (CD and DE), and three single repeats (C, D, and E), and examined their interactions with cyclin T1 and Tat.
First we tested the interactions in the yeast two-hybrid system by monitoring growth on selective medium and ␤-galactosidase activity (Fig. 4). As shown previously, GrnPGFBA and GrnCDE interacted with both cyclin T1 and Tat (18). GrnCD also interacted with both cyclin T1 and Tat. GrnC and GrnD did not interact detectably with either Tat or cyclin T1. Strikingly, GrnDE and GrnE interacted selectively with Tat but not with cyclin T1 compared with the control of vectors alone.
We confirmed the selective interaction of GrnDE and GrnE with Tat but not cyclin T1 in co-immunoprecipitation assays. Human 293T cells were transfected with various FLAG-tagged granulin constructs and anti-FLAG immunoprecipitates were probed for the presence of endogenous cyclin T1. GrnDE and GrnE, as well as GrnD (not shown), failed to co-immunoprecipitate with endogenous cyclin T1 (Fig. 5A). GrnDE and GrnE, but not GrnD, co-immunoprecipitated HA-tagged Tat (Fig. 5B). The binding of GrnCD to cyclin T1 and Tat, observed in the yeast two-hybrid assay, could not be confirmed by co-immunoprecipitation because of background problems, but transactivation data support the existence of these interactions in mammalian cells (see Fig. 7 below). To determine whether the interactions are direct, we tested the binding between Tat and granulins in a GST pull-down assay. The granulins were labeled with 35 S by in vitro synthesis in a wheat germ system. GrnDE and GrnE, but not GrnD, interacted with GST-Tat72 (Fig. 5C).
Finally, to document the specific interactions between GrnDE and Tat in living cells, we examined their intracellular co-localization. Previously we observed that GrnCDE co-localized in the nucleus of transfected COS7 cells with cyclin T1, but GrnDE did not (18). Like GrnCDE, GrnDE was predominantly detected in the cytoplasm in the absence of Tat, but co-localized with Tat in the nucleoli (Fig. 6). Thus, although the major Nand C-terminal segments of the granulin molecule (GrnPGFBA and GrnCDE) both bind cyclin T1 and Tat, differential specificity was observed within GrnCDE. Tat binding is a property shared by GrnCD, GrnDE, and GrnE, but only the GrnCD di-repeat is able to bind cyclin T1. Fig. 2 showed that granulins can inhibit expression from the HIV-1 LTR without binding cyclin T1, implying that the granulin-Tat interaction is also inhibitory. Having found that GrnDE and GrnE cannot directly bind cyclin T1 but can bind Tat, we were able to test this inference by asking whether these granulins are able to repress the HIV-1 promoter.

GrnDE and GrnE Inhibit Tat Transactivation and Viral Gene Expression-The transfection experiments of
Vectors expressing GrnDE and GrnE and other granulin constructs were co-transfected into 3T3 cells with human cyclin T1 and an HIV-1 LTR-luciferase reporter. Firefly luciferase expression was normalized to expression of Renilla luciferase from the RSV promoter. GrnDE and GrnE reduced gene expression by 4-to 6-fold, comparable to the effects of GrnCD and GrnCDE, which interact with both Tat and cyclin T1 (Fig. 7A).
GrnD, which does not interact with either Tat or cyclin T1, did not inhibit to a significant degree. Thus interaction of GrnDE and GrnE with Tat, or perhaps with the Tat⅐P-TEFb complex, is sufficient for inhibition of transcription from the HIV-1 promoter. We note that GrnDE and GrnE, despite their relatively low expression (Figs. 5B and 7A), repress Tat transactivation to an extent comparable to that of the other inhibitory granulins. The granulins PGFBA, CDE, DE, and E, but not GrnD, were also inhibitory toward luciferase expression from a transfected noninfectious HIV-1 molecular clone (Fig. 7B).
We showed previously that full-length granulin and GrnCDE inhibit reporter gene expression by about 2-fold when the Gal4cyclin T1 fusion protein is tethered to a modified HIV-1 promoter via Gal4 binding sites (18). GrnPGFBA, which also interacts with cyclin T1, gave a similar inhibition, but GrnDE did not cause a significant inhibition in this tethered system (Fig. 7C). These Interactions of granulins with cyclin T1 or Tat were analyzed in the yeast two-hybrid system, by co-expressing GAL4 activation domain (AD) fusion proteins with GAL4 binding domain (BD) fusion proteins. Interactions were scored by monitoring growth on selective medium lacking leucine, tryptophan, and histidine (ϩ, growth; Ϫ, no growth), and were quantified using a liquid ␤-galactosidase assay. The interaction between CDK9 and cyclin T1 served as a positive control. Vectors expressing the AD and BD alone were used as a negative control.

FIG. 5. Granulins DE and E interact with Tat but not with cyclin T1.
A, extracts were prepared from 293T cells transfected with equivalent amounts of FLAG-granulin plasmids. Complexes immunoprecipitated with anti-FLAG antibody were immunoblotted using anticyclin T1 antibody (upper panel). The extracts were also analyzed directly by immunoblotting using anti-FLAG antibody (lower panel). B, extracts were prepared from 293T cells co-transfected with equivalent amounts of FLAG-granulin plasmids and 5 g of HA-Tat vector. Immunoprecipitation with anti-FLAG antibody was followed by immunoblotting with anti-HA (Tat) antibody (upper panel). Expression of HA-Tat (middle panel) and FLAG-granulins (lower panel) in WCE was detected by direct immunoblotting with anti-HA antibody and anti-FLAG antibody, respectively. To detect granulins D and E, the right portion of the lower panel was exposed four times longer. C, the binding of 35 S-labeled granulins DE, D, and E to GST-Tat-(1-72) or GST was tested in pull-down assays.

results are consistent with a model in which GrnDE and GrnE inhibit expression from the HIV-1 LTR via interaction with Tat.
GrnDE and GrnE Interact with the Tat⅐P-TEFb Complex-Results presented above (see Fig. 2) suggested that granulin targets the Tat⅐P-TEFb complex, which is the modulator of HIV-1 transcription. Therefore, we sought to determine whether granulin interacts with this complex in cells. The inability of GrnDE and GrnE to bind directly to cyclin T1 (Figs. 4 and 5A) allowed us to determine whether Tat can form a bridge between these proteins by forming ternary complexes containing Tat, P-TEFb, and granulin. To this end, 293T cells were co-transfected with GrnDE (or GrnE) and HA-Tat. GrnDE (or GrnE) immunoprecipitates should contain cyclin T1 if, and only if, P-TEFb is complexed with Tat bound to granulin. Fig.  8A shows that FLAG-tagged GrnDE and GrnE precipitated with endogenous cyclin T1 in the presence of HA-Tat. GrnD, which does not bind to cyclin T1 or Tat, served as a negative control and GrnPGFBA and GrnCDE, which bind both cyclin T1 or Tat, were positive controls. CDK9 was also detected in the immunoprecipitated complex containing GrnDE and HA-Tat (Fig. 8B), confirming that these complexes contain both subunits of P-TEFb. Thus, Tat can bridge between P-TEFb and GrnDE (or GrnE).
This finding was confirmed using a Tat mutant, TatK41A, that fails to interact with cyclin T1 (21, 41) but still binds to granulin (see Fig. 8D). Cells were co-transfected with wild-type or mutant HA-Tat together with FLAG-tagged GrnCDE or GrnDE. Immunoprecipitates with anti-FLAG antibody were probed for the presence of endogenous cyclin T1 and HA-Tat. Fig. 8C shows that the mutant Tat bound both granulins (top panel). Cyclin T1, which co-precipitated with GrnDE in the presence of wild-type Tat (Fig. 8, A and C), was absent from complexes containing GrnDE and Tat K41A (second panel). Thus, when the Tat-cyclin T1 interaction is abrogated by the mutation, ternary complexes containing granulin DE, P-TEFb, and Tat are not observed. Cyclin T1 was also present in com- plexes formed with GrnCDE regardless of the Tat mutation (Fig. 8, A and C), as a result of the direct binding of GrnCDE to cyclin T1 (Figs. 4 and 5A). The existence of ternary complexes FIG. 7. GrnDE and GrnE inhibit Tat transactivation and gene expression from HIV-1 molecular clone. A, NIH 3T3 cells were co-transfected with 1.5 g of vector expressing equivalent amounts of the FLAG-granulins specified, together with 20 ng of RSV-Tat72, 100 ng of FLAG-cyclin T1 (T1), and 100 ng of the reporter plasmids pLTRluciferase (firefly) and pRSV-luciferase (Renilla) as indicated. Cell extracts were assayed for firefly and Renilla luciferase activities. Results are expressed as -fold activation of firefly luciferase activity normalized to Renilla luciferase activity and relative to the control lacking Tat. Data are the average of three experiments in duplicate with standard deviations shown. The inset panel shows expression of granulins PG-FBA, CDE, CD, and DE, detected by immunoblotting. Granulins D and E, which are expressed at low levels in 293T cells (Fig. 5B), were undetectable by immunoblotting of 3T3 cell extracts. B, Jurkat cells were co-transfected with 1 g of HIV-1 molecular clone, 1 g of pRSVluciferase (Renilla), and 10 g or equivalent amounts of the granulin plasmids specified. Firefly luciferase expression from the molecular clone was normalized to Renilla luciferase activity. Data are the average of three experiments in duplicate with standard deviations shown. C, NIH 3T3 cells were co-transfected with 0.25 g of Gal4-cyclin T1, 100 ng of G5-HIV-luciferase (firefly) reporter plasmid, and 1.5 g or equivalent amounts of the specified FLAG-granulins. Results are expressed as -fold activation of firefly luciferase relative to the control lacking Gal4-cyclin T1. Data are normalized to protein levels as some granulins caused a small decrease (ϳ30%) in the expression of Renilla luciferase from the RSV promoter. Data are the average of two experiments in duplicate with standard deviations shown.

FIG. 8. Granulins E and DE bind to the Tat⅐P-TEFb complex in vivo.
A, extracts of 293T cells co-transfected with plasmids expressing HA-Tat and the FLAG-granulins indicated were immunoprecipitated with anti-FLAG antibody. Immunocomplexes were analyzed by immunoblotting using anti-cyclin T1 antibody. Input was 293T whole cell extract (WCE). B, same as in A, except that immunocomplexes were analyzed by immunoblotting using anti-CDK9 antibody. C, human 293T cells were co-transfected with plasmids expressing wild-type (WT) or mutant Tat (HA-Tat WT or HA-Tat K41A) and FLAG-GrnCDE or FLAG-GrnDE as indicated. Complexes containing the FLAG-granulins were immunoprecipitated with anti-FLAG antibody and examined by immunoblotting using anti-HA antibody (upper panel) and anti-cyclin T1 antibody (second panel). The expression of HA-Tat and FLAG-GrnCDE or FLAG-GrnDE in the cells was detected by direct immunoblotting of WCE with the corresponding antibodies (bottom two panels). D, human 293T cells were co-transfected with vectors expressing FLAG-GrnCDE and HA-Tat WT or the HA-Tat mutants C30G and K41A. Complexes containing the HA-Tat proteins were immunoprecipitated with anti-HA antibody and examined by immunoblotting using anti-FLAG (GrnCDE), anti-cyclin T1, and anti-HA (Tat) antibodies as indicated.
containing granulins DE or E, Tat and P-TEFb, is consistent with the ability of these granulins to inhibit HIV-1 transactivation by targeting the Tat⅐P-TEFb complex.
Lastly, we asked whether granulin can bridge between Tat and P-TEFb. Like TatK41A, TatC30G does not significantly interact with cyclin T1 (35), but neither mutation interfered with the binding of Tat to granulin. Cells were co-transfected with constructs expressing FLAG-GrnCDE and HA-tagged Tat wild-type and mutants. Tat-containing complexes were immunoprecipitated with anti-HA antibody and probed with anti-FLAG antibody. GrnCDE coimmunoprecipitated with Tat C30G and TatK41A, indicating that the point mutations allow interaction with granulin (Fig. 8D, top panel, lanes 4 and 5). Cyclin T1 was not detected, however, suggesting that GrnCDE cannot bridge between Tat and cyclin T1 (Fig. 8D, middle  panel, lanes 4 and 5). The amount of GrnCDE complexed with each of the mutant Tat proteins was considerably greater than that with wild-type Tat (Fig. 8D, top panel, lane 3), implying that GrnCDE and cyclin T1 compete for binding to wild-type Tat. This competition could also contribute to the inhibition of Tat transactivation.

DISCUSSION
Granulin interacts with both cyclin T1 (18) and HIV-1 Tat (18,28) and inhibits the P-TEFb-dependent activation of the HIV promoter (18). This inhibition can occur via cyclin T1 in the absence of Tat (18). We show here that granulin binding to Tat can also be sufficient for inhibition of Tat transactivation. Furthermore, we discovered that two granulins, GrnDE and GrnE, bind to Tat but not to cyclin T1, and inhibit Tat transactivation and viral gene expression from the HIV-1 molecular clone. As discussed below, granulin E contains specific amino acids that are unique and conserved.
Specificity of Granulin Interactions with Tat and Cyclin T1-The activation domain of Tat (aa 1-48), which is sufficient for binding to cyclin T1 (21), also binds granulins (18). The interactions of Tat with the two proteins are different, however. First, the acidic domain of Tat (aa 1-21) is necessary for binding to cyclin T1 but not to granulin (Fig. 3B). Second, the cysteine-rich region of Tat is sufficient for binding to granulin but not to cyclin T1 (Fig. 3B). Third, specific point mutations in the cysteine-rich and core domains of Tat that abrogate its interaction with cyclin T1 do not affect its binding to granulins (Fig. 8D). Fourth, Tat interactions with cyclin T1 are zinc-dependent (38,39), whereas zinc has no effect on Tat binding to granulins (Fig. 3C). On the other hand, as predicted (18), zinc greatly enhanced the binding of granulins to full-length cyclin T1 or its His-rich domain (Fig. 3, D and E). These differences suggest that different surfaces of Tat are involved in its interactions with cyclin T1 and granulin, allowing it to bind simultaneously to both cellular proteins. Substitution of three single cysteine residues in the cysteine-rich region of HIV-1 Tat had no discernible effect on granulin binding (Fig. 8D). In contrast, the substitution of each of three cysteines of caprine arthritis encephalitis virus Tat with serine residues results in loss of granulin binding activity (29). The cysteine-rich region of HIV-1 Tat contains 7 cysteines so we speculate that a minimum of three may be required for the Tat-granulin interaction.
Comparison of the interactions between various granulin repeats and cyclin T1 or Tat revealed additional specificities: several granulin repeats interact with both Tat and cyclin T1 (namely GrnCDE, GrnPGFBA, and GrnCD), but GrnE and GrnDE interact only with Tat. GrnC and GrnD did not bind to either Tat or cyclin T1. Because GrnD alone does not bind to Tat, the Tat interaction with GrnDE is probably via GrnE (Figs. 4 and 5). Based on their shared 12-cysteine motif sequence, all granulins are predicted to form the same three-dimensional structure, a novel super-helix consisting of four stacked ␤-hairpins linked by an axial rod containing six intramolecular disulfide bridges (13). The findings that Tat binds to GrnE but not to GrnD or GrnC (this report), or to GrnB or GrnA (28), indicate that, in addition to the general structure of the granulin motif, one or more specific structural features that are present in GrnE are responsible for its interaction with Tat (see below). None of the single granulins tested (C, D, and E) interacted with cyclin T1, suggesting that more than one repeat is required for this interaction. Although the di-repeat GrnCD bound to cyclin T1, GrnDE did not, arguing that the general structure of the di-repeat is not sufficient per se for the interaction with cyclin T1. Presumably, specific residues and structural features to be defined are responsible for the binding differences observed.
Ternary Complex Formation and Inhibition of HIV-1 Gene Expression-The existence of multiple dual interactions among the three proteins, granulin, Tat and cyclin T1, raised the possibility that granulin can form a ternary complex with both Tat and cyclin T1 as well as binary complexes with either cyclin T1 or Tat alone. Taking advantage of the inability of GrnDE and GrnE to bind cyclin T1, we identified a ternary complex in which Tat acts as a bridge between the cyclin T1 and granulin components (Fig. 8).
The co-localization of Tat with GrnCDE and GrnDE (Fig. 6) demonstrates that these proteins interact in cells and that Tat can recruit these granulins to the nucleus and nucleolus. Granulins can inhibit HIV-1 transcription by binding Tat, as evidenced here by experiments using cyclin T1 deletions that cannot bind to granulin (Fig. 2) and experiments using granulins DE and E that cannot bind full-length cyclin T1 (Figs. 4, 5, and 7). In principle, the mechanism of transcription inhibition via Tat binding could involve Tat sequestration and/or the formation of an inhibitory ternary complex containing P-TEFb, Tat, and granulin. Our data suggest that the dominant interaction is between Tat and P-TEFb. In pull-down experiments, P-TEFb binds more avidly to GST-Tat than to GST-granulin (18), and much less granulin is co-immunoprecipitated with Tat when the Tat is also able to bind P-TEFb (Fig. 8). Thus, Tat sequestration is not likely to be the predominant mechanism of suppression.
The transcriptional inhibition extends to expression from the HIV-1 molecular clone and displays specificity in terms of both the granulin effector and the target promoter. GrnD, which does not bind Tat or cyclin T1, had no significant effect on gene expression from the HIV LTR or molecular clone (Fig. 7). Furthermore, in the tethered system, granulins that bind to cyclin T1 (GrnPGFBA and GrnCDE) inhibited cyclin T1-dependent transcription from the LTR whereas GrnDE had little effect (Fig. 7). GrnDE and GrnE are not toxic to cells and had no effect on transcription from the cellular PCNA promoter. 2 Granulin Repeat Comparisons-To discern unique features in GrnE, we aligned the sequences of the human granulins and looked for conserved amino acids (shown shaded in Fig. 9A). In addition to the extended cysteine-rich motif cited above, all the granulins share a conserved threonine residue immediately before the first double cysteine. The alignment reveals differences between granulin E and the other granulins, for example at positions 8 and 10 of GrnE. In granulin E positions 8 and 10 are occupied by phenylalanine and histidine, respectively, whereas the other granulins have serine and proline in these positions. Although the overall sequence identity among E granulins of human, mouse, and rat is 72%, phenylalanine and histidine are fully conserved at positions 8 and 10 (Fig. 9B).
Inspection of the sequence logos in Fig. 9C reveals the existence of additional amino acids that are unique to the E granulins, as emphasized by the large dominant letters in the logo. It is also apparent that the C-terminal region of the E granulins is rich in basic amino acids compared with the other granulins. The non-E granulins share several distinct common features, such as the proline in position 10, although serine in position 8 is not highly conserved.
It is well established that the HIV-1 host range can be determined by single amino acid changes in cellular proteins that interact with viral components (42). A classic example lies in the interactions of HIV-1 Tat and TAR with cyclin T1, which depends on the presence of cysteine in position 261 of human cyclin T1 (38,39). This cysteine, absent from mouse cyclin T1, is essential for the formation of the ternary complex Tat⅐TAR⅐P-TEFb. More recently, a single amino acid in the host defense factor APOBEC3G was found to control its ability to bind HIV-1 vif, thereby affecting viral species specificity among primate viruses (43). Differences among the E granulins of the three mammals are apparent in the sequence alignment shown in Fig. 9B. There are nine unique changes in human granulin E compared with mouse and rat. Of these, three are nonconservative (residues 21, 35, and 37), one is weakly conservative (residue 19), and five are strongly conservative (residues 17, 30, 46, 48, and 57). It is notable that both of the serine residues present in mouse and rat granulin E are changed to non-phosphorylatable residues in human GrnE (asparagine and alanine at positions 19 and 48, respectively) and that full-length granulin can be phosphorylated by P-TEFb in vitro (18). Recent work has identified serine residues as the sole site of granulin phosphorylation in vivo and that human GrnE cannot serve as a substrate for P-TEFb in vitro. 2 It remains to be determined whether Tat interacts with mouse or rat E granulins and whether any of these residues play a role in species specificity. These questions will be explored in future experiments, in conjunction with homology modeling of the E granulins based on the three-dimensional structure of carp granulin-1 and human GrnA (13,44).
The inhibition by GrnE of expression from the molecular clone, which is transcribed like provirus, raises the possibility that specific agents directed against HIV-1 could be developed based on the granulin motif. The granulin structure is very stable, and a synthetic polypeptide corresponding to its 30residue N-terminal subdomain spontaneously forms a structure composed of two stacked ␤-hairpins (45). Thus, granulin E could be exploited as a lead for developing specific therapeutic peptides and to design novel peptidomimetic drugs targeting Tat and the Tat⅐P-TEFb complex. FIG. 9. Comparison of granulin repeat sequences. Human (NP_002078), mouse (NP_032201), and rat (NP_058809) GEP sequences were identified by the HMMER search tool. Motif sequences were aligned by ClustalX using default settings. A, alignments of human granulin repeats that have been tested for Tat binding (only E is positive). Identical residues in the alignment are shaded and marked with asterisks (*); strongly and weakly conserved residues are marked with colons and periods (: and .), respectively. B, alignments of mouse, rat, and human GrnE sequences, displayed in the same format. C, sequence logos for GrnE of human, mouse, and rat (top) and all other granulin motifs (bottom) were created by WebLogo (available at weblogo.berkeley.edu/).