Basic Residues in the Nucleocapsid Domain of Gag Are Required for Interaction of HIV-1 Gag with ABCE1 (HP68), a Cellular Protein Important for HIV-1 Capsid Assembly*

During human immunodeficiency virus, type 1 (HIV-1) assembly, Gag polypeptides multimerize into immature HIV-1 capsids. The cellular ATP-binding protein ABCE1 (also called HP68 or RNase L inhibitor) appears to be critical for proper assembly of the HIV-1 capsid. In primate cells, ABCE1 associates with Gag polypeptides present in immature capsid assembly intermediates. Here we demonstrate that the NC domain of Gag is critical for interaction with endogenous primate ABCE1, whereas other domains in Gag can be deleted without eliminating the association of Gag with ABCE1. NC contains two Cys-His boxes that form zinc finger motifs and are responsible for encapsidation of HIV-1 genomic RNA. In addition, NC contains basic residues known to play a critical role in nonspecific RNA binding, Gag-Gag interactions, and particle formation. We demonstrate that basic residues in NC are needed for the Gag-ABCE1 interaction, whereas the cysteine and histidine residues in the zinc fingers are dispensable. Constructs that fail to interact with primate ABCE1 or interact poorly also fail to form capsids and are arrested at an early point in the immature capsid assembly pathway. Whereas others have shown that basic residues in NC bind nonspecifically to RNA, which in turn scaffolds or nucleates assembly, our data demonstrate that the same basic residues in NC act either directly or indirectly to recruit a cellular protein that also promotes capsid formation. Thus, in cells, basic residues in NC appear to act by two mechanisms, recruiting both RNA and a cellular ATPase in order to facilitate efficient assembly of HIV-1 capsids.

priate name for this protein, which is the sole member of the ATPbinding cassette protein family E and is present in many species that do not encode RNase L (16). In this study, the term ABCE1 will be used instead of HP68. ABCE1 is highly conserved and ubiquitously present in eukaryotes, including yeast, as well as archaebacteria. In a variety of eukaryotic species, ABCE1 is critical for ribosome biogenesis (5)(6)(7)(8)(9). As is the case with ribosomes, capsids are large multiprotein complexes that contain RNA and are capable of self-assembly in vitro but probably assemble in a regulated fashion within cells (18 -21). Analogous to ribosomes, some viral capsids may utilize ABCE1 to chaperone and coordinate their assembly.
Depletion-reconstitution studies in a cell-free system and dominant negative mutant studies in cells indicate that ABCE1 plays an important role in assembly of immature HIV-1 capsids (4). In addition, co-immunoprecipitation with antibody to endogenous ABCE1 demonstrates that ABCE1 is associated with Gag polypeptides present in assembly intermediates isolated from primate cells expressing HIV-1 or other primate lentiviruses (1,4). ABCE1 releases from Gag upon completion of immature capsid assembly and is therefore not present in significant amounts in HIV-1 particles released from cells (1,4). Analysis of a recently reported crystal structure of ABCE1 suggests that the two nucleotide binding domains in ABCE1 act with a hinge region to undergo a clamp-like motion, with an estimated combined movement of ϳ40°during the ATP-driven power stroke (15). Together these findings suggest a model in which ABCE1 binds post-translationally to HIV-1 Gag in capsid assembly intermediates and promotes ATP-dependent conformational changes important for assembly, thereby functioning as a chaperone during HIV-1 capsid formation. The exact manner in which ABCE1 acts during assembly has yet to be determined but could include conformational changes associated either with formation of the capsid shell, packaging RNA into the capsid, or targeting assembling Gag to membranes.
The 55-kDa HIV-1 Gag polypeptide is composed of four major domains (from N to C terminus): matrix (MA), capsid (CA), nucleocapsid (NC), and p6. Initial analysis of the Gag-ABCE1 association suggested that NC is required for interaction of Gag with ABCE1, whereas p6 is dispensable (1,4). These data are consistent with previous studies demonstrating that NC contains determinants (previously termed the I domain) required for Gag polypeptide multimerization, whereas p6 is not needed for capsid assembly but is required for budding (reviewed in Ref. 22). NC contains two Cys-His boxes that form zinc fingers, are highly conserved among retroviruses, and are known to be important for specific incorporation of HIV genomic RNA (reviewed in Ref. 23). In addition, NC is highly basic; 15 of its 55 amino acids are arginines and lysines dispersed throughout the domain. These basic residues have been shown to bind RNA nonspecifically (24,25). Evidence suggests that RNA binding by these basic residues promotes Gag multimerization, with RNA acting as a nucleator or scaffold (24, 26 -28). However, these basic residues could also act by other mechanisms to promote capsid assembly.
Here we demonstrate that the basic residues within NC are necessary to recruit endogenous ABCE1 into Gag-containing assembly intermediates. Furthermore, we find that the cysteine and histidine residues in NC are not required for the Gag-ABCE1 interaction and that large regions of the MA and CA domains of Gag are also dispensable. Velocity sedimentation analyses demonstrate that two NC mutants that fail to form fully assembled immature capsids (including a mutant that fails to bind ABCE1 and one that binds ABCE1 poorly) are arrested at early points in the assembly pathway in cells. Together, our findings reveal that the basic residues in NC that bind nonspecifically to RNA also are important for the interaction of Gag with the cellular protein ABCE1, which facilitates HIV-1 capsid formation. Thus, NC appears to act by two mechanisms to promote efficient HIV-1 capsid assembly in the complex environment of the cytoplasm.

EXPERIMENTAL PROCEDURES
Plasmids-For mammalian cell transfection plasmids, Gag mutations were engineered into the pSVGagRRE-R construct, which was obtained from David Rekosh (29) and encodes Gag and the Rev response element from the BH10 strain of HIV-1. Truncations were engineered by introduction of two stop codons after the amino acid in Gag indicated in the construct name, using site-directed mutagenesis (Stratagene). To make other constructs, a SacI site was engineered into the parental construct immediately downstream from the Gag coding region by site-directed mutagenesis. Gag mutations were engineered using standard PCR procedures and inserted into the SacI sites on either side of the Gag coding region in the modified pSVGagRRE-R plasmid. The KR10A construct was engineered in an analogous manner by fusing the NC-p6 domains from a template plasmid encoding the KR10A mutations (M1-2/BR (24)), obtained from Jeremy Luban, to the MA-CA domains in pSVGagRRE-R using standard PCR procedures.
Plasmids encoding WT Gag, Tr361, and Tr437 for in vitro transcription have been described previously (1)(2)(3). Plasmids encoding CH1A, CH2A, and CH1/2A for in vitro transcription were engineered using standard PCR procedures into the previously described pSPGag plasmid (WT Gag), which encodes SF2 Gag downstream from the SP6 promoter and a Xenopus 5Ј-untranslated region (1-3, 30, 31). The plasmid encoding KR10A for in vitro transcription was generated by amplifying the M1-2/BR Gag mutant coding region described above (24) and inserting it into the NcoI and EcoRI sites of pSPGag using standard PCR methods. Because M1-2/BR was generated from the HXB2 and BH10 strains of HIV-1 (see Ref. 24 for details), a matched HXB2/BH10 pSP WT Gag plasmid was also generated and found to behave the same as the SF2 Gag plasmid. The FLAG-tagged ABCE1 construct has been described previously (1). The plasmid pcDNA-APO3G, which encodes wild-type human Apobec3G with a myc-HisA tag at the C terminus, was generated by Drs. Klaus Strebel and Sandra Kao (54) and obtained from the NIH AIDS Research and Reference Reagent Program.
To create the GST-ABCE1 plasmid, PCR was used to amplify the previously described ABCE1 coding region (4). This coding region was inserted into the EcoRI and BamHI sites of the pGEX4T-1 vector (Amersham Biosciences). Coding regions of all plasmids generated in this study were confirmed by sequencing.
Transfections and Immunoprecipitations-COS-1 cells (from African green monkey) or human 293T cells were transfected in 60-mm dishes with 4 l of pSVGagRRE-R encoding WT or mutant Gag and 0.5 l of pCMVRev (32, 33) using 24 l of Lipofectamine (Invitrogen) or 15 l of Lipo 2000 (Invitrogen). Cells were harvested 45 h post-transfection in 250 l of Nonidet P-40 buffer (containing 0.625% Nonidet P-40, 10 mM Tris acetate, pH 7.4, 50 mM potassium acetate, and 100 mM NaCl), supplemented with 10 mM EDTA and protease inhibitor mixture for mammalian cells (Sigma). Cell lysates were clarified by centrifugation at 1000 rpm for 10 min in a GH 3.8 rotor using an Allegra 6R centrifuge (Beckman Coulter) and centrifugation in a microcentrifuge at 18,000 ϫ g for 30 s. Clarified lysates were divided equally and subjected to immunoprecipitation using affinity-purified ␣-ABCE1, rabbit IgG, or affinitypurified ␣-Apobec3G, as noted, coupled to protein A immobilized on Tris-acryl beads (Pierce) as described previously (1,4). The antibody to Apobec3G (34) is directed against the C-terminal 29 amino acids in Apobec3G and was affinity-purified against the peptide immunogen and coupled to Tris-acryl beads in parallel with ␣-ABCE1. For two constructs (⌬CA and ⌬MACA), twice as much cell lysate was input into immunoprecipitations to compensate for reduced expression. Immunoprecipitates were analyzed by SDS-PAGE, transferred to nitrocellulose (MSI), and subjected to immunoblotting using a monoclonal antibody to the capsid domain of Gag (Dako) as previously described (1). Aliquots representing 2% of the cell lysate input were analyzed in parallel with immunoprecipitations on all blots.
RNase A Treatment-Clarified lysates of COS-1 or 293T cells, transfected and harvested as above, were treated with RNase A (Qiagen) at the indicated concentrations for 10 min at 37°C. Lysates were subjected to centrifugation at 18,000 ϫ g for 30 s, and equivalent aliquots of the supernatant were used to program immunoprecipitations.
Quantitative RT-PCR-RNA was purified from cell lysates by adding 20 l of lysate to 200 l of RNAqueous lysis buffer (Ambion). After mixing, 20 l of a control lysate (murine EL4 cells; see below) was spiked into the reaction to serve as a control for RNA purification and reverse transcription efficiency. RNA was then isolated per the manufacturer's protocol. Eluates were then treated with rDNase 1 (Ambion) and subsequently subjected to reverse transcription using random DNA primers and Superscript II reverse transcriptase (ϩRT; Invitrogen). Quantitative PCR was performed using iQ SYBR green supermix (Bio-Rad). Serial dilutions of a corresponding DNA template were run in parallel. A standard curve of C T versus Ϫlog[DNA] was calculated, in which the [DNA] of the highest dilution was arbitrarily set at 1, and sample values were extrapolated from the standard curves. Reactions minus RT (ϪRT) were processed in parallel, and ϪRT values were subtracted from the ϩRT values. Values are reported as 0 if the ϪRT value was greater than the ϩRT value. Sequences for the primers are 5Ј-gactatgtagaccggttctat-3Ј (forward) and 5Ј-caaaactcttgccttatggccgggtcctcc-3Ј (reverse) for HIV-1 (Gag), 5Ј-cacggctgcttccagc-3Ј (forward) and 5Ј-ggaaggctggaagagt-3Ј (reverse) for human actin, and 5Ј-cactgccgcatcctct-3Ј (forward) and 5Ј-ggaaggctggccaaga-3Ј (reverse) for murine actin. For EL4 cell lysate, EL4 cells were lysed in lysis buffer (RNaqueous; Ambion) at a concentration of 5 ϫ 10 4 cells/l.
Cell-free Assembly Reactions-In vitro transcription and cell-free translation using wheat germ extract and Tran 35 S-label (ICN Biochemicals) were performed as described previously (1,2). Cell-free reactions were programmed using a mixture of 40% FLAG-ABCE1 transcript and 60% WT or mutant Gag transcripts, as described previously (1). Cellfree translations were diluted 300-fold in Nonidet P-40 buffer and subjected to immunoprecipitation with FLAG antibody coupled to beads (Sigma) or mouse IgG (Sigma) with protein G beads (Pierce), as described previously (1). Immunoprecipitations were analyzed by SDS-PAGE and autoradiography in parallel with aliquots of total cell-free reaction representing 5% of input.
GST-ABCE1 Pull-down Assays-Competent BL21 Escherichia coli (Novagen) were transformed with GST-ABCE1. E. coli were grown to FIGURE 1. The Gag-ABCE1 interaction is progressively reduced upon truncation of Gag in NC. A, diagram showing NC residues in BH10 Gag, with arrows indicating the last amino acid present in each truncation construct. Lines demarcate where the NC domain meets the spacer regions, SP1 and SP2. Basic residues in NC are shown in gray. B, Gag truncation constructs are diagrammed on the right with major domains labeled above and Cys-His boxes (CH1 and CH2) shown in gray. Lysates of COS-1 cells transfected with the indicated constructs were subjected to immunoprecipitation with ␣-ABCE1 (immune, I) or nonimmune control antibody (N) and followed by immunoblotting with an antibody to Gag. Equivalent aliquots of total input (T) are shown to indicate migration and level of expression. Immunoprecipitations were performed under either native conditions (Native) or after denaturation (Denat). For each construct, all lanes shown were taken from a single exposure. Experiments in each panel were repeated three times, and representative data are shown. C, the reduction in ABCE1 interaction relative to wild-type Gag upon progressive truncation in NC was quantitated from three repeats of this experiment (statistical significance; *, p Ͻ 0.05; **, p Ͻ 0.01). Indicated below the bar graph for each construct are the number of lysines and arginines in NC (#KR) and the presence or absence of the first and second Cys-His boxes (CH1/CH2).

FIGURE 2. Altering Lys and Arg residues but not Cys and His residues in NC eliminates the Gag-ABCE1 interaction.
Lysates from COS-1 cells expressing the indicated mutants were subjected to immunoprecipitation (IP) under native conditions with ␣-ABCE1 (immune, I) or nonimmune control antibody (N) and followed by immunoblotting with an antibody to Gag. Constructs examined include mutants containing Cys-His box deletions (A) and amino acid substitutions (B). Equivalent aliquots of total input (T) are shown to indicate migration and level of expression. All lanes were taken from a single exposure. Diagrams in A show amino acids in NC for each construct, with major domains labeled above, lysines and arginines in gray, cysteines and histidines outlined, and substituted amino acids indicated with black dots. Experiments in each panel were repeated three times, and representative data are shown. C, the reduction in ABCE1 interaction relative to wild-type Gag for each construct was quantitated from three repeats of each experiment (statistical significance; *, p Ͻ 0.05; **, p Ͻ 0.01). Indicated below the bar graph for each construct are the number of lysines and arginines in NC (#KR) and presence or absence of first and second Cys-His boxes (CH1/CH2). late log phase, induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside (Sigma) for 2 h, centrifuged at 6000 rpm for 10 min in a SLA-1500 rotor (Sorvall), and resuspended in 8 ml of cold phosphate-buffered saline. Phenylmethylsulfonyl fluoride was added to 100 mM to the resuspended cells, and six pulses (30 s each) with a sonicator were used to lyse the bacterial cells. Bacterial lysate was clarified by centrifugation in an SS-34 rotor (Sorvall) at 12,000 rpm for 10 min, and 1.5 ml of the resulting supernatant was incubated with 140 l of packed glutathione-Sepharose beads (Amersham Biosciences) for 45 min at room temperature. Before incubating with cell lysate, GST-ABCE1-bound glutathione-Sepharose was washed three times with 1 ml of phosphate-buffered saline plus 1% Sarcosyl and then twice with 1 ml of phosphate-buffered saline alone. COS-1 cell lysates were harvested in Nonidet P-40 buffer supplemented with 4 mM magnesium acetate and incubated with the washed GST-ABCE1 glutathione-Sepharose columns for 1 h at 4°C (300 l of lysate/column). After incubation with COS-1 cell lysate, glutathione-Sepharose was washed five times with 1 ml of Nonidet P-40 wash buffer (0.625% Nonidet P-40, 10 mM Tris acetate, 50 mM potassium acetate, 100 mM NaCl, adjusted to pH 8.5) and twice in 1 ml of the same buffer lacking detergent. Three 1-ml elutions were carried out using buffer containing 10 mM Tris acetate, pH 7.4, 50 mM potassium acetate, 100 mM NaCl, and 40 mM glutathione.
Velocity Sedimentation-Calibration of gradients with markers and for calculation of approximate S values has been described previously (2). For analysis of immature released capsids, medium was collected 45 h post-transfection and passed through a 0.22-m filter. Nonidet P-40 was added to a final concentration of 1% to the filtered media, and 100 l was layered onto discontinuous step gradients containing 500 l each of 20, 40, 50, and 75% sucrose in Nonidet P-40 buffer supplemented with 4 mM MgAc and subjected to velocity sedimentation in a TLS55 rotor at 45,000 rpm for 45 min at 4°C (Beckman Coulter). Fractions (200 l) were serially collected from the top of the gradient, precipitated with trichloroacetic acid, and analyzed by immunoblotting as described above.
For analysis of cellular complexes, 100 l of cell lysates were layered onto step gradients containing either 500 l each of 20, 40, 50, and 75% sucrose (for separation of 10/80 S complexes from 500/750 S complexes) or 400 l each of 5, 10, 15, 30, and 40% sucrose (for separation of 10 S complexes from 80 S complexes) in Nonidet P-40 buffer without MgAc. Gradients were subjected to velocity sedimentation under the same conditions as above. Fractions (150 l) were collected serially from the top of the gradient, and equivalent amounts of gradient fractions were analyzed by immunoblotting.
Quantitation-Quantitation was performed by digitizing immunoblots using an Agfa Duoscan T1200 scanner and Photoshop 5.5 software (Adobe Systems Inc.). Mean band densities were determined and adjusted for band size and background. Dilution standards were performed to ensure that immunoblots were in the linear range for semiquantitative analysis.

Progressive Truncation in NC Leads to a Corresponding Reduction in the Gag-ABCE1
Interaction-Previously, we demonstrated that an antiserum directed against endogenous primate ABCE1 (␣-ABCE1) co-immunoprecipitates wild-type HIV-1 Gag from primate cells infected with HIV-1 or transfected with an HIV-1 proviral construct. Abundant cellular proteins such as tubulin and actin are not co-immunoprecipitated by ␣-ABCE1 (4). Since earlier findings suggested that the NC domain of Gag is important for the Gag-ABCE1 interaction in mammalian cells (4), we wanted to identify the features in NC that are required for the Gag-ABCE1 interaction. Therefore, we engineered HIV-1 BH10 Gag mutants that contained progressive truncations in NC (Fig. 1A) into previously described transfection constructs that allow expression of the HIV-1 Gag precursor in the absence of other HIV-1 gene products besides Rev (33). In COS-1 cells, transfection with these constructs results in production and release of virion-like particles containing immature HIV-1 capsids (1,2,4,33). Whereas most truncation mutants expressed at levels comparable with wild-type Gag in COS-1 cells, mutants that were truncated within Cys-His boxes expressed poorly in COS-1 cells and were not examined further (data not shown).
Truncation constructs were expressed in COS-1 cells, and lysates were subjected to immunoprecipitation with ␣-ABCE1 under native conditions. Co-immunoprecipitation of Gag was assessed by immunoblotting (Fig. 1B, left panels, Native). The two longest truncation constructs (Tr437 and Tr427) associated with ABCE1 as well as wild-type Gag. Four constructs of intermediate length (Tr412, Tr410, Tr409, and Tr405) interacted with ABCE1 to a moderate degree but not as well as wild-type Gag. Finally, the two constructs truncated most proximally (Tr388 and Tr361) failed to associate with ABCE1 (Fig. 1B, left panels, Native), even when long exposures were examined (data not shown). Quantitation revealed that these differences in ABCE1 interaction were significant (Fig. 1C). When lysates were denatured to disrupt proteinprotein interactions, ␣-ABCE1 immunoprecipitated endogenous ABCE1 present in COS-1 cells in all cases (data not shown) but failed to co-immunoprecipitate WT Gag or any of the Gag truncation mutants (Fig. 1B, right panels, Denat). This control confirms that native interactions are required for association of HIV-1 Gag with ABCE1, consistent with previous findings (1,4). Together, these data demonstrate that loss of NC residues correlates with a reduction in association of Gag with ABCE1.
Analysis of features in these truncated constructs revealed that the two constructs that interacted with ABCE1 as well as wild-type (Tr427 and Tr437) contained both Cys-His boxes and 14 or 15 basic residues. The four constructs that interacted to an intermediate extent (Tr405, Tr409, Tr410, and Tr412) contained one Cys-His box and 7-11 basic residues. Finally, the two constructs that failed to interact with ABCE1 (Tr361 and Tr388) contained no Cys-His boxes and four or fewer basic residues. The shortest truncation mutant from our series that associated with ABCE1, Tr405, contained one Cys-His box and seven basic residues in NC (see Fig. 1, A and C (bottom)).
Substitution of Lysines and Arginines but Not Cysteines and Histidines in NC Eliminates ABCE1 Association-Since both the zinc fingers and the basic residues in NC were altered upon truncation of NC, the mutations described above failed to distinguish whether only one of these features was critical for the Gag-ABCE1 interaction. Therefore, we engineered additional mutations to determine whether the Cys-His boxes or the dispersed basic residues in NC govern association with ABCE1. First we assessed the effect of deleting the Cys-His boxes. Gag mutants containing deletions of either the first Cys-His box or the second Cys-His box (Gag⌬CH1 versus Gag⌬CH2) associated with ABCE1, as indicated by immunoprecipitation of COS-1 lysates with ␣-ABCE1 ( Fig. 2A). Deletion of both Cys-His boxes (Gag⌬CH1/2) reduced the Gag-ABCE1 interaction to very low but detectable levels ( Fig. 2A).
Each Cys-His box in NC contains not only the cysteine and histidine residues that are critical for zinc finger formation but also a few basic residues. Thus, our deletion of both Cys-His boxes (Gag⌬CH1/2) resulted in the loss of five of the 15 lysines and arginines present in NC (see diagrams in Fig. 2A). Consequently, results obtained using the Gag⌬CH1/2 construct did not allow us to precisely define the contri- FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 bution of the Cys-His boxes versus the basic residues. Therefore, we engineered point mutants that would allow us to completely dissociate the contribution of cysteines and histidines from that of lysines and arginines in NC. In constructs CH1A and CH2A, all of the cysteines and histidines in either the first or second Cys-His box were replaced with alanines, whereas CH1/2A contained the same substitutions in both Cys-His boxes. In all of these constructs, the lysines and arginines present in NC were unaltered. As shown in Fig. 2B, when expressed in COS-1 cells, all three constructs were co-immunoprecipitated by ␣-ABCE1, indicating that the Gag-ABCE1 interaction was maintained even upon substitution of all the cysteines and histidines in NC. Coimmunoprecipitation of CH1A and CH2A by ␣-ABCE1 was similar to wild-type Gag. In contrast, substitution of all of the cysteines and histidines in NC (CH1/2A) resulted in reduced association with endogenous ABCE1, but the association was consistently detectable (Fig. 2C).

Basic Residues in NC Are Critical for Gag-ABCE1 Association
Having established that substitution of cysteine and histidine residues critical for zinc finger formation did not eliminate the Gag-ABCE1 interaction, we next addressed the contribution of basic charge substitutions. Previously, Cimarelli et al. (24) found that mutation of the 10 arginine and lysine residues located in and between the Cys-His boxes resulted in loss of viral replication, virion production, and Gag-Gag interactions. When we substituted the same 10 lysines and arginines in NC with alanines (KR10A), complete elimination of the Gag-ABCE1 interaction was observed, despite expression levels similar to wild-type Gag (Fig. 2B, T lanes). Even very long exposures of immunoblots failed to detect co-immunoprecipitation of KR10A by ␣-ABCE1. These results were confirmed by quantitation (Fig. 2C). Together, these data indicate that the association of HIV-1 Gag with ABCE1 is dependent on basic residues in NC and does not require intact zinc fingers in NC.
To evaluate the number of basic charge substitutions that are required to abolish the interaction of Gag with ABCE1, we engineered Gag constructs containing substitutions of fewer than 10 basic residues. Because initial experiments revealed that alanine substitution of three or four basic residues in the region between the two Cys-His boxes only had small effects on the ABCE1 interaction (data not shown), we engineered a more extreme charge disruption by substituting 3 or 4 lysines and arginines in this linker region of NC with glutamic acids (KR3E and KR4E, respectively). Upon expression of these constructs in COS-1 cells, association with endogenous ABCE1 was observed by co-immunoprecipitation, albeit at reduced levels (Fig. 3A). These data suggest that between 5 and 10 basic residues in NC need to be mutated (leaving 5-10 basic residues intact) to completely eliminate association of fulllength Gag with endogenous ABCE1. The results are similar to findings obtained using truncation mutants in Fig. 1, in which constructs encoding seven or more basic residues were found to associate with ABCE1. Since the exact number of residues required for loss of the Gag-ABCE1 interaction is likely to vary with the exact position of the mutated residues within NC, the choice of substituting amino acid, the length of the Gag construct used, and the exact coding sequence of the Gag strain, we did not attempt finer resolution mapping of basic residues.
Deletion of Large Regions of MA and CA Does Not Eliminate the Gag-ABCE1 Interaction-Others have reported that deletions of large regions of MA and CA do not have significant effects on infectivity, as long as the myristoylation signal and associated charged residues in MA are left intact (35,36). Consistent with this, we found that deletion of a large region of CA (residues 143-276 in Gag; ⌬CA) or deletion of most of MA and CA (residues 12-282 in Gag; ⌬MACA) reduced but did not eliminate the ability of Gag to associate with ABCE1 (Fig. 3B). In addi-tion, deletion of the major homology region (residues 285-304 in Gag; ⌬MHR) within CA had little effect on the Gag-ABCE1 interaction (Fig.  3B). Together, these findings indicate that most of MA and CA are not essential for the Gag-ABCE1 interaction. However, the reduction in association observed with the large deletions raises the possibility that these domains may contain residues that modulate the Gag-ABCE1 interaction.
Confirmation of the Role of Basic Charge in NC for Gag-ABCE1 Interaction Using Other Approaches-The interaction of ABCE1 with Gag was initially identified by biochemical dissection of a cell-free system that supports de novo assembly of capsids closely resembling immature HIV-1 capsids produced in cells (1,2,4,37). Immunodepletion-reconstitution experiments in this system were also used to demonstrate the critical role of ABCE1 in post-translational events in HIV-1 capsid formation (4). To validate findings obtained using ␣-ABCE1 in COS-1 cells, we examined the co-immunoprecipitation of HIV-1 Gag mutants in this cell-free HIV-1 capsid assembly system. FLAG-tagged human ABCE1 was co-translated either with HIV-1 wild-type Gag or with selected Gag mutants in parallel reactions and then subjected to immunoprecipitation under native conditions with antibody to FLAG (␣-FLAG). Consistent with our previous observations, ␣-FLAG co-immunoprecipitated wild-type Gag and Tr437 but not Tr361 (Fig. 4A). In addition, ␣-FLAG co-immunoprecipitated CH1A, CH2A, and CH1/2A to varying extents (Fig. 4A) but did not immunoprecipitate the KR10A mutant (Fig. 4B). Thus, these NC mutations have comparable effects on the Gag-ABCE1 interaction when assessed in different cellular contexts (primate cells versus a cell-free system programmed with wheat germ extract) or using different antibodies (␣-ABCE1 to detect endogenous ABCE1 versus ␣-FLAG to detect epitope-tagged ABCE1).
We also examined the interaction of WT and mutant Gag with a recombinant GST-ABCE1 fusion protein (encoding human ABCE1) produced in E. coli using a pull-down assay. Lysates of COS-1 cells expressing either wild-type or mutant Gag constructs were incubated with GST-ABCE1 purified from E. coli and bound to glutathione beads.  FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 GST-ABCE1 and proteins bound to GST-ABCE1 were eluted from the beads using glutathione, and the presence of Gag and ABCE1 in eluates was assessed by immunoblotting. Following incubation of lysates containing WT Gag, final washes contained no Gag or ABCE1, whereas glutathione eluted both GST-ABCE1 and Gag, indicating that WT Gag was bound to GST-ABCE1 (Fig. 4C). In contrast, after incubation with lysates expressing the negative control Tr361 (Gag truncated proximal to NC; see Fig. 1A), recombinant GST-ABCE1 was detected in eluates, but Tr361 was not (data not shown). KR10A expressed in COS-1 cell lysates also failed to bind to GST-ABCE1, whereas CH1A, CH2A, and CH1/2A from cell lysates bound to intermediate levels (Fig. 4C). Thus, comparable results were obtained when constructs that separately abolish zinc finger motifs versus basic residues in NC were examined by three different approaches, including co-immunoprecipitation of endogenous ABCE1 in COS-1 cells, co-immunoprecipitation of epitope-tagged human ABCE1 expressed in a cell-free assembly system and pull-down from COS-1 cells with recombinant GST-ABCE1. Therefore, we conclude that the basic residues in the NC domain of Gag are critical for association of Gag with ABCE1.

Basic Residues in NC Are Critical for Gag-ABCE1 Association
The Interaction between Gag and ABCE1 Is Resistant to RNase A Treatment-Many studies have shown that the basic residues in the NC domain of Gag are responsible for binding to nonspecific RNA (reviewed in Ref. 23). Association of Gag with other proteins, such as Apobec3G, via an RNA bridge has been demonstrated (e.g. see Refs. 38 and 39). Because ABCE1 may have RNA binding elements (6,15), it is possible that the charged residues in NC allow Gag to associate with ABCE1 by an RNA bridge.
To address this, we examined the effect of RNase A on the Gag-ABCE1 interaction, using the Gag-Apobec3G interaction as a control for RNase sensitivity. Cellular lysates expressing the nearly complete HIV genome (HIV-1⌬env) and Apobec3G were treated in parallel with different concentrations of RNase A and then subjected to immunopre-FIGURE 5. The ABCE1-Gag interaction is insensitive to concentrations of RNase A that degrade HIV-1 RNA and cellular RNA. 293T cells were transfected to express either genomic HIV-1 and Apobec3G constructs or Gag constructs alone, as indicated. A, lysates were harvested, and equivalent aliquots were treated with RNase A (final concentrations of 0 -1000 g/ml, as indicated). Treated lysates were subjected to immunoprecipitation with either ␣-ABCE1 (I), nonimmune control antibody (N), or ␣-Apobec3G (A), as noted. Immunoblotting was performed using an antibody to Gag. Equivalent aliquots of total input (T) are shown to indicate migration of cleaved or uncleaved Gag bands and level of expression. All lanes for each construct were taken from a single exposure. The experiment was repeated three times, and representative data are shown. B, real time quantitative RT-PCR was used to determine the relative amounts of HIV-1 (Gag) and human (Hu) actin RNA for selected lysates shown in A. An aliquot of untreated murine lysate was added to each transfected cell lysate after RNase treatment, and the amount of murine (Mu) actin was also determined as a control for RNA isolation and RT efficiency. cipitation using antibody to endogenous ABCE1, a nonimmune control antiserum, or antibody to Apobec3G (Fig. 5, lanes I, N, and A, respectively). Immunoblotting was used to assess coimmunoprecipitation of Gag by these antibodies. The association of Gag with Apobec3G was lost upon treatment with increasing concentrations of RNase A (Fig. 5A,  top two panels, A lanes), consistent with previous reports (e.g. see Refs. 38 and 39). In contrast, treatment with RNase A had no effect on the association of Gag with ABCE1 (Fig. 5A, top two panels, I lanes). Similar results were obtained in the presence or absence of Vif (HIV-1⌬env⌬vif; Fig. 5A, compare top two panels), which is known to neutralize the activity of Apobec3G (40). We also transfected cells to express wild-type Gag alone, the CH1/2⌬A Gag mutant, or the ⌬CH1/2 Gag mutant. These two mutants were chosen, since both bind to ABCE1 at somewhat reduced levels (see Fig. 2). Even in the absence of other HIV-1 proteins, the Gag-ABCE1 interaction was resistant to RNase A (Fig. 5A,  bottom panels). These data reveal that the Gag-ABCE1 interaction is resistant to RNase A treatment when compared with the RNase A-sensitive Gag-Apobec interaction.
To further confirm that RNase A was effective in these experiments, we subjected an aliquot of selected lysates used for immunoprecipitation in Fig. 5A to reverse transcription followed by quantitative PCR. HIV-1-specific RNAs and actin mRNA were virtually eliminated upon treatment with even 1 g/ml RNase A (Fig. 5B), indicating that RNase A was indeed active. Murine actin, from untreated murine cell lysate that was mixed in after RNase treatment to serve as a positive control for RNA extraction and RT efficiency (see "Experimental Procedures"), remained relatively constant between all samples (Fig. 5B). Thus, the Gag-ABCE1 interaction is maintained even when lysates are treated with concentrations of RNase in excess of that required to reduce cellular and HIV-1 RNA to undetectable levels.
Gag Mutants That Fail to Release Completed Capsids Also Fail to Progress through the Capsid Assembly Pathway-Studies by Cimarelli et al. (24) showed that Gag mutants encoding the KR10A substitutions do not support Gag-Gag interactions. Furthermore, they demonstrated that cells expressing these mutants in the context of the complete HIV genome do not release significant numbers of virions. The few virions released have abnormal cores by electron microscopy (24). To determine whether mutants that exhibited reduced ABCE1 binding were able to release immature capsids, we examined the media from cells expressing wild-type and mutant constructs. Immunoblotting of lysates from transfected COS-1 cells revealed that the NC mutants expressed to similar levels (Fig. 6A). Media from these cells were treated with detergent to remove the envelopes of released viral particles and then subjected to velocity sedimentation as previously described (1,2,4). As expected, wild-type immature capsids were detected in the fractions corresponding to ϳ750 S by immunoblotting with antibody to Gag (Fig.  6B). Since the HIV-1 protease was not expressed, only immature capsids were released. Cells expressing the negative control construct Tr361 did not release immature capsids. Cells expressing the KR10A construct also failed to release immature capsids into the medium as judged by velocity sedimentation, although KR10A contains an intact p6 domain required for budding. Both CH1A and CH2A released ϳ750 S capsids into the medium. In contrast, CH1/2A did not release immature capsids into the medium (Fig. 6B).
Immature capsid release by CH1A and CH2A is consistent with the findings of others showing that mutations in only one Cys-His box have minimal effects on particle release (41). In addition, capsid release by these constructs fits with our observation that CH1A and CH2A interact well with ABCE1 (Fig. 2, B and C). Conversely, the absence of capsid release by CH1/2A and KR10A, which interacted with ABCE1 poorly or not at all (Fig. 2, B and C), suggests defects in intracellular events during immature capsid formation. Previously, we have demonstrated that wild-type Gag and assembly-competent Gag mutants progress through the entire assembly pathway, progressively forming the ϳ10, ϳ80, ϳ150, and ϳ500 S intracellular assembly intermediates before forming completed ϳ750 S capsids (2). In contrast, assembly-defective Gag mutants are arrested at different points along the immature capsid assembly pathway with accumulation of assembly intermediates that precede the point of blockade (1)(2)(3)30). To determine why KR10A and CH1/2A fail to produce virions, we examined intracellular capsid assembly intermediates formed by KR10A and CH1/2. Lysates of COS-1 cells expressing WT and mutant Gag constructs were analyzed using velocity sedimentation gradients in which early assembly intermediates (ϳ10 and ϳ80 S) migrate at the top of the gradient, and late (ϳ500 and ϳ750 S) assembly intermediates migrate at the bottom. Analysis of the assembly-competent WT Gag and CH2A constructs revealed the presence of both early and late assembly intermediates at steady state (Fig.  7A, left panels). Similar results were obtained for CH1A (data not shown). In contrast, the assembly-defective KR10A and CH1/2A constructs formed only early assembly intermediates and accumulated FIGURE 6. Velocity sedimentation analysis reveals that capsids are not released from cells expressing the KR10A and CH1/2. A, COS-1 cell lysates of wild-type and indicated mutant Gag constructs were harvested, and equivalent aliquots were analyzed by immunoblotting to demonstrate level of expression of each construct in cells. B, at the same time, medium from each plate was collected, treated with detergent to remove envelopes from viral particles, and analyzed by velocity sedimentation. Immunoblotting was performed on each fraction to detect Gag. Gradients are shown with fractions from the top of the gradient, representing smaller S values at the left. The position of ϳ750 S immature capsids is shown by a dark bar. The experiment was repeated twice, and representative data are shown. FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 these in large amounts (Fig. 7A, left panels). These data indicate that both KR10A and CH1/2A fail to produce completed capsids because they are arrested at early points in the immature capsid assembly pathway.

Basic Residues in NC Are Critical for Gag-ABCE1 Association
Since KR10A fails to interact with ABCE1, whereas CH1/2A interacts with ABCE1 poorly, it is likely that they exhibit different defects during early events in immature capsid formation. We have previously shown that the ϳ10 S complex does not contain ABCE1, whereas the ϳ80 S complex represents the first assembly intermediate in which ABCE1 binds to Gag (1,4). Moreover, the ϳ80 S assembly intermediate is a critical one, since depletion of ATP causes arrest of Gag in the assembly pathway, with accumulation of the ϳ80 and ϳ150 S assembly intermediates (2). Thus, formation of and/or exit from the ϳ80/150 S assembly intermediates is likely to be ratelimiting and critical for assembly.
To distinguish between early assembly intermediates, cell lysates were analyzed by velocity sedimentation using a gradient that completely separates the first two assembly intermediates (ϳ10 S from ϳ80 S). WT Gag and CH2A, both of which are assembly-competent, contained clear peaks corresponding to both the ϳ10 and ϳ80 S assembly intermediates at steady state (Fig. 7B). In contrast, the KR10A mutant formed only the ϳ10 S complex. The failure of KR10A to form the ϳ80 S assembly intermediate (Fig. 7B) is consistent with its inability to interact with ABCE1 (Fig. 2, B and C). CH1/2A formed a small but reproducible peak in the ϳ80 S region of the gradient (Fig. 7B), superimposed on a trail of the ϳ10 S assembly intermediate. The significantly reduced binding of CH1/2A to ABCE1 (Fig. 2, B and C) with abnormal or inadequate formation of the ϳ80 S assembly intermediate (in which Gag associates with ABCE1) could explain the failure of CH1/2A to progress beyond the ϳ80 S stage of the assembly pathway. Thus, these data suggest that although the cysteines and histidines in the zinc fingers are not required for minimal binding of ABCE1, the presence of at least one intact Cys-His box in NC may be important for reaching threshold levels of ABCE1 binding and consequently for proper ABCE1 function in the intracellular capsid assembly pathway. . Velocity sedimentation analysis reveals that KR10A and CH1/2 are arrested in the form of early assembly intermediates. A, COS-1 cell lysates of wild type and the indicated mutant Gag constructs were analyzed by velocity sedimentation on gradients that separate early from late assembly intermediates. Fractions were immunoblotted, and the amount of Gag in each lane was quantitated and graphed as percentage of total Gag. Positions of early (ϳ10 to ϳ80 S) and late (ϳ500 to ϳ750 S) assembly intermediates are shown. B, the same lysates were analyzed as in A but using a different velocity sedimentation gradient that separates the ϳ10 S assembly intermediate from the ϳ80 S assembly intermediate, as indicated. Positions of early ϳ10 and ϳ80 S assembly intermediates are shown. Experiment was repeated twice, and representative data are shown.

DISCUSSION
Studies presented here demonstrate that basic residues in the NC domain of Gag are critical for the association of the cellular ATP-binding protein ABCE1 (previously known as HP68 or RLI) with HIV-1 Gag. Substitution of 10 of 15 lysines and arginines in NC completely abolished the interaction of Gag with endogenous ABCE1 present in COS-1 cells. In contrast, substitution of all of the cysteines and histidines in NC did not eliminate the Gag-ABCE1 interaction, indicating that the Gag-ABCE1 interaction can occur in the absence of intact zinc fingers. Large regions of MA and CA are also not essential for the Gag-ABCE1 association per se. Taken together with our previous finding that the p6 domain of Gag is dispensable for this interaction (1,4), these data indicate that basic residues in NC are important determinants of the Gag-ABCE1 association in primate cells. More detailed mapping revealed that Gag constructs containing a minimum of 6 -10 lysines and arginines in NC were able to interact with endogenous ABCE1 in primate cells, albeit at reduced levels.
Whereas basic charge in NC appears to be critical for the Gag-ABCE1 interaction, one caveat to note is that other motifs and regions of Gag could modulate this interaction and could also be needed for ABCE1 to function properly during assembly. Notably, the data presented here do not address whether the N terminus of MA plays a role in Gag-ABCE1 binding. Furthermore, the reduction in the Gag-ABCE1 interaction seen with deletions in MA and CA raise the possibility that residues in these regions influence recruitment and function of ABCE1 during capsid formation. Similarly, substitution of as few as four basic residues with glutamic acids (KR4E) and substitution of all the cysteines and histidines (CH1/2A) resulted in significant reduction in ABCE1 binding. In the case of CH1/2A, assembly is arrested at the ϳ80 S assembly intermediate, raising the possibility that reduced ABCE1 binding prevents ABCE1 from functioning properly during capsid assembly. Further investigation of this possibility will require a better understanding of how ABCE1 acts to promote capsid formation.
NC from HIV-1 and other retroviruses is known to interact nonspecifically with RNA as well as in a highly specific manner with genomic RNA (reviewed in Ref. 23). The nonspecific RNA interaction is important for promoting Gag multimerization and capsid formation and appears to be governed by basic charge in NC (24,25,(42)(43)(44)(45). In contrast, packaging of the genome of HIV-1 and many other retroviruses requires intact zinc fingers within NC (46 -51). These and other findings support a widely accepted model in which association of basic residues with RNA promotes multimerization of the ϳ5000 Gag polypeptides that are needed to form a single immature capsid, whereas the Cys-His boxes govern encapsidation of genomic RNA (24,26,27,52,53). However, the finding that RNA plays a critical role in assembly does not exclude the possibility that other factors may also play an important role in promoting proper assembly. Recruitment of proteins that promote capsid assembly could be important in cells, where the concentration of Gag is typically quite low, barriers to efficient assembly are likely to exist, and efficiency of virion production is critical for propagation.
Although it remains to be determined whether primate ABCE1 has RNA-binding properties, our data suggest that the Gag-ABCE1 interaction is not dependent on an RNA bridge. We have previously found that the Gag-ABCE1 interaction is relatively resistant to 1 g/ml RNase A (4). Here we demonstrate that the interaction is resistant at to RNase A at 1000 g/ml, which greatly exceeds the concentration of RNase A required to fully degrade cellular and HIV-1 RNA and disrupt another known RNase-sensitive interaction in the same extract. One explanation for these findings is that the Gag NC domain may be associated with ABCE1 largely through protein-protein interactions. An alternate possibility is that RNA binding by NC may promote Gag-Gag interactions that in turn alter the conformation of Gag, thereby exposing a binding site for ABCE1 elsewhere in Gag. In this model, after ABCE1 is bound to Gag, RNA is no longer required to maintain the altered conformation that exposes the ABCE1 binding site. Further investigation will be required to distinguish between these models in which the basic charge in NC acts either directly or indirectly to promote the Gag-ABCE1 interaction. Note that other primate lentiviral Gag proteins interact with ABCE1 during assembly (1). Studies suggest that HIV-2 and SIV Gag proteins also utilize basic residues to recruit endogenous ABCE1 in primate cells, although the exact residues involved have not been mapped (1,30).
In summary, our data suggest the following model. Whereas a wide variety of unrelated viruses bind nonspecifically to RNA to promote capsid formation, HIV-1 and other primate lentiviruses appear to have evolved a mechanism in which the same RNA-binding residues also act directly or indirectly to recruit ABCE1, a cellular protein that further facilitates the capsid assembly process. By promoting efficient capsid formation in cells, ABCE1 may act as a molecular chaperone in concert with RNA to ensure Gag multimerization under circumstances where assembly is not favored. Since ABCE1 is involved in promoting ribosome assembly (5)(6)(7)(8)(9), it would be present in an ideal location to associate with and act on newly synthesized Gag polypeptides.