Inhibition of HIV-1 particle assembly by 2',3'-cyclic-nucleotide 3'-phosphodiesterase

CNP inhibits the generation of infectious HIV-1 particles

We have recently described large-scale screens in which individual ISGs were delivered to cells using bicistronic HIV-1-based vectors (Schoggins et al., 2011). During this study we identified ~50 ISG-encoding lentiviral vectors that achieved only poor transduction. Although many instances of poor transduction may have been due to cis-acting vector-intrinsic problems, we noted that some genes with previously described antiviral activity such as APOBEC3G, tetherin and MOV10 (Furtak et al., 2010; Neil et al., 2008; Sheehy et al., 2002) were present in this ISG subset. We therefore hypothesized that other genes from this subset might specifically interfere with infectious HIV-1 production.

To investigate this possibility, we subcloned these ~50 ISGs into a conventional mammalian expression plasmid and measured their ability to inhibit infectious HIV-1 based lentiviral vector production when expressed in trans (Figure 1A). Simultaneously, HIV-1 Gag protein expression in cells, and particulate capsid (CA) protein in culture supernatant was monitored by western blotting (Figure 1B). Control antiviral genes, specifically rhesus macaque tetherin (MmTHN, Figure1A, B) and human tetherin mutated to resist Vpu-antagonism (delGI T45I), (McNatt et al., 2009), reduced infectious particle yield without affecting Gag expression (Figure1 A, B).

Conversely, most of the ~50 screened ISGs had little effect on HIV-1 vector production (Figure 1A). Human tetherin, a well-characterized antiviral restriction factor (Neil et al., 2008) had only a modest effect because the GagPol expression vector used also encodes Vpu, which antagonizes its activity. Of the genes that did reduce infectious virion yield, most decreased the levels of both Gag and coexpressed HA-GFP proteins (e.g. SLFN12, RARRES or OAS3 (Figure 1B)). This was likely a result of global or non-specific effects, induced by expressing these genes.

The screen correctly identified APOBEC3G and MOV10 as antiviral ISGs, both of which reduced infectious virion yield without inhibiting Gag expression or physical particle production, (Furtak et al., 2010; Sheehy et al., 2002). Notably, we also found that one additional ISG, CNP, also substantially inhibited infectious particle production (Figure 1A), This effect of CNP occurred in the absence of any defect in Gag protein expression (Figure 1B). However, in contrast to APOBEC3G and MOV10, and like tetherin, CNP reduced the abundance of particulate CA protein in the supernatant of transfected cells, suggesting that CNP functioned at a stage of the virus replication cycle between Gag expression and virion release.

CNP inhibits replication-competent HIV-1 at physiological expression levels

To determine whether CNP could inhibit replication competent HIV-1 in the presence of the normal repertoire of accessory proteins, we coexpressed an HIV-1 proviral plasmid (NL4-3) with varying amounts of CNP. Infectious particle production was decreased in a CNP-dose dependent manner, with ~100-fold reduction in infectious virion yield at the highest level of CNP expression (Figure 2A). In accordance with findings during the initial screen (Figure 1), CNP expression substantially reduced particulate CA protein in the supernatant of transfected cultures by, up to ~10 to 50-fold (Figure 2B and Figure S1A,B), but did not affect the total level of cell associated Gag protein (Figure 2C). However, Gag processing appeared to be reduced in cells coexpressing CNP, as the presence of CNP led to an apparent ~two-fold reduction in cell-associated CA protein and a corresponding increase in the levels of unprocessed Pr55 Gag precursor in cell lysates (Figure 2C). This apparent reduction in cell associated Gag processing was likely a secondary effect of decreased particle assembly, because CA protein in released virions was processed normally (Figure 2B). Furthermore, no defect in intrinsic particle infectiousness, or envelope incorporation into virions, was observed in the presence of CNP (Figure S1C), again suggesting that CNP affected the quantity and not the quality of the particles that were released.

CNP possesses well-defined catalytic activity (Lee et al., 2001) and also contains a CAAX box prenylation signal at its C-terminus, which is thought to drive membrane localization. Catalytically inactive mutant CNP (H251A, T253A, H330A, T332A) inhibited virion production as effectively as wild-type (wt) CNP, (Figure 2A, B, C). In contrast, introduction of a point mutation (C418A) into the prenylation motif had little effect on CNP expression levels, but abolished antiviral activity (Figure 2A, B, C). These data suggest that CNP prenylation is essential, but catalytic activity is dispensable, for inhibition of HIV-1 production.

CNP is only weakly induced by type I IFNs (Bartee et al., 2009); we therefore assessed CNP expression levels in various tissues to determine whether any expressed sufficient levels to inhibit particle production in vivo (Figure 2D). Note that the two forms of CNP visible in Figure 2C and D and throughout this study are generated from alternate translation initiation codons (O'Neill et al., 1997) (see also Figure 4A). In accordance with the use of CNP as a marker of differentiated oligodendrocytes (Vogel and Thompson, 1988), CNP was abundantly expressed in brain tissue from adult mice (Figure 2D). Notably, expression levels in adult brain tissue exceeded the maximum level of coexpressed CNP used in transfected 293T cells (Figure 2). Thus, CNP expression levels in oligodendrocytes greatly exceeded those required to observe antiviral activity. Interestingly, significant CNP expression was detected in spleen, thymus and lymph node (Figure 2D) raising the possibility that cells expressing inhibitory levels of CNP might also reside in these tissues.

Several, but not all, mammals encode a CNP that can inhibit a subset of lentiviruses

Because retroviral restriction factors often display species-dependent variation in antiviral activity that can be exploited in their characterization, we tested CNP from a panel of mammalian species for their ability to inhibit a panel of retroviruses. Comparable levels of all CNP variants were expressed (Figure 3 A–H) and most of those tested (macaque, African green monkey, cow, sheep) exhibited antiretroviral activity. However, murine CNP was inactive against any of the viruses tested (Figure 3 A–H).

With the exception of murine CNP, the spectrum of viruses targeted by CNP was the same for all the variants tested (Figure 3 A–H). Primate lentiviruses such as SIV_MAC239 (Figure 4B), HIV-2 7312A (Figure 3D) and SIV_AGMSAB (Figure 3E) were all sensitive to CNP to varying degrees. Indeed, SIV_MAC239 seemed particularly sensitive to inhibition by CNP, with infectious particle production being reduced over 100-fold when coexpressed with multiple CNP variants (Figure 3B). Furthermore, the apparent inhibition of HIV-1 Gag processing, evident in the presence of CNP (Figure 2A and Figure 3A), was more striking in the case of SIV_MAC239 (Figure 3B). Indeed, processed SIV_MAC239 CA was barely detectable in the presence of inhibitory variants of CNP.

The specificity of CNP-mediated inhibition was highlighted by the resistance of both the non-primate lentivirus FIV (Figure 3G) and the gammaretrovirus MLV (Figure 3H). Furthermore, CNP was able to distinguish between closely related viruses. For example, SIV_MAC239 and HIV-2 7312A were sensitive to multiple CNP variants (Figure 3D) whereas HIV-2 ROD10 was completely resistant (Figure 3C). Similarly, whereas SIV_AGMSAB was inhibited by CNP, SIV_AGMTAN was unaffected (Figure 3F). Importantly, these data indicate that the antiviral activity of CNP is specific to particular retroviruses, and is therefore unlikely to be due to generalized perturbation of cell physiology.

A single amino acid residue that differs in murine CNP is required for antiviral activity

The observation that murine CNP lacked antiviral activity against the viruses tested led us to investigate which CNP domains were critical for its activity. We used SIV_MAC239 in these experiments because it was particularly sensitive to CNP (Figure 3B). Human CNP is a 422 amino acid protein that shares approximately 85% amino acid identity with murine CNP (Figure 4A). We first constructed chimeric CNP molecules by exchanging amino acids 1–224 of mouse and human CNP (Figure 4B). A chimeric CNP encoding murine N-terminal and human C-terminal domains (MmHs) had no antiviral activity, whereas the reciprocal chimera (HsMm) was fully active (Figure 4C). Thus, determinants of the antiretroviral activity of CNP mapped to the N-terminal half of CNP, which includes the P-loop domain.

Most of the differences between human and murine CNP N-termini were confined to three regions encompassing amino acids 69–75, 89–99 and 109–118 (Figure 4A). Replacing amino acids 69–75 (but not amino acids 89–99 or 109–118) in human CNP with murine counterparts abolished antiretroviral activity, without affecting expression level (Figure 4B, C). Moreover, when individual amino acid substitutions were made in this region, a single conservative substitution (D72E, effectively a three atom (CH₂) insertion into the amino acid side chain) rendered human CNP completely inactive, without affecting expression levels (Figure 4B, C). While the identity of residue 72 was critical for CNP activity, it was not the sole determinant because the corresponding E72D change in murine CNP failed to confer CNP antiviral activity (data not shown).

CNP targets the MA domain of HIV-1 Gag protein following its association with the plasma membrane

The fact that Gag expression is sufficient to drive VLP formation led us to consider whether CNP might directly target Gag. In fact, CNP inhibit the production of VLPs assembled using either native codon or codon-optimized (synGag) Gag expression plasmids, without affecting Gag protein expression levels (Figure 5A). Because synGag expression is achieved in the absence of HIV-1 genomic RNA sequences, this result suggested that CNP acts against HIV-1 Gag directly, with no requirement for any other virally encoded component.

The moderate Gag processing defect that was observed under conditions where CNP reduced particle yield (Figure 2C, 3A, 3B) is reminiscent of the effects of MA mutations that reduce membrane binding (Bryant and Ratner, 1990) suggesting the possibility that CNP might target MA. Indeed, particle formation driven by an HIV-1 Gag construct, Gag-δGH, which lacks the majority of MA (amino acids 7–110) (Perez-Caballero et al., 2004), was resistant to inhibition by CNP. (Figure 5B). Based on these results, we next examined whether transferring the MA region of HIV-1 to MLV, a virus that is ordinarily CNP-resistant, (Figure 3I), might confer sensitivity. Indeed, this manipulation conferred CNP-sensitivity, and CNP reduced the yield of infectious virions assembled using the chimeric Gag by ~50 fold (Figure 5C). Thus, the MA region of Gag is the key determinant of sensitivity to CNP.

The efficiency of Gag multimerization, and thus assembly, is greatly enhanced by Gag association with both the plasma membrane (Kutluay and Bieniasz, 2010) and with viral or cellular RNAs. Because (i) MA directs Gag to membranes and was a key determinant of CNP sensitivity (Figure 4), (ii) CNP is a membrane anchored RNA binding protein and (iii) the CNP membrane anchor (prenylation) was essential for antiviral activity (Figure 2), we investigated whether CNP affected the ability of HIV-1 Gag to associate with membranes, or inhibited the ability of Gag to recruit HIV-1 genomic RNA. Interestingly, under conditions where particle assembly was blocked by CNP, there was little effect on the subcellular distribution of Gag, as assessed by fractionation of Gag-expressing cells on a membrane flotation gradient (Figure 5D). Specifically, Gag was readily detectable in the membrane fraction (fraction 3) of cells coexpressing either WT CNP or inactive non-prenylated CNP (Figure 5D). In contrast, the abundance of the non-prenylated mutant of CNP in membrane fractions was greatly diminished (Figure 5D). These data indicate that even though MA is a key sensitivity determinant, CNP does not inhibit the ability of Gag to associate with membranes, Nonetheless, CNP-membrane association seemed to be required for antiviral activity.

We previously described an assay in which RT-qPCR is coupled with membrane flotation analysis to measure the distribution of HIV-1 RNA in infected cells (Kutluay and Bieniasz, 2010). Using this assay, we found that CNP did not affect the ability of Gag to recruit genomic RNA to membranes (Figure 5E). Overall, these findings indicated that CNP inhibits particle production subsequent to a key early event in virion assembly, namely the association of Gag and viral genomic RNA with the plasma membrane.

To further delineate the step at which CNP acts, we used a microscopy-based approach. Assembly of morphologically accurate, individual HIV-1 VLPs can be readily visualized by fluorescence microscopy using coexpressed Gag and Gag-GFP fusion proteins (Jouvenet et al., 2008; Larson et al., 2005). Normally, Gag-GFP that is initially diffuse concentrates at sites of virion assembly, forming bright fluorescent puncta. In the presence of non-prenylated, inactive CNP, distinctive puncta of Gag-GFP could be readily observed (Figure 5F) and quantified (Figure 5G). Moreover, extracellular VLP production proceeded efficiently (Figure 5H). However, in the presence of CNP, a striking paucity of puncta was observed on the cell surface (Figure 5F). Quantitative analysis revealed that the abundance of VLPs had been reduced ~20-fold by CNP coexpression (Figure 5G). A simultaneous, dramatic reduction in extracellular VLP yield was also observed, while the overall level of Gag/Gag-GFP expression was unaffected by CNP (Figure 5H). Once again, Gag-GFP accumulation at the plasma membrane appeared abundant even in the presence of CNP, however the Gag-GFP was diffusely distributed at the plasma membrane and failed to coalesce into puncta (Figure 5F). These data reinforce the conclusion that CNP inhibits HIV-1 virion assembly subsequent to the association of Gag with the plasma membrane.

Selection of a CNP-resistant HIV-1 mutant

The rapidity of HIV-1 evolution often makes it possible to isolate mutants in vitro that are resistant to inhibitors. We asked whether HIV-1 could evolve resistance to CNP following cultivation in the presence of CNP. Derivatives of MT4 T-cells expressing similar levels of either active CNP or inactive (non-prenylated) CNP were generated (Figure 6A) and then infected with a replication-competent HIV-1 clone in which Nef was replaced with EGFP (NHG, accession: JQ585717). MT4 cells are unusually permissive to HIV-1 replication, and unmodified cells, or cells expressing non-prenylated CNP, became nearly 100% infected between the 3^rd and 4^th day after inoculation (Figure 6B). However, HIV-1 replication was delayed in MT4 cells expressing the intact CNP protein, with the majority of cells remaining uninfected on day 5 (Figure 6B). This result highlights the ability of CNP to inhibit the spread of HIV-1.

HIV-1 propagated in the presence of CNP was harvested on day 6, and following 3 subsequent passages in MT4 cells expressing CNP, an isolate was obtained that replicated with indistinguishable kinetics in the presence or absence of CNP (Figure 6C). Interestingly, this CNP-adapted isolate also replicated with near-identical kinetics to the parental HIV-1 clone in unmodified MT4 cells (Figure 6B, C). Sequencing of PCR-amplified proviral gag DNA identified a single G to A transition that had apparently been selected to near uniformity in the CNP-resistant isolate. This single nonsynonymous change altered codon 40 in the MA region of Gag from glutamate (5’-GAG-3’) to lysine (5’-AAG-3’) (Figure 6D). This change is henceforth referred to as E40K. When this single amino acid change was introduced into the parental HIV-1 clone, replication in primary lymphocytes was impaired only marginally in one donor, and not at all in a second (Figure 6E). Moreover, the E40K mutation did not substantially alter envelope incorporation into virions (Figure S2A, B, C). This suggested that resistance to CNP was acquired without incurring a major replicative fitness cost.

Importantly, the E40K mutation was sufficient to confer complete resistance to the inhibitory effects of CNP on infectious virion production from both 293T cells (Figure 6F) and MT4 cells (Figure S2D, E). E40 is exposed on the surface of the globular head of myristoylated MA (Tang et al., 2004), and is positioned close to, but not precisely coincident with the positively charged surface that is thought to participate in membrane binding (Figure 6G). Moreover, E40 is positioned toward the interior of the trimer structure that is adopted by MA in crystals (Hill et al., 1996). A survey of the variation at E40 in an alignment of primate lentiviruses (http://www.hiv.lanl.gov) revealed that this position is dimorphic in primate lentiviruses and the amino acid identity at this position (Figure 6D) correlated precisely with primate lentivirus sensitivity to CNP (Figure 3). Whenever E40 was present, the virus was sensitive to CNP (HIV-1, SIV_MAC239, SIV_AGMSAB and HIV-2 7312A), and whenever K40 was present, the virus was resistant (HIV-2 ROD10 and SIV_AGMTAN) (Figure 3, 6D). Taken together, these data strongly suggest that HIV-2 ROD10 and SIV_AGMTAN are resistant to CNP because they encode a lysine residue at the equivalent position to HIV-1 E40K. Notably, the vast majority of primate lentiviruses encode glutamate at position 40 and are therefore predicted to be CNP sensitive. Only rare examples of HIV-2, SIV_AGM and SIV_RCM encode K40 and are therefore predicted to be CNP-resistant.

CNP is expressed at low, sub inhibitory levels in many cell types. For example, CNP is barely detectable in 293T cells (Figure 2). Nevertheless, CNP is among the more abundant proteins detected by mass spectroscopic analysis of purified HIV-1 virions derived from 293T cells (Table S1) and is also found in macrophage derived virions (Chertova et al., 2006). Therefore, we examined the degree to which CNP was incorporated into HIV-1 particles when expressed at inhibitory levels, using the active (human) and inactive (murine) CNP proteins as well as sensitive (E40) and resistant (K40) HIV-1 strains. Because overexpressed CNP inhibits the assembly of wt HIV-1, there were fewer particles generated by cells coexpressing human CNP (Figure 6F). However, when equal numbers of virions were analyzed (based on particulate CA quantities, Figure 6H), human CNP was easily detected in wt virions while the inactive murine CNP was largely absent (Figure 6H). In addition, CNP was also largely absent from E40K mutant virions. Importantly, virion associated CNP was resistant to digestion by subtilisin indicating that it is packaged into the interior of virions (Figure 2F). Together, these data suggest that CNP associates with virions, specifically when active CNP and a CNP-sensitive Gag protein are present and is efficiently incorporated into virions if assembly is completed.

CNP binds directly to HIV-1 Gag and blocks an early step in virion assembly

To further examine whether CNP associates with assembling virions, we used fluorescent microscopy to determine whether CNP and Gag-mCherry fusion proteins colocalize in transfected cells. Because antiviral doses of CNP inhibit the formation of Gag puncta (Figure 5F, G) which would complicate analyses of colocalization, we used a low, sub-inhibitory dose of coexpressed CNP. A high degree of colocalization between CNP and Gag-mCherry was observed, with a mean Pearson’s colocalization coefficient of ~0.6 (Figure 7A,C). Notably, CNP exhibited a greatly reduced degree of colocalization with the CNP-resistant E40K Gag mutant, with a mean colocalization coefficient value of ~0.2 (Figure 7B,C). The data suggest that wt, but not CNP-resistant, Gag recruits CNP to sites of particle assembly.

Cumulatively, our findings suggested that CNP might inhibit particle assembly by binding directly to Gag. Importantly, the identification of an inactive point mutant CNP (D72E) and a CNP-resistant Gag point mutant (E40K) provided specificity controls to investigate this question. Our initial attempts to specifically coprecipitate CNP with a Gag-glutathione-S-transferase (GST) fusion protein, coexpressed in 293T cells, proved unsuccessful. However, when cells were treated with a crosslinking reagent (Ethylene glycol bis[sulfosuccinimidylsuccinate], EGS) prior to glutathione-sepharose precipitation, crosslinked CNP:Gag-GST complexes were readily detected in the precipitates (Figure 7D). The crosslinked proteins included one species of a size (~130kDa) that suggested the formation of a complex containing only one CNP molecule (~48kDa) one Gag-GST (82kDa) molecule, as well as higher molecular weight forms that may have contained multiple copies of Gag-GST and/or CNP or other proteins (Figure 7D). The apparent formation of a ~130kDa 1:1 CNP:Gag-GST complex, using an EGS crosslinker that is only ~16 angstroms in length, strongly suggested that CNP and Gag-GST bound directly to each other. Importantly, the formation of the CNP:Gag-GST complexes was remarkably specific. Only trace amounts of CNP(D72E):Gag-GST or CNP:Gag(E40K)-GST complexes were formed. (Figure 7D). Thus, the specificity with which CNP:Gag-GST complexes formed precisely recapitulated the specificity which the antiviral effects of CNP were observed.

The findings that CNP bound to Gag (Figure 7D), and impeded the coalescence of Gag-GFP fluorescent puncta (Figure 5F,G) led us to consider whether CNP might induce the accumulation of incomplete budding structures or mis-formed assembly intermediates that were not detectable by fluorescence microscopy. To examine this possibility, we utilized MT4 cells stably expressing CNP (Figure 6A–C, 7E). In this context, the CNP expression level is comparable to intermediate levels used in cotransfection experiments (Figure 7E) and causes sub-maximal inhibition of particle assembly. We reasoned that this condition would increase the likelihood of observing abnormal virions or assembly intermediates. CNP-expressing cells were infected with equivalent titers of CNP-sensitive (wt) or CNP-resistant (E40K) HIV-1 (both encoding GFP in place of nef) yielding ~65% infected cells (Figure S3A) and examined using thin section electron microscopy (Figure 7F, G, H). The fraction of cells on which virions were observed was reduced ~2-fold in wt HIV-1 infected cells as compared to CNP-resistant (E40K) HIV-1 infected cells (Figure 7G). Moreover, when the number of virions associated with producing cells was counted, this number was about 5-fold reduced for wt HIV-1 infected cells as compared to CNP-resistant (E40K) HIV-1 infected cells (Figure 7H). Thus, the overall observed reduction in the number of virions observed in wt HIV-1 infected cells as compared to CNP-resistant (E40K) HIV-1 infected cells was about ~10-fold, entirely consistent with the reduction in wt HIV-1 particle production observed in the same cells using western blot assays (Figure 7I). Importantly, the morphology of virions and budding structures were indistinguishable regardless of whether CNP-sensitive or resistant viruses were considered. In particular, although wt virions were fewer in number than CNP resistant virions, they were generally of the correct size and morphology (Figure S3B). These data suggest that when CNP successfully inhibits HIV-1 assembly it does so at an early stage, i.e. before electron dense structures (Figure 7F, G H) or fluorescent punta (Figure 5F, G) become detectable. ‘Breakthrough’ particles that do assemble in the presence of CNP apparently achieve normal morphology and infectiousness.

Plasmid construction

The library of ISGs (Figure 1) were inserted into a CMV-promoter driven expression vector (pcDNA DEST40, Invitrogen) using the Gateway procedure, according to manufacturers instructions. CNP cDNAs from the various mammalian species were amplified by PCR and cloned using oligonucleotide derived SfiI sites into Gateway® compatible entry vectors prior to transfer to a CMV promoter driven expression vector (pcDNA™-DEST40, Invitrogen) for transfection. Mutant CNP expression plasmids, and a mutant HIV-1 proviral plasmid (NHG, that carries GFP in place of nef) bearing the E40K MA mutation, were generated using PCR based mutagenesis and oligonucleotides detailed in the supplemental information.

Cell lines and transfection

Cell lines and peripheral blood mononuclear cells were maintained under standard conditions. MT-4 cell clones expressing CNP were generated by transduction with retroviral vectors, puromycin selection and single cell cloning by limiting dilution. Clones were subsequently maintained in medium supplemented with puromycin. All transfection experiments were performed in 293T cells, which were seeded in 24-well plates at a concentration of 2.5×10⁵ cells/well and transfected the following day using polyethylenimine (PolySciences).

Virus release and infectivity assays

For the ISG screen (Figure 1), 293T cells were co-transfected with individual ISG expression vectors and plasmids required to make VSV-G pseudotyped, GFP-expressing HIV-1 vector particles. In addition, a plasmid expressing HA-GFP was also co-transfected.

To measure the effect of CNP on virus particle production, cells were transfected with increasing or fixed amounts of a CNP expression plasmid along with plasmids required to generate infectious retrovirus particles or VLPs.

In all transfection experiments, the total amount of DNA was held constant within the experiment by supplementing the transfection with an empty expression vector. At 48 hrs post transfection, virion-containing culture supernatants were harvested, clarified by low speed centrifugation and filtered (0.22 µm). Particle release was evaluated using western blot analysis of cell lysates and pelleted virions using an anti HIV-1 p24 CA antibody. Infectious virions were quantified by inoculating TZM-bl indicator cells followed by β-galactosidase assay or, in cases where the viral genome encoded GFP, by inoculating MT4 cells and enumerating infected cells by FACS.

HIV-1 adaptation and replication assays

MT4 cells were infected with HIV-1 (NHG) and infection monitored by FACS analysis. Culture supernatants were used to inoculate fresh MT4 cells when the majority of the cells became infected. Replication of WT or E40K mutant HIV-1 in human peripheral blood mononuclear cells was monitored using reverse transcriptase assays.

Western blot analyses

Glutathione-sepharose precipitates, virion and cell lysates, or tissue extracts were separated on 4–12% acrylamide gels, blotted onto nitrocellulose membranes and probed with anti-HIV-1-p24CA (183-H12-5C), anti-CNP (Millipore) or anti-tubulin primary antibodies (Sigma). Subsequently, blots were probed with peroxidase labeled goat anti-mouse antibodies or with fluorescent goat anti-mouse IRdye 800CW antibodies (Li-COR) prior to visualization by exposure to film or quantitation using an Odyssey infrared imaging system (Li-COR).

CNP expression in tissues

Tissues from 16 day and 1-day-old C57BL/6 mice were harvested, and digested where necessary with proteinase K prior to isolating single-cell suspensions using a cell strainer. The tissue samples were then lysed in RIPA buffer (150mM NaCl, 50 mM Tris pH 7.4, 0.1%SDS, 1mM EDTA, 1% Igepal CA-630, 1% Na deoxycholate and protease inhibitors), diluted in protein sample buffer and sonicated. CNP expression was assessed by western blotting as described above.

Membrane flotation assay

Cells expressing HIV-1 Gag, a packageable viral genome and CNP were generated by transfection of HEK 293T cells with a vector genome expressing plasmid (V1B (Kutluay and Bieniasz, 2010)) along with plasmids expressing Gag and CNP. They were then subjected to membrane flotation assay as previously described (Kutluay and Bieniasz, 2010; Spearman et al., 1997) (as detailed in the Supplemental Information). Gradient fractions were probed for the presence of Gag and CNP proteins using western blot assays, as well as viral RNA, using quantitative RT PCR assays as described previously (Kutluay and Bieniasz, 2010).

Fluorescent and electron microscopy

Cells, seeded in glass-bottomed dishes were transfected with HIV-1 Gag, Gag-GFP, Gag-mCherry and CNP expression vectors. They were then immunostained where necessary, using an anti-CNP antibody and observed by deconvolution microscopy, as described previously (Jouvenet et al., 2006). Puncta of Gag-GFP from randomly selected fields were counted and Pearson’s colocalization coefficients derived, as described in the Supplemental Information. Thin section EM of HIV-1 infected MT4 cells was carried out using previously described methods (Zhang et al., 2009), as detailed in the supplemental information. The number of virions per cell was determined by counting the number of virus like structures (virions and budding structures) per cell, using randomly selected cells that were adjudged to have at least one definitive virus-like structure. The percentage of virion producing cells per square of a 200 Mesh EM grid was calculated by counting the number of cells with or without definitive associated virion structures. Approximately 20 cells each from 5 different squares (a total of ~100 cells) were sampled for each condition.

Crosslinking and precipitation analyses

293T cells expressing CNP and Gag-GST were treated with Ethylene glycol bis[succinimidylsuccinate] (EGS) crosslinker, and glutathione-sepharose precipitates prepared as previously described (Eastman et al., 2005; Kutluay and Bieniasz, 2010). The precipitates were analyzed by western blotting, as described above.

Analysis of protein incorporation into virions

To assess the incorportation of Env protein or CNP into virions, virions were harvested from transfected 293T cells and subjected to western blot analysis, as described above, and detailed in the supplementary information. To assess whether CNP was incorporated into the interior of particles, purified virions were treated with subtilisin prior to western blot analysis, as detailed in the supplemental information.

For mass spectrometry analysis of virions, particles were harvested from the supernatant of infected 293T cells by ultracentrifugation onto a sucrose cushion, then separated on an Optiprep gradient. Proteins in each of ten gradient fractions were precipitated with trichloroacetic acid and separated by gel electrophoresis. A fraction containing the majority of the virions was identified and gel pieces comprising the entire lane were analyzed using a using an Orbitrap XL mass spectrometer, as described in the supplemental information.