HIV-1 NEF BINDS A SUBPOPULATION OF MHC-I THROUGHOUT ITS TRAFFICKING ITINERARY AND DOWNREGULATES MHC-I BY PERTURBING BOTH ANTEROGRADE AND RETROGRADE TRAFFICKING

The HIV protein, Nef is thought to mediate immune evasion and promote viral persistence in part by downregulating major histocompatibility complex class I protein (MHC-I or HLA-I) from the cell surface. Two different models have been proposed to explain this phenomenon: 1) stimulation of MHC-I retrograde trafficking from and aberrant recycling to plasma membrane versus 2) inhibition of anterograde trafficking of newly synthesized HLA-I from the ER to the plasma membrane. We show here that Nef simultaneously uses both mechanisms to downregulate HLA- I in PBMCs or HeLa cells. Consistent with this, we found using Fluorescence Correlation Spectroscopy that a third of diffusing HLA-I at the ER, GOLGI/TGN and the plasma membrane (PM) was associated with Nef. The binding of Nef was similarly avid for native HLA-I and recombinant HLA-I

The HIV protein, Nef is thought to mediate immune evasion and promote viral persistence in part by downregulating major histocompatibility complex class I protein (MHC-I or HLA-I) from the cell surface. Two different models have been proposed to explain this phenomenon: 1) stimulation of MHC-I retrograde trafficking from and aberrant recycling to plasma membrane versus 2) inhibition of anterograde trafficking of newly synthesized HLA-I from the ER to the plasma membrane. We show here that Nef simultaneously uses both mechanisms to downregulate HLA-I in PBMCs or HeLa cells. Consistent with this, we found using Fluorescence Correlation Spectroscopy that a third of diffusing HLA-I at the ER, GOLGI/TGN and the plasma membrane (PM) was associated with Nef. The binding of Nef was similarly avid for native HLA-I and recombinant HLA-I A2 at the PM. Nef binding to HLA-I at the PM was sensitive to specific inhibition of endocytosis. It was also attenuated by cyclodextrin disruption of PM lipid micro-domain architecture, a change, which also retarded lateral diffusion and induced large clusters of HLA-I. In all, our data support a model for Nef downregulation of HLA-I that involves both major trafficking itineraries and persistent protein-protein interactions throughout the cell.
Viral persistence is a fundamental feature of HIV infection; it is due to an ineffective immune response induced directly by the virus. Multiple mechanisms have been proposed to explain this, including depletion of and immune evasion by HIV-infected CD4+ T cells. The HIV-encoded membrane protein Nef is thought to be an important mediator of immune evasion (1). The underlying mechanisms of Nef immunomodulation have not been fully delineated, but appear to involve Nef-induced downregulation of major immune cell receptors from the plasma membrane, including CD4, MHC I and II, CD28 and DC-SIGN (2)(3)(4)(5). Nef-dependent downregulation of MHC-I may be particularly important since this may impede clearance of HIV-infected cells by MHC-restricted CD8 + cytotoxic T lymphocytes (1,6) (and cited reviews). Animal viruses have evolved different strategies to subvert nearly every step in the assembly, maturation, trafficking and antigen presentation by MHC-I (7,8). The study of how Nef perturbs trafficking of MHC-I and other immune receptors is important since it may provide new insights into both basic mechanisms regulating HIV persistence and general mechanisms of antigen presentation, as well as offer new strategies for vaccine development. Knowledge of Nef mechanisms may help guide therapeutic schemes based on the decryption of persistently infected clones Genetic and biochemical studies have shown that at least three distinct sub-domains of Nef are required for downregulating HLA-I: an N-terminal α helix with a conserved MET at position 20 (M20), an SH3-binding polyproline motif 72 PXXP 75 , and an acidic domain, 62 EEEE 65 (9)(10)(11)(12). However, a consensus has not yet been established at either the molecular or cellular level for how Nef downregulates HLA-1. Data supporting defective retrograde receptor trafficking, also known as recycling, by internalization of receptor from the plasma membrane and/or return of internalized receptors to the plasma membrane, have been published (3,10,13). In contrast, there is evidence for defective anterograde trafficking, in which newly synthesized MHC-1 complex is misrouted in an AP-1 dependent manner from the TGN to endolysosomes for premature degradation (14)(15)(16). The two models envision different roles for the three functional motifs of Nef with regard to HLA-I downregulation. In the retrograde model, the three motifs of Nef are required in a hierarchical manner 1) for binding to PACS-2 in the TGN through the 62 EEEE 65 motif; 2) for binding and activation of a TGNlocalized Src family kinase through the 72 PXXP 75 motif; and 3) for inhibiting HLA-I recycling through the M20 motif (10). In contrast, the anterograde mechanism postulates that the same three Nef motifs are required for facilitating and stabilizing a ternary complex between Nef, the cytoplasmic tail of HLA-I and AP-1 adapter (17, 18). Importantly, the proponents of this model affirm that Nef does not bind to MHC-I at the cell surface (14) and that Nef mutants that lack an effect on MHC-I do not bind the receptor (19). A significant limitation in studies supporting both viewpoints is that the binding analyses did not involve live cell conditions to establish subcellular distribution, but rather steady state interactions in cell lysates. These models are not mutually exclusive; and they have not been evaluated simultaneously in the same cell systems. Aberrant MHC-I trafficking as proposed by each model may have a different immunological outcome. If Nef were to exclusively disrupt the anterograde transport of nascent MHC-I, no HIV-I antigens will be presented for developing a CTL repertoire. If, however, the defect lies in the retrograde transport, the reduced levels of HIV antigen loaded MHC-I at the cell surface may compromise CTL surveillance and killing of infected cells. In this work, we have addressed the gaps in the knowledge on how Nef may impact MHC-I traffic through a combined biochemical, biophysical and cell biological study of Nef influence on native and recombinant HLA-I trafficking in human PBMCs, the human T cell line Jurkat and the epithelial cell line HeLa.

Two Photon Two Color Fluorescence Cross Correlation Spectroscopy (TPTCFCCS).
For TPTCFCCS experiments, low-level expression was achieved by using 1/4 th the regular amount of plasmid DNAs and limiting expression to 4-6 h. For visualizing ER, Golgi or the plasma membrane, transfectants were stained at 25 o C for 30min with ER-Tracker Red, BODIPY TRceramide or wheat germ agglutinin conjugates (Invitrogen Corp) and used within the next 30-45 min. Two-photon imaging and TPTCFCCS measurements were carried out using a modified Alba II system (ISS Inc.) in the Ultrafast Laser Microscopy facility of NHLBI. The excitation source was a 100 fs pulsed tunable titanium sapphire laser (MaiTai, Spectra-Physics-Newport). The wavelength was set to 920 nm, for the HLA-I A2-Venus/eYFP (HLA-I A2-V) and Nef-Cerulean (Nef-CerFP) experiments. The excitation power was set to <6 mW at the microscope entrance. The microscope was a Zeiss Axiovert 135M using an E750SP 2P dichroic filter (Chroma Inc) to eliminate the IR exciting light. The objective was a 100X Plan-Neofluar oil objective (Zeiss Inc) with NA 1.3.
A piezo scanner (100µm X 100µm travel, Mad City Labs) and a z-scanning PIFOC (100µm, Mad City Labs) controlled by the ISS software was used to image Hela cells at a resolution of 0.2 ms or 1.0 ms per pixel. For the HLA-I A2-Venus and Nef-CerFP experiments, a 515 nm dichroic mirror was used to split the detected light onto 2-channels. Additional 580±30 nm bandpass emission filters were placed before channel 1 (eYFP/Venus channel) to minimize the contribution of the Cerulean signal. For Cerulean detection, a 470±40 nm filter was used in Channel 2 (Nef-CerFP channel). For excitation of the organelle specific dyes at 850nm, a 630 LP filter was used in Channel 1 instead of the 580 nm bandpass used for eYFP detection. Confocal volume was calibrated by measuring the diffusion of rhodamine 110 and Alexa 488 after excitation at 920 nm. The beam waist, w o , was found to be 0.3 µm, while the axial waist z o was 1.4 µm. A diffusion coefficient for cerulean of 14±4 µm 2 /sec was obtained in the cytoplasm at room temperature. Using these filters and 920 nm excitation (favors eYFP and Venus over Cerulean and eCFP), there was less than 5% bleed through with little to no cross correlation in the eYFP/Venus channel arising from eCFP/Cerulean. The diffusion of Nef-CerFP was found to be 8.7±2.8 µm 2 /sec in the cytoplasm. Other controls included measuring the diffusion coefficient of cells cotransfected with Cerulean and Venus tagged LL/AA CD4 mutant (0.90±0.7 µm 2 /sec). This value can be compared to diffusion coefficient of Venus tagged CD4LA alone, 1.05±0.37 µm 2 /sec. Other controls included: Nef-CerFP by itself at the PM, 3.4±2.1 µm 2 /sec. Also, we measured a two-component diffusion of HA2Y at the plasma membrane, 6.0±0.8 µm 2 /sec and 0.21 ±0.06 µm 2 /sec. For the MβCD and ikarugamycin experiments, a 570 LP filter was used instead of the 560±30 nm bandpass filter to capture as much as possible of the VenusFP signal. For the Alexa 647 labeled antibodies for HLA experiments, a 660 dichroic (Semrock Inc) was used to separate the cerulean and Alexa 647 channels. The excitation wavelength was 850 nm for experiments involving Alexa 647 labeled antibodies against HLA-I-ABC (HA) or HLA-I-A2 (HA2) and Nef-CerFP. A 692±40 nm filter (Semrock Inc) was used in Ch1 to detect the Alexa 647 signal in this case. The same filter mentioned above was used for cerulean detection in these experiments. The calibration at 850nm yielded a focal volume with waists w 0 = 0.25 µm and z 0 =1.6 µm. The molecular weight of the monomeric antibody is about 150kD with a diffusion coefficient in water of 56±16 µm 2 /sec. Data were acquired at a sampling rate of 100kHz with the ISS Alba II system. The cells were imaged at 0.2 or 1.0 ms/pixel and up to three locations per cell and data from at least 5 or 6 cells from each of 3-6 independent transfections were selected for TPTCFCCS acquisition. Each point was observed sequentially five times lasting 15-20 secs. The cells were analyzed 4-5 h after transfection, when protein levels are low and only weak FP images are possible. The experiments were repeated 5-6 times on different days. The correlation function is calculated from the intensity trace F(t) by (23) € G τ   performed at 514nm by 1 scans at 0.078 sec at full laser power. 10 scans at 0.078 sec intervals were acquired before bleaching and post-bleaching image acquisition was set for 200 scans at 0.078 sec interval. Two regions of interest (ROI) were set in the ER, Golgi or plasma membrane of each cell, one ROI for image acquisition post-photobleaching, while the other ROI was set as non-bleached control. In each cell, fluorescence recovery data collected in bleached ROI were normalized against values collected in non-bleached ROI and averaged for values form 5 -10 cells corresponding to 3-5 independent experiments. The data was also normalized to the loss of signal due to the bleach pulse. The Origin Pro software (v. 7, OriginLabs Inc) was used to generate the fitted curve and t-half (t 1/2 ) from the normalized fluorescence recoveries. FRAP data for HLA-I A2-Venus recovery were fit to a two component exponential decay given by: y = A1*exp(-x/t1) + A2*exp(-x/t2) + y0, where t1 and t2 are the diffusion times, A1 and A2 are the pre-exponential factors and y0 isF the immobile fraction. The diffusion coefficients were calculated according to Axelrod et al (25) given by τ D = ω 2 γ/4D (26). Statistical analysis. Biochemical and immunological experiments were done in duplicate. Quantitative microscopic experiments were done in duplicate three times or more as indicated under the appropriate sections. Results are expressed as mean±SEM. Significance levels of differences in means were determined by unpaired two-tailed Student's t-test using Prism 5 or Microsoft EXCEL.

Nef downregulates MHC-I with variable efficacy in different cell types.
Recombinant Nef was able to downregulate native MHC-I and recombinant HLA-I A2 and other HLA-I alleles in epithelial and hematopoietic cells, but with variable efficacy. In particular, Nef was less effective in HeLa cells than in the T cell line Jurkat or in PBMCs. Differences in efficacy for downregulation of native vs. recombinant HLA-I in T cells and epithelial cells were observed among many different Nef alleles derived from clinical isolates of HIV-1 and SIV. The variability appears to be due to cellular factors, since in all cases all Nef alleles tested had the same efficacy in any given cell type (Supplemental Results, Suppl. Table I and Suppl. Figures 1-2). HIV-1 Nef downregulation of native HLA-I relies on multiple molecular mediators of endocytosis. We evaluated the relative contribution of endocytosis in Nef-induced HLA-I downregulation in PBMCs using two types of genetic inhibitors of endocytic trafficking: 1) GFP-or YFP-tagged dominant-negative inhibitor forms of vesicular GTPases (GDP bound) or constitutively active mutants (GTP bound) of the GTPases dynamin, Eps15, Rab5, Rab11, Rab7, and Arf6. In quiescent PBMCs, surface expression of HLA-I was reduced to 50±3% of control by Nef in the absence of these endocytic inhibitors, but only to 76±4%, 75±3%, 86±6%, 81±5%, 76±4% and 77±4% of control in the presence of the inhibitors, Eps15 deletion mutant; K44A dynamin mutant, Rab5-S34N, Rab7-T22N, Rab11-S25N and Arf6-T27N mutants, respectively (Figure 1 A). To further define the biochemical mechanisms regulating the Nef effect we used HeLa cells, a frequently used cell system for isolating Nef mechanisms. It is important to recall, however, that HeLa cells are epithelial cells and therefore not a target of HIV-1. We first inquired whether Nef effect on HLA-I is enhanced at lower temperature where one may expect reduced intracellular HLA-I transport rate (14,27). As shown in Figure 1 B, Nef induced only ~20% loss of HLA-I at 37 o C, versus significant (2-3 fold) response (to 50± 6 or 39±12% of control by NL4-3 or NA7 allele) upon 12h downshift to 26 o C (Figure 1 B). Downshifting to 26 o C did not, however, alter the Nef effect either in T cell lines or in quiescent PBMCs (not shown). We measured the density of native HLA-I at the PM of HeLa cells co-expressing Nef or a null mutant (NX) with GFP or GFP/YFP fusion proteins of various vesicular GTPases. At 26 o C, Nef down-regulated HLA-I to 48±4% of control, NX. In a pairwise comparison of HLA-I MFVs in the Nef (+) versus (-) population from the same transfection, or separate wt vs NX transfections, the dominant-negative inhibitors of Arf6 (T27N) and the constitutively active mutants of Arf6 (Q67L) and Rab11 induced significant reversal of NA7 and NL4-3 Nef effects from 48±4 to 110±8 or 91±8% for Arf6T27N; to 84±8% or 69±6% for Arf6Q67L; and 111±10 or 87±7% for Rab11GTP (Figure 1 C). The above results were corroborated by immunofluorescence microscopy.
Steady-state distribution of HLA-I was examined for Nef and GFP or the GFP/YFP fusion proteins described above. In the absence of functional Nef expression, HLA-I normally resided at the PM whereas in the Nefexpressing GFP (+) cells substantial HLA-I was present in perinuclear vesicles with the remaining receptor distributed in a speckled pattern near the PM (Figure 1 D, compare panels labeled NX & GFP and Nef & GFP). In cells expressing GFP or YFP-tagged K44A dynamin, Eps15 deletion, and Rab5-GDP mutant, HLA-I trapping in the perinuclear vesicles was markedly reduced with most of the receptor now redistributed to more peripheral punctate vesicles, resembling clathrin coated vesicles (CCVs) at the plasma membrane. Though Arf6 G-proteins substantially reversed Nef effect at the PM (Figure 1 C, right), HLA-I was still found in some large vesicular and tubular structures (Figure 1 D) that have been termed Arf6 endosomes (28).
Nef induced clearance of recombinant HLA-I A2 from the PM was partially reversed by some genetic inhibitors of endocytosis. We inquired whether the Nef effect on recombinant HLA-I A2 would mimic that on native HLA-I. In HeLa cells, NA7 and NL4-3 Nefs were more effective on the A2 allele than native, reducing the MFV of A2 to 33±6 or 34±5% at 37 o C and 22±4 or 14±3% at 26 o C (Figure 2 A, top left). Further, genetic inhibitors of endocytosis were not as effective in reversing the Nef effect on A2. At 26 o C, there was a modest reversal of both the NA7 and NL4-3 Nef effects: from 41±3 and 39±3% to 72±8 and 64±6% for ARF6T27N; to 73±7 and 68±6% for ARF6Q67L and to 71±7 and 66±6% for Rab11-Q67L (Figure 2 A, top and bottom right). Minor reversals of Nef effect were also observed with other Rab5, Rab7 and Rab11 mutants. At 37 o C, only the dominant negative Rab11-S25N induced a more modest reversal of the NL4-3 and NA7 effects, to 54±6 or 59±5%, respectively (Figure 2 A, bottom left). The genetic inhibitors were ineffecive in Jurkat cells (data not shown). The stronger Nef effect on recombinant HLA-I A2 vs. innate HLA-I was supported by immunofluorescence microscopy (which also added important spatial information). Absent Nef, almost all the HLA-I A2 was observed at the PM; in the presence of Nef, HLA-I A2 was trapped into perinuclear vesicles ( There was more extensive co-localization of HLA-I A2 with CD63 and LAMP-1 markers for late endosomes and lysosomes. There was also modest colocalization with mannose-6-phosphate receptor (M6P-R), which transports the TGN resident proteins from the Golgi complex. HLA-I A2 was not colocalized with TGN46. Given rapid decay of CD4 in Nef (+) cells, we inquired whether CD4 would appear in the same transport vesicles. In Nef-expressing cells, CD4 was redistributed in a pattern similar to that of HLA-I A2.
There was considerable co-localization of CD4 with CD63, M6P-R and TGN46 positive vesicles, and less co-staining with LAMP-1 positive vesicles and some colocalization with early endosomal antigen (Suppl. Figure 3). Both HLA-I A2 and CD4 colocalized with β-COP vesicles, more so for CD4. It appears that notwithstanding the modus operandi, Nef causes both the recombinant HLA-I A2 and CD4 to be sequestered in post Golgi and endolysosomal vesicles. This may presage a β-COP dependent pathway as proposed (16). Nef down-regulation of native HLA-I in PBMCs and T cell lines was reversed by siRNA knockdown of AP1 subunits or clathrin heavy chain. Nef has been proposed to induce TGN retention and eventual lysosomal degradation of HLA-I by recruiting and stabilizing AP-1:HLA-I complexes (15,16). Subsequently, it was shown by siRNA knockdown that clathrin was also crucial to HLA-I diversion (27). Treatment of Jurkat cells with siRNAs against the µ2 or α (not shown) chain of AP2, µ1 or γ (not shown) chain of AP1, δ chain of AP3 or clathrin heavy chain (CHC) induced a marked depletion of AP2, AP1, AP3 and clathrin positive vesicles respectively as shown by immunofluorescence (Figure 3 A, top left). Knockdown of the AP1 µ1 or γ chains substantially reversed the Nef effect from 31 ±2% or 38 ±2% to 86 ±6% or 88 ±8% (n=4, p<0.01) of control levels for Jurkat cells and quiescent PBMCs respectively ( Figure 3 A and B, histograms on the right, data for γ chain not shown). Clathrin knockdown fully restored the PM HLA-I levels. Ablation of AP2 µ or α (not shown) chain resulted in a mild, but statistically significant enhancement of Nef induced HLA-I downregulation; from 31 ±2% or 38 ±2% to 26 ±2% or 32 ±3% in Jurkat cells and PBMCs respectively (n=5, p<0.05). AP3 δ chain knockdown also led to slight yet significant reversal of Nef effect (from 31 ±2% or 38 ±2% to 40 ±3% or 55 ±4% in Jurkat cells and PBMCs; n=5, p<0.05) ( Figure 3 A and B, histograms on the right). Differential effects of siRNA knockdown of PACS-1, ARF6 and ARNO on Nef effects on HLA-I. First we confirmed that Nef clearing of the recombinant HLA-I A2 was also dependent on the same adapter subunits and clathrin as native HLA-I (Suppl. Results and Suppl. Figure 4). Nef effect differs at 26 o vs. 37 o C in HeLa cells vs. T cells lines; to see if this reflected different intracellular itineraries, we undertook a more extensive siRNA knockdown analysis of Nef effect on HLA-I and CD4. Following siRNA knockdown, we cotransfected Nef or the inactive NX with CD8 and allowed the cells to grow for 18-24 h at 37 o and 26 o C after plasmid transfection. We quantified PM density of native or recombinant HLA-I and CD4 in CD8 (either NA7 or NL4-3 allele) gated populations. Cell aliquots were saved for immunoblot quantitation of the respective siRNA target. As shown in Figure 4, panels B1-B4 and C1-C3, the various siRNAs knocked down the cognate proteins to almost undetectable levels. Nef induced a slight loss (to 58±5% of control MFV) of native HLA-I in HeLa cells at 37 o C. There was a significant improvement in the Nef effect at 26 o C, with the PM HLA-I reduced to 38±3% of control. AP1 µ1 (Figure 4 A, top left) or γ (not shown) chain knockdown led to a significant reversal of Nef mediated HLA-I downregulation in HeLa cells to MFVs approaching 84±6.8 or 79±5.8% at 37 or 26 o C respectively. CHC knockdown also led to restoration of HLA-I in Nef expressers to 106±9 or 94±8% at 37 or 26 o C, as was also shown for Jurkat cells and PBMCs ( Figure 3) at 37 o C. siRNA knockdown of AP2 µ chain enhanced HLA-I downregulation slightly; from 58±5 or 38±3 to 38±2 or 31±3% at 37 and 26 o C respectively (Figure 4 A, top left). Similar results were obtained in two experiments with AP-2 α chain knockdown (not shown). Knockdown of AP1 and clathrin in Jurkat cells resulted in a marked reversal of Nef effect on native HLA-I at 37 o and 26 o C, notwithstanding that Nef was much more potent in Jurkat (29±3 or 28±3% vs. 58±5 or 38±3% in HeLa cells at 37 or 26 o C) cells (Figure 4 A, bottom left). AP2 ablation resulted in a mild enhancement (<20%) of the Nef effect in both cells.
Nef induced robust downregulation of recombinant HLA-A2 in HeLa cells to 30±3 or 24.2±1% of control at 37 o C or 26 o C. AP-1 µ1 or γ (not shown) chain or CHC siRNA knockdown induced a substantial reversal of Nef effect to 75±6 or 74.4±8% at 37 o C and to 72±4 or 69±6% at 26 o C for AP-1 or CHC respectively (Figure 4 A, top right). In Jurkat cells, Nef induced a marked loss (to 8±0.5% of control) of recombinant HLA-I A2 at the PM at 37 o or 26 o C. AP1 or CHC knockdowns led to moderation of this strong Nef effect: 51±3 or 54±5% for AP1 and 55±5 or 52±2% for CHC knockdown at 37 o and 26 o C respectively (Figure 4 A, bottom right). In this background, AP-2 knockdown still induced a slight enhancement of Nef mediated HLA-I A2 downregulation (to 5.5±0.3% of control). AP3 δ chain knockdown did not significantly reverse Nef induced downregulation of native or recombinant MHC-I in HeLa or Jurkat cells at either temperature (data not shown).
Ablation of PACS-1 led to significant reversal of Nef effect on native HLA-I in HeLa cells at 37 o C (from 58±5 to 79±7%) and 26 o C (from 38±3 to 77±4.8%) ( Earlier in this manuscript, we showed that both Arf6Q67L and Arf6T27N mutants partially rectified the Nef mediated downmodulation of native HLA-I in HeLa cells and PBMCs. To confirm these observations we evaluated the effects of siRNA knockdown of Arf6 and ARNO GEF. Ablation of Arf6 using a mixture of three different siRNAs led to a weak reversal of Nef effect on native HLA-I in HeLa cells at 37 o C (from 58±5 to 66±5% of control MFV) and more at 26 o C (from 38±3 to 57±4%). Likewise, ARNO knockdown with a pool of three siRNAs in HeLa cells induced a modest reversal of Nef effect (from 58±5 to 75±6% at 37 o C and from 38±3 to 56±4% at 26 o C) (Figure 4 A, top left). These findings were validated using single species of siRNAs against Arf6 and ARNO (not shown). Knockdown of either Arf6 or ARNO had no effect of Nef mediated downmodulation of recombinant HLA-I A2 in HeLa cells or of native or recombinant HLA-I in Jurkat lymphocytes (Figure 4 A). We considered bystander (off-site) effects of siRNA knockdown strategy by examining the effect of siRNA knockdown on native CD4 in Nef (+) Jurkat and HeLa cells. As illustrated in Suppl. Figure 5, only AP2 and clathrin depletion resulted in a significant reversal of Nef induced CD4 downregulation in Jurkat lymphocytes and HeLa cells.
Thus, our portfolio of cofactor manipulation findings indicated that Nef subverts both antero-and retrograde trafficking MHC-I alleles, more predominantly the anterograde trafficking of recombinant HLA-I A2 in T cells than in HeLa cells while impairing the endocytic traffic of native HLA-I in quiescent PBMCs and HeLa cells. Live Imaging and Dynamics by TPTCFCCS and FRAP. The biochemical methods described above quantify the net effects of Nef, integrated over various long incubation times. We inquired whether Nef effects on innate or recombinant HLA-I (and on the trafficking of CD4 vs. HLA-I) would be visible in real time observations of complex formation between Nef and these receptors in the subcellular organelles. We acquired snapshots (~2 min) for the interactions of Nef-Cerulean Fluorescent Protein (Nef-CerFP) with YFP or Venus tagged HLA-A2, wt CD4 or LL/AA CD4 mutant proteins in the TGN, ER and PM in HeLa cells at 25 o C using TPTCFCCS. The labeling was innocuous as shown by the Suppl. Figure 6. Nef-CFP fusion protein downregulated CD4 and HLA-I A2 as well as unmodified Nef, and conversely HLA-I A2-Venus fusion was downregulated by Nef as well as untagged HLA-I A2. TPTCFCCS records the fluctuations in fluorescence intensity caused by the entry to and exit of labelled molecules from a very small open focal volume. This femtoliter excitation zone is optically defined by the instrument confocal pinhole (for a one-photon FCS instrument) or, in our case, by the shape of the twophoton excitation probability (29). The fluctuations track instantaneous local changes in the chromophore concentration. Both the concentration and diffusion coefficients of the fluorescent molecules are thus determined (http://fcsxpert.com/) (30,31). The diffusion times τ D recovered are functions of both the size of the diffusing complex and the effective viscosity and topology of the local milieu. Cytoplasmic 'free' diffusion rates can be benchmarked and calibrated with known proteins; hence a measured τ D (see methods) provides molecular weight. On a two-dimensional surface like a smooth membrane, the overall complex size becomes less important than the girth of the portion that is stuck in viscid membrane; hence the cross section of the membrane-spanning element could instead be deduced. In convoluted, vesicular surfaces, however, anomalous diffusion (proteins that must follow winding roads -or wander into their dead ends -rather than cross them) yields τ D that is more difficult to link to a particular complex size. Worse, FPs trapped in sub-micron vesicles may even bleach during FCS, excluding them from the average (Stasevich et al, submitted). Nevertheless, τ D is often a useful measure of relative hydrodynamic radii of the complex(es). Two photon, two color Cross-correlation: Simultaneous excitation and detection of two fluorophores with different emission maxima allows cross-correlation of fluorescence intensity vs. time. Cross correlation is simply a measure of how often the two colors blink together, signaling they are transiting the focal spot together. If two fluorescent species do not diffuse together (i.e. are non-interacting), the fluctuations in (cross) intensity will be entirely uncorrelated. If there is 100% complex formation, the fluctuations will be 100% cross-correlated in time. We often refer to the cross correlation as a measure of comobility; i.e. two different colored proteins need not directly bind each other, but if they bind any common partner(s) they will diffuse together and the crosscorrelated blinking will occur. This permits us to examine such binding at very low particle concentrations. Since the ROIs examined by FCS are in the 0.1µ 2 range, number fluctuations are significant, as average occupancy of only a few molecules is common (provided that fluorescent protein expression is maintained at modest sub-µM levels). With correlation in time, TPTCFCCS measurements reflect solely the mobility characteristics of the target(s) in the various subcellular locales. By analyzing the ACF and CCFs from ROIs with different intensities (i.e. concentrations, C i ), the concentrations of free (C C and C V ) and bound (C CV ) Cerulean and Venus tagged proteins can in principle be calculated. FCS has been used to evaluate in 'real' time (minutes) the avidity of macromolecular interactions at the plasma membrane (32)(33)(34) and complex dynamics inside organelles such as mitochondria (35) and the Golgi network (36). While many of these studies have elucidated parameters relating to macromolecular transport and assembly etc., few, if any have used two-photon excitation with its facile crosscorrelation. Cross-correlation of Nef and HLA-I A2 in subcellular organelles. HeLa transfectants expressing HLA-I A2 fused with fluorescent protein (eYFP, Venus or Cerulean) and Nef fused with CFP or Cerulean were used to obtain TPTCFCCS data by single focal point acquisitions of emitted photon count traces. To find and target the ER, GOLGI/TGN and the PM, the organelles were counterstained with Bodipy TR conjugated glibencamide, Bodipy TR ceramide and TR conjugated WGA respectively. The organelle specific dyes were two-photon excited at 840 nm. Then, HLA-I A2-eY/Venus and Nef-CerFP were detected by excitation at 920 nm (avoiding the counterstain). ACF profiles for HLA-I A2-Venus in blue, Nef-CerFP in green and cross correlation (CC) in red are shown for representative data in Figure 5. There was little or no cross correlation between controls Nef-CerFP and Venus-FP or between controls HLA-I A2-Venus and Cerulean-FP (not shown). Since the analysis was limited to the Nef bound fraction of HLA-I A2, the ACF data were fit to a simple diffusion model without any constraints. The cross correlation (CC) profile followed both the Nef & HLA-I A2 ACFs throughout the time range (Figure 5  A). Thus in these organelles the cross-correlating (comobile) species have similar transport properties to the average (both free+bound) proteins. The degree of binding in a simple ACF/CCF plot is directly extracted by comparing the intercept (G o , G(O)) of the CCF fit to the lowest G o of the two ACFs; i.e., the CCF is compared to the most concentrated of the two ACF partners. The CCF G o cannot exceed that ACF Go value, and G o (CCF)/ G o (ACF) determines fraction bound ( i.e. the portion of the more abundant species that is tied up in comobile complexes). Thus, one can deduce that about one third of the more abundant fluorescent species (which in each case happened to be A2-Venus) was complexed with Nef (37% in the ER, 23% in the GOLGI/TGN network and 30% at the PM). As an aside, attempted FCS of Nef and HLA-I A2 within endolysosomal vesicles was unsuccessful, owing to the small vesicular volumes (ibid.) and high nonrandom mobility, which resulted in widespread intensity heterogeneities causing 'stepped' intensity time traces. In the environs where many labeled proteins are confined to actively moving vesicles, FCS data reduction can be disrupted by flow artifacts and "spiking" (i.e. the passage of an enormous fluorescent aggregate through the focal volume); the correction of these is beyond the scope of this manuscript. Comparison of Comobility of Nef with the native vs. recombinant HLA-I at the PM. We inquired whether the native HLA-I would also exhibit ~30% binding for Nef at the PM. Cells expressing Nef-CerFP alone or with HLA-I A2 were stained with Alexa 647 labeled W632 or BB7.2 mAbs against native HLA-I or the A2 allele (Figure 5 B). The fraction (~35%) of native HLA-I (via W632 mAb binding) found complexed with Nef-CerFP at the plasma membrane was comparable to the ~30% HLA-I A2-Venus binding level with Nef. A slightly larger fraction (~40%) of BB7.2 mAb (specific for HLA-I A2) was complexed with Nef, suggesting this bivalent antibody might engage a bit more Nef (perhaps through receptor crosslinking). Table 1 summarizes all the results for HLA-I A2-Venus and Nef-CerFP in the ER and Golgi membrane network and the plasma membrane. Results obtained from studies using fluorescent antibodies targeting either the native HLA-I or recombinant HLA-I A2 are also included. The calculated diffusion coefficients were from fits to a single diffusing complex undergoing 3D diffusion for studies within the ER and TGN; in contrast, 2D diffusion was assumed in the PM. Nef expression induced minor decreases in the diffusion coefficient and a 10% increase in the amount of the faster HLA-I species at the PM ( Table 3). The Ab labeled HLA-I/Nef complexes are large (Table 3), though not quite as large as the largest covalent complexes. On occasion, a small amount of fast Ab component (presumed free) was visible. Both wt and LL/AA mutant CD4 at the plasma membrane were found bound to Nef. Physical interaction of HIV-1 Nef and CD4 has been demonstrated in vitro (37,38) and in vivo in insect and mammalian cell expression systems-expressing both proteins (39). Stable complexes between Nef resistant LL/AA CD4 mutant and Nef have been demonstrated suggesting that the CD4 di-leucine may not be critical for Nef binding, but may instead be needed to interact with the adapter complex (40). We examined the distribution of wt or LL/AA CD4 on live and permeabilized cells expressing wt Nef-GFP (Suppl. Figure 7 A1). In live cells expressing Nef-GFP, wt CD4 was markedly reduced at the PM. In contrast Nef resistant LL/AA CD4 was abundant at the PM. In permeabilized cells, both wt and LL/AA CD4 was visualized in intracellular vesicles in Nef-GFP cells. While the intracellular CD4 represented most if not all of wt CD4, only minor amounts of LL/AA CD4 were sequestered intracellularly (Suppl. Figure 7 A1, compare patterns of EC CD4 with IC CD4). This is consistent with the FACS analysis, showing marked loss of wt but not the LL/AA CD4 at the PM in cells expressing Nef or Nef-CerFP (Suppl. Figure 7 A2 and B). Both others (41) and we (20) have shown that LL/AA CD4 does not undergo constitutive endocytosis. By TPTCFCCS, we found almost all CD4 is sequestered in comobile complexes with Nef: 70±20% or 90±30% of wt or LL/AA CD4 mutant moved in Nef bound complexes (Table 1). From the shape of the CCF (cross correlation function), one can even infer that Nef recruited both CD4 proteins from some sort of larger, slower complex. ACFs of LA/AA CD4 alone, wt CD4 with Nef and LL/AA CD4 with Nef were plotted together. Differences were apparent in the shape of the curves (Figure 5 C). The diffusion coefficient for LL/AA CD4 alone was apparently monodisperse at 0.9 ± 0.7 µm 2 /sec, but in the presence of Nef there were two discernable components with diffusion coefficients of 6.6±2 µm 2 /sec and 0.32±0.17 µm 2 /sec, in approximately equal amounts. For wt CD4 in the presence of Nef, the diffusion parameters were 7.8±4 µm 2 /sec for a 31±12% fraction and 0.37±0.18 µm 2 /sec for a 69±22% fraction (Table 3). This confirms that Nef mobilizes a portion of CD4 into protein (or lipoprotein) complexes that are smaller than those found absent Nef. Unfortunately, attempts to evaluate the crosscorrelation potential between Nef and wt or LL/AA CD4 in intracellular organelles were unsuccessful since 1) there was a paucity of LL/AA CD4 in the intracellular organelles relative to Nef and 2) some intracellular wt CD4 was sequestered in small and highly mobile endolysosomal vesicles, which are photobleached readily.
In all, the TPTCFCCS minute-by-minute snapshots revealed that during the downregulating process, Nef is comobile with about a third of HLA-I (in several cellular locations). Avidity of Nef for CD4 is even stronger in realtime-nearly all CD4 is comobile with Nef at the PM, breaking away from large native complexes to wander the surface more quickly with Nef. FRAP analysis of Nef effect on the lateral mobility of HLA-I A2. FRAP, unlike FCS, measures the diffusion of proteins over distances comparable to the sizes of organelles. We measured the two-dimensional diffusion coefficients (D t ) of various mobile fraction(s) (M f ) of HLA-I A2-Venus in the ER, GOLGI/TGN and at the PM in the presence or absence of Nef by monitoring the FRAP. Three fractions were identified in our study, an immobile fraction, and two others with a difference of 10-20 times in diffusion times. The corresponding diffusion coefficients (D t ) were simply labeled 'slow FRAP' and 'very slow FRAP'. We choose these terms to acknowledge the relative timesale sensitivity of FRAP vs FCS. Fast and slow FCS rates correspond to 10s of µm 2 /sec and ~µm 2 /sec, respectively. Thus slow FCS rates are comparable to the fastest FRAP rates seen. The FRAP very slow and immobile terms would be lost to bleaching in FCS. The percent immobile fraction in FRAP was also measured and included in Tables 2 A and B. In the ER, 44.2% of HLA-I A2-Venus had a D t of 0.34 µm 2 /sec, and another 34% had a D t of 0.0164 µm 2 /sec, and these rates were not significantly altered by Nef expression with the corresponding diffusion rates 0.3 µm 2 /sec for the 50% slow fraction and 0.02 µm 2 /sec for the 34% very slow fraction. Nef slightly reduced the amount of the immobile fraction (22 vs. 16%) ( Figure 6 A and Table 2 A). At 0.34 µm 2 /sec and 0.02 µm 2 /sec, the diffusion rates of the two mobile HLA-I A2 fractions in the Golgi network were remarkably similar to those in ER. However, there was a substantial doubling of the immobile fraction at the expense of the fast moving fraction in the Golgi. Nef induced a slight and statistically insignificant change to 0.22 µm 2 /sec and 0.01 µm 2 /sec. (Figure 6 B and Table 2 A). Nef did not significantly change the diffusion rates of either the slow or the very slow moving HLA-I at the plasma membrane but it significantly decreased the quantity of 'slow' diffusing molecules from 40% to 25% leaving the 'very slow' diffusing quantity almost unchanged (from 40±2% to 44±2 %). This necessarily increased the immobile fraction at the PM from 20±1% to 31±2% (Figure 6 C and Table 2 B). FRAP analysis of Nef effects on the lateral mobility of wt and LL/AA mutant CD4 at the PM. Our efforts at FRAP analysis of wt CD4-Venus in HeLa cells co-expressing Nef-CerFP were unsuccessful, since most if not all the tagged CD4 protein exhibited rapid constitutive endocytosis with rapid diffusion in and out of tiny fast-moving vesicles much less than the typical 1µm×1µm ROI for FRAP analysis. Therefore, we used a LL (413-414)/AA CD4 mutant which does not undergo constitutive endocytosis and is resistant to Nef. Nef did not alter the relative distributions of the mobile and immobile fractions or the diffusion rates of either the slow or the very slow moving fraction of LL/AA CD4 (Figure 6 D and Table 2 B). LL/AA CD4, which lacks constitutive endocytosis, had a D t of 0.073 µm 2 /sec for the faster fraction, almost four time slower than the 0.266 µm 2 /sec rate for the slow HLA-I A2 at the plasma membrane. Nef effects on 2D diffusion at the PM Two diffusion components were observed for CD4 by FCCS analysis, with most of the binding between CD4 and Nef occurring in the more mobile fraction. It is important to remember that FCCS cannot quantify the 'immobile' and the 'slowest' fractions of FRAP (due to bleaching). Most of the Nef at the plasma membrane was found to bind CD4 dileucine mutant. In contrast, only about 30% of HLA-I-A2 at the plasma membrane bound to Nef. We compared slow diffusion parameters obtained for HLA-I A2-and LL/AA CD4 at the plasma membrane by TPTCFCCS with those obtained from FRAP experiments. As illustrated by Figure 7, the slow diffusion coefficient obtained by FRAP is comparable to the slow component observed in FCCS. However, it must be noted that while the in vivo dynamics of proteins or membranes and physical interaction(s) of proteins can be evaluated by FCCS or FRAP, the smallest ROIs evaluated by FRAP are in the µm 2 range, enough to encompass whole sub-cellular organelles, vesicles and cytoskeletal elements. As a result, the relative D t measured by FRAP can reflect complex binding and "sieving" interactions of the fluorescent species with its environment, or sequestration within ≤1 µ structures. Thus FCCS is best at quantifying "neighborhood " transport, dominated by diffusion, while FRAP extends over wider areas ("citywide") more dependent on traffic patterns and jams. Conversely, FCCS is hampered in studying slower diffusion by the possibility that long residence times in the FCS spot link to photobleaching (ibid.). FCS is also less tolerant of non-diffusive transport that may be visible but averaged on the extended timescale of FRAP. Nevertheless, the Nef effect was quite pronounced in FCS experiments and had only subtle effects on FRAP curves.

Inhibition of clathrin dependent endocytosis partially reversed Nef induced loss of HLA-I, retarded lateral mobility of HLA-I and CD4 and decreased their binding to
Nef. HLA-I is endocytosed both by clathrin dependent and independent pathways (42)(43)(44). Nef is presumed to act as a connector between CD4 and endocytic machinery (45,46) through components of clathrin-coated vesicles (CCV). This suggests that an inhibitor of clathrin-mediated endocytosis could reverse Nef-mediated CD4 downregulation. The macrolide antibiotic, ikarugamycin (IKA) from Streptomyces phaeochromogenes sub-sp. ikaruganensis that inhibits the uptake of clathrin-dependent PM receptors without affecting their internal trafficking (47) was in fact shown to block CD4 downregulation in response to PMA or Nef (48).
After confirming that IKA was a specific inhibitor of clathrin dependent endocytosis (Suppl. Figure 8), we tested the effects of IKA treatment on the PM levels of CD4, native HLA-I or HLA-I A2 in Nef (+) vs. Nef (-) HeLa, Jurkat cells or quiescent PBMCs. In HeLa cells, Nef reduced the PM levels of CD4 to 24±3 or 31±2% of control at 37 o or 26 o C. Treatment with 4 or 6 µM IKA for 4 h partially reversed this CD4 effect of Nef; with the PM CD4 levels increasing to 59±7 and 53±8 with 4 µM or 51±6 and 63±7% of control levels at 6 µM IKA at  Table 3). The recovered values for a fit to a two-component diffusion model undergoing 2D diffusion were 2.3±0.8 µm 2 /sec and 0.2±0.1 µm 2 /sec. The faster component had a considerably reduced D t compared to the 4.9 µm 2 /sec D t for untreated cells (Table 3). Perhaps coincidentally, this change correlated with the appearance of larger complexes of HLA-I A2 at the PM unable to be endocytosed (Figure 8 D, left). The slower component did not change significantly in IKA treated cells. IKA marginally reduced the cross correlation (comobility) between Nef and HLA-I A2, resulting in a calculated 19±10% comobility, to be contrasted with the 30±4% found in the untreated cells. IKA affected the D t for plasma membrane associated LL/AA CD4 more profoundly with the rate of the faster component dropping from 6.6 µm 2 /sec to 2.3±0.8 µm 2 /sec (Table 3). This change was accompanied by a drop in the percent bound from 70±20% to 40±10% in the treated cells. Even the D t of the slower component was reduced by >3 fold (Figure  8 D, right and Table 3). IKA treatment has been shown to increase the residence time of CD4 at the plasma membrane (48). In Nef expressing cells, CD4 is cleared from the plasma membrane within 2 hours. IKA treatment prolonged the CD4 presence at the plasma membrane by 60%. Although the overall bound fraction of Nef was reduced (Figure 8 D, right and Table 3), the considerable decrease in the diffusion rates probably resulted from an increase in the large complexes formed at the plasma membrane between Nef, unknown cellular partners and HLA-I A2 or LL/AA CD4 when clathrin function was persistently impaired. Multimolecular interactions between Nef and HLA-I A2 or LL/AA CD4 at the PM were disrupted by cholesterol depletion. MHC-I and MHC-II proteins associate with DRMs and co-distribute with different raft markers upon antibody clustering or during immunological synapse formation (49)(50)(51)(52). In accordance, endocytic trafficking of MHC molecules was dependent on such lipid raft integrity (44,50,53). While we are agnostic about the 'raft' concept in our measurement, cholesterol depletion inhibited both the constitutive and Nef enhanced internalization of HLA-I, but not of CD4 (Supplemental Results and Suppl. Figure 9). MβCD extraction altered the HLA-I A2 & Nef ACF and CCF profiles significantly. The ACFs obtained at representative points on the PM of cells before and after MβCD extraction are shown on Figure 9, left. The diffusion was fit to a 2D model for two mobile species. The diffusion coefficients obtained for untreated cells were composed of a faster component of 4.9±1.1 µm 2 /sec and a slower one at 0.23±0.14 µm 2 /sec. The average diffusion coefficients for cross correlation curves post treatment were 1.76±0.28 and 0.46±0.15 µm 2 /sec (Table 3). These values are graphically represented by the histograms in Figure 10 A. In this case, we have a small increase of the slow diffusion coefficient and a halving of the fast diffusion coefficient. Thus cholesterol depletion induced significant HLA-I clustering and disrupted its interaction with Nef, and this pattern implied that Nef prefered the faster smaller version of these two species. ACF data for LL/AA CD4-eY and Nef-CerFP obtained before and after treatment with MβCD of double transfected cells are shown in Figure 9, right. The average diffusion coefficients obtained for the pretreated cells were 6.6±2.4 and 0.32±0.17 µm 2 /sec for the cross correlation. The post-treatment gave values of 3.7±2.2 and 0.21±0.07 µm 2 /sec (Table 3) For diffusing membrane-spanning lipoprotein complexes, D t should depend strongly on membrane viscosity and the size of the (presumed cylindrical) spanning segment; hence transmembrane segment radii can be calculated from the corresponding diffusion coefficient(s). Presuming lateral diffusion is the only process and modeling our complex as a cylindrical or ellipsoidal structure, we applied the Saffman and Delbruck equation (54,55): Where a is the radius, D L is the lateral diffusion, h is the thickness of the membrane, η w is the viscosity of the surrounding phase, η is the viscosity of the membrane, T is temperature and γ is the Euler constant. We obtained a value of 6 cPoise for η w from measuring the diffusion of cerulean in the cytoplasm of HeLa cells. Viscosity at the PM has been determined to be 1.3 Poise (55,56). Using this value in the formula, we obtained for a diffusion coefficient of 0.2 µm 2 /sec for a membrane spanning protein (5 nm) and a viscosity of ~0.13 Pa⋅sec (1.3 Poise), a radius of ~40 nm. Figure 10 is a graphic illustration depicting the diffusion coefficients of the various mobile fractions together with their relative abundances determined for the different experimental conditions. The numerical values in Table 3 include in addition the putative cylinder radii of various mobile complexes of HLA-I and CD4s at the plasma membrane. As water is much less viscous than membrane, all these calculations pertain only to the membrane-spanning portion of these species. FRAP analyses of cholesterol depletion effects comport with FCS findings. By FRAP analysis we observed that cholesterol depletion increased the size of the immobile fraction of HLA-I A2 at the PM almost two fold (from 20 to 36%) and correspondingly decreased the size of fast moving fraction (Table 4 A).
Cholesterol extraction induced similar changes in the slow and immobile fractions LL/AA CD4 (Table 4 B). Nef expression did not modify these changes induced by cholesterol depletion. Thus the results of FRAP analysis agreed in general with the conclusions deduced from the ACF curves. For FCS, the diffusion coefficients obtained for many of the complexes are similar to those reported for Golgi resident proteins (57) and MHC Class I proteins in the ER (58) and the PM (59). The values obtained for 3D diffusion inside of the organelles yield apparent MW in the millions assuming a cytoplasmic viscosity of 6 cP (60), calculated from the presumed free diffusion of cerulean in the cytoplasm of HeLa cells at room temperature. These unlikely sizes suggested that simple free diffusion models could not be applied to the large multimolecular complexes observed within convoluted or vesicular structures. The degree to which they remain membrane bound on convoluted surfaces vs. diffusing in some nonideal media is unclear; we simply wish to avoid leading readers to numbers for "complex sizes" when effective viscosity and free path is unknown here.

DISCUSSION
The present work establishes that HIV Nef downregulation of MHC-I may involve both proposed mechanisms in human PBMCs: 1) accelerated receptor endocytosis from and aberrant recycling to the plasma membrane and 2) misrouting of newly synthesized MHC-I from TGN to endolysosomes for degradation. These results support the relevance of both the anterograde and retrograde trafficking pathways for Nef-induced MHC-I downregulation established by previous authors (3,13) (10) (14,15), and extend them in four ways: 1) by defining their relative importance in primary HIV target cells, 2) by identifying differential utilization in different cell types, 3) by establishing their relative importance for native versus recombinant HLA-I, and 4) by delineating the differential molecular requirements for the Nef effect in multiple cell types. Our work also provides the first biophysical association measurement between Nef and MHC-I in living cells, and quantifies (minute-byminute) the extent of interaction at the subcellular level. Consistent with the functional effects of Nef on MHC-I downregulation through effects on both anterograde and retrograde trafficking itineraries, we found that Nef associates with the target molecule(s) throughout the cell.
Our data provide an expanded list of specific factors that mediate the Nef effect on native HLA-I in PBMCs, including clathrin, AP1 adapter, the GTPases Arf6, Rab7 and Rab11, as well as dynamin and Eps15. Eps15 is tightly linked to clathrin-dependent endocytosis (61), and dynamin is critical for both clathrin-dependent and -independent endocytosis barring those pathways regulated by Arf6, Cdc42 or RhoA (62)(63)(64)(65). AP1 adapter regulates intervesicular transport, and Arf6 regulates clathrin-independent endocytosis, whereas Rab11 and Rab7 regulate recycling endosomes and endolysosomal fusion, respectively.
As in PBMCs, the mechanism for Nef downregulation of native HLA-I in HeLa cells involved Arf6, Rab7 and Rab11 GTPases, clathrin, and AP1 adapter(s), but, in addition, was also dependent on PACS-I and ARNO. These observations confirm previous reports, and extend knowledge of regulation Nef effect to Rab7 and Rab11.
With regard to recombinant HLA-I A2, we confirmed distinct molecular regulation of the Nef effect in different cell contexts. Thus, in Jurkat cells Nef-induced downregulation of recombinant HLA-I A2 was only reversed by siRNA knockdown of AP1 adapter(s) and clathrin, whereas in HeLa cells knockdown of PACS-1 adapters and genetic inhibitors of Arf6 and Rab GTPases could also reverse the Nef effect. The reason for these differences is not yet known.
Previous in vitro and in vivo investigations using tagged proteins (including Nef-MHC-I hybrids) and co-precipitations identified physical interactions between Nef and HLA-I. In particular, Nef has been reported to bind hypophosphorylated MHC-I early in the secretory pathway, but not mature MHC-I at the cell surface (14). Consistent with this, Nef mutants lacking the ability to downregulate HLA-I failed to coprecipitate with the receptor (19).
Structural determinants of Nef binding to cellular partners include Nef residues 62 EEEE 65 (standardize nomenclature) and to a lesser extent 72 PXXP 75 for direct formation of a ternary complex with the µ1 subunit of AP1 and the tyrosine-containing cytoplasmic domain of HLA-I (17,18,66).
A limitation of this type of approach is that the results report only steady-state measurements without spatio-temoral information. Minute by minute FCCS analysis overcomes these limitations while still providing quantitative information.
Using this technique, we found that nearly a third of HLA-I A2-Venus fusion protein at the plasma membrane and in intracellular organelles (ER and GOLGI/TGN) was present in Nef-bound complexes. Consistent with this, Nef also bound equivalent fractions of fluorescent antibody-tagged native (and recombinant) HLA-I at the cell surface. If only AP1 subunit were critical for stabilizing Nef and HLA-I interaction as has been reported (17,18,66), then FCS should have identified more extensive interactions in the GOLGI/TGN vesicles.
Under steady-state conditions, it has been previously reported that MHC-I molecules at the cell surface are nearly evenly divided into two populations, with mean half-lives of 1 and 20 h; the short-lived fast-trafficking population corresponds to receptor that has not bound peptide or has bound lowaffinity peptide (67). It remains to be determined whether the subpopulation of MHC-I we found associated with Nef by FCCS corresponds to one or the other of these kinetically distinct MHC-I species. MHC-I loaded with fluorescent peptides of different affinities (and colors) could be employed to evaluate binding of peptide loaded MHC-I to Nef, but this is beyond the scope of this manuscript.
Using FCCS, we also found that both wild type CD4 and the endocytosis-resistant LL/AA mutant of CD4 had strong affinity for Nef, with 70±20% of wt or 90±30% of the mutant at the PM associated with Nef. The greater affinity (during the clearance process) of Nef for CD4 vs. HLA-I is in agreement with the endpoint observations that Nef is much more potent in downregulating CD4 vs. HLA-I from the cell surface.
To cross-validate the molecular dynamics of MHC-I downregulation by Nef we used FRAP imaging. Analysis of steady state recombinant HLA-I A2 in HeLa cells lacking Nef identified an immobile fraction and two mobile fractions that had a 10-20 fold difference in diffusion coefficients. The relative distribution of these components was similar to the previous reports of murine H2L in L cells (59) and human MHC Class I on HeLa cells (68), and the diffusion coefficient of the very slow species matched the comparable species identified by single particle tracking (SPT) studies (69). While Nef did not change the diffusion coefficients measured by FRAP of either the slow or the very slow-diffusing fraction of HLA-I at the plasma membrane significantly, there was a significant decrease in the amount of the slowdiffusing fraction and a reciprocal increase in the amount of the immobile fraction. These data suggest that Nef induces clustering of surviving HLA-I and reduces the average lateral mobility of HLA-I. In contrast, Nef did not alter the relative distributions of the mobile and immobile fractions or the diffusion rates of either the slow or the very slow moving fraction of LL/AA CD4, which is not susceptible to endocytosis whether Nef is present or not.
Ikarugamycin (IKA) that has previously been shown (48) to block Nef-mediated downregulation of CD4 in various cell types. Consistent with our clathrin knockdown data, we found that this agent could partially reverse Nef-induced downregulation of native HLA-I in both HeLa cells and resting PBMCs. Moreover, our FCCS analysis indicated that IKA treatment significantly reduced the Nef-bound fraction of both HLA-I and CD4 (by ~50%) and decreased their lateral mobility (by >60%), while facilitating receptor sequestering near or on the cell surface. This analysis in living cells is consistent with previously published observations using hypertonic saline or dominant negative dynamin expression to inhibit clathrin-coated vesicle traffic (43). In a recent report using total internal reflection fluorescence microscopy (70), Nef at the plasma membrane was esclusively associated with clathrin-coated pits and vesicles (CCPs and CCVs). We too observed that wt and three different Nef mutants extensively colocalized with CCVs (Suppl. Figure 10). If all Nef positive spots at the PM were associated with CCP(V)s then Nef binding fractions of CD4 and MHC-I might represent receptors being recruited into the CCVs and If Nef association with clathrin were critical for binding to CD4 and MHC-I, then clathrin depletion or inactivation would eliminate Nef binding to either receptor. This was what was observed in IKA treated cells.
Our data from acute cholesterol depletion experiments further suggested that Nef binding to HLA-I in complexes at the plasma membrane is an important determinant of the Nef effect. Cholesterol depletion inhibited Nef downregulation of HLA-I and changed the auto-and cross-correlation profiles significantly. There was a marked decrease in the diffusion coefficient of the faster species of HLA-I in the cross correlation profile, suggesting significant disruption in the normal Nef:HLA-I interaction. Cholesterol depletion has been shown previously to reduce lateral mobility and increase clustering of HLA-I, resulting from redistribution of plasma membrane PIP2 and reorganization of the cytoskeleton (71,72). The marked decrease in the diffusion of faster HLA-I species in the autocorrelation profile suggests unusual HLA-I clustering, as does the large increase in the immobile fraction of HLA-I seen in the FRAP analysis. In HeLa cells, cholesterol depletion reduced constitutive and Nef enhanced internalization of native HLA-I more than recombinant HLA-I A2, but the reason for this is unknown. (Suppl. Figure 9 C vs. D).
It is important to note that our endocytosis assay values are net results of internalization and recycling. Recycling rates of HLA-I are difficult to quantify since both β2 microglobulin and peptide dissociate from internalized HLA-I, and about half of the internalized receptor is degraded under physiological conditions (53,73). Nevertheless, treatments that inhibited "net" endocytosis increased HLA-I clustering and inhibited Nef binding potential.
If Nef enhances endocytosis of native HLA-I, and depletion of clathrin and AP1 can reverse the Nef effect, then why did depletion of AP-2 not also reverse the Nef effect? Clathrin-coated vesicles are involved in endocytosis and intracellular vesicle traffic between TGN and late lysosomes (74). AP-2 facilitates clathrindriven endocytosis whereas AP-1 facilitates clathrinmediated vesicle trafficking between the trans-Golgi network (TGN) and endosomes (although there is still some question about directionality) (75). In the absence of AP1, both endocytosed HLA-I and newly synthesized HLA-I in transit to the cell surface would have recycled or followed the default pathway to the plasma membrane, rather than being diverted to the TGN and to MVBs/lysosomes. In contrast, depletion of AP2 might divert HLA-I to clathrin-independent endocytic pathways (76), such as through Arf6 that may in turn be subverted by Nef.
It was of interest that mutations at the three Nef motifs deemed critical for the HLA-I effect only induced a partial reversal of Nef effect on both the innate and recombinant HLA-I levels at the plasma membrane in T cell lines. In Hela cells and quiescent PBMCs, mutation at the E62-65 or the M20A motif of Nef completely restored the native, but not the plasmid-expressed HLA-I levels (Suppl. Fig. 11). Whether the EEEE at position 62 is a PACS binding motif (10,12) or whether residues E62-65 and to a less extent P78 in Nef are required for direct interaction with the AP1 µ1 subunit and HLA-I (18,66), the incomplete rescue with the Nef mutants suggests that these critical motifs may be players in both antero-and retrograde trafficking defects.
A limitation of FCCS and other live cell imaging analyses is the uncertainty whether two associated proteins actually bind directly and/or interact functionally, versus whether the correlation is indirect and facilitated by unidentified adapters. In this regard, we are currently attempting to define ternary Nef interactions with HLA-I and candidate vesicle adapters, including clathrin. A related limitation in live cell imaging is artifacts induced by protein overexpression. We have tried to avoid these by limiting recombinant protein expression to modest promotion levels and for short times, and have not noted any vesicular changes (77,78); of course we cannot yet rule out over expression artefacts. We should more hopefully note that our FCS measurements are effectively done at low submicromolar concentration levels, thus these studies have been conducted under minimal perturbation conditions.
Given the HeLa cell volume of ~2 picoliter, our data have been collected at Nef copy levels well under 10 5 per cell, which is within a range compatible with the Nef levels anticipated in early infection.
Further, while we purposely avoided calculating Kd values for MHC or CD4 in unknown multiprotein complexes, it is clear that Nef must have nanomolar affinity for these unknown complexes, as both the complexes and Nef are seen to be present at nM levels in our cross-correlation experiments.
In conclusion, our data support a model in which Nef may associate with a subpopulation of MHC-I throughout the cell and downregulate MHC-I from the plasma membrane through effects on both anterograde and retrograde itineraries. We have extended biophysical knowledge of this important HIV immune evasion system by defining diffusion parameters of HLA-I and CD4 at the PM, yielding the apparent size of the molecular complexes. These systems can be used to further evaluate the molecular basis of the Nef effect on MHC-I dynamics, for example in cells lacking various vesicle transport proteins. In vivo kinetics are essential to gain mechanistic insight of cellular processes. Auto and cross correlation (G Nef or G HLA / G cc ) data was fit to one diffusing component for the Golgi and ER, while a two component 2D diffusion model was used to analyze PM data. Effect of HIV-1 Nef on the diffusion rates of HLA-I-A2 in different subcellular organelles deduced from FCS assay. Dt fast or Dt slow and % fast or slow represent diffusion rates of the fast or slow components and their relative abundances (in %) are given.      Figure 1 A. Nef mediated clearance of innate HLA-I from the plasma membrane of quiescent PBMCs was reversed by genetic inhibitors of endocytosis. Cells were cotransfected with NL-3 Nef or a null mutant (NX) at a twofold molar excess over GFP or G/YFP-tagged dominant negative inhibitors of endocytosis. HLA-I expression for G/YFP-gated cells was measured using Alexa 647 conjugated W/632 mAb. FACS histogram profile of HLA-I for Nef (black) cells is overlaid with the corresponding one for non-Nef (gray) transfectants. HLA-I MFVs of GFP (or YFP) gated populations for the null and the Nef transfectants averaged from four experiments are plotted as histograms (with error bars) in sets of two, with the value for null Nef arbitrarily set to 100. † n=3, p<0.008; * n=4, p<0.008.  HeLa cells were cotransfected with NL-3 (black) or NA7 Nef (gray) or a null mutant (white) at a twofold molar excess over GFP or G/YFP-tagged dominant negative inhibitors of endocytosis. After shifting the transfectants to 26 o C for 18 h, HLA-I expression for G/YFPgated cells was measured using Alexa 647 conjugated W/632 mAb. HLA-I MFVs of GFP (or YFP) gated populations for the null and the two Nef transfectants averaged from four experiments are plotted as histograms (with error bars) in sets of three, with the value for null Nef arbitrarily set to 100. † n=4, p<0.008; * n=5, p<0.008. Figure 1 D. Subcellular distribution of native HLA-I in HeLa cells co-expressing Nef or NX and GFP or GFP (or YFP) tagged effectors or inhibitors of endocytosis. HeLa cells were transfected with a twofold molar excess of HIV-1 Nef over GFP or GFP/YFP fusion protein plasmids. At these ratios, Nef expression was present in all the GFP/YFP + cells. At 18 h after transfection, cells were downshifted to RT (25 o C) for 4-6, h rinsed, fixed in 4% PFA and permeabilized in 0.1% Triton X-100, stained with Alexa-647 conjugated polytropic HLA-I mAb, W632 before processing for microscopy. Individual channels corresponding to GFP and W632 fluorescence extracted from the RGB images are shown.   Hela cells were transfected with an HLA-I A2 IRES GFP bicistronic vector. HLA-I A2 was stained with Alexa 647 conjugated BB7.2 mAb. β-COP, EEA-1, clathrin, furin, and M6P-R were detected with the respective rabbit antisera, TGN with sheep anti-TGN 46, and LAMP and CD63 with the respective murine mAbs followed by secondary staining with Alexa 568 conjugated anti-rabbit, -sheep or -mouse IgGs respectively. Individual channels corresponding to HLA-I A2, the respective organelle and GFP fluorescence extracted from the RGB images are shown below composite pictographs of HLA-I A2 in green, and organelle in red. Scale bars are 10 µ. Figure 3 A. Nef induced downregulation of native HLA-I was reversed by siRNA knockdown of AP-1 subunits or CHC in Jurkat cells and quiescent PBMCs. A). Selected subsets (Mock, AP1 µ1a, AP2 µ2, AP3 δ or CHC siRNA transfected) of Jurkat cells were analyzed by immunofluorescence microscopy using antibodies against AP1 γ chain, AP2 α chain, AP3 δ chain or CHC (top left). Following siRNA knockdown cells were transfecetd with bicistronic plasmids encoding GFP and Nef or NX mutant. Relative HLA-I MFVs for GFP gated cells are plotted pairwise for Nef (+) and Nef (-) cells, with the latter values arbitrarily assigned 100% in each pair. FACS histogram profiles of HLA-I for selected siRNA transfectants. MFVs for NL4-3 Nef (black) expressers are overlaid with the corresponding one for the non-Nef (gray) cells (bottom left). HLA-I MFVs of CD8 gated populations for the null (gray) and the Nef expressers (in the context of siRNA knock down) averaged from four experiments are plotted pairwise as histograms (with error bars) in sets of three, with the value for null Nef arbitrarily set to 100 (right). † n=3, p<0.03 and * n=3, p<0.005 for the indicated siRNAs vs. none in Nef (+) cells. Figure 4 A. Differential HLA-I response to Nef in Jurkat versus HeLa cells in the context of siRNA knockdown of vesicular adapter proteins, clathrin, PACS-1, Arf6 and ARNO. A) Histograms showing average MFVs (with error bars) of native HLA-I (left) or plasmid expressed A2 allele (right) in HeLa (top) or Jurkat (bottom) cells expressing Nef in the context of siRNA knockdown of the indicated proteins. Jurkat or HeLa cells were transfected for 36 h with the respective siRNAs twice before cotransfection with Nef or N/X (vector) plasmid with one expressing CD8, CD4 and/or HLA-I A2 expression plasmid(s) and incubated at 37 o or 26 o C for 18-24 h. ‡ n=4, p<0.04, † n=4, p<0.02, * n=4, p<0.008 for the indicated siRNA knockdowns vs. no siRNA (CON) in Nef (+) cells. Figure 4 B1-C3. Approximately 2X10 6 (HeLa or Jurkat) cells were disrupted in 0.5 ml of lysis buffer and initially 50 µl of 15,000 g supernatants were immunoblotted for actin. After adjusting the volume to constant actin values, the indicated vesicular or adapter proteins were detected by immunoblotting extracts of cells treated with the siRNAs listed below the repective blots, using rabbit antibody against AP1 µ1 chain (B1) and AP2 µ and murine mAbs against AP2 α chain (B2), AP3 δ chain (B3), clathrin heavy chain (B4), rabbit antiserum against PACS-I (C1) and murine mAbs against Arf6 (C2) and ARNO (C3).

Figure 5 A)
Single point FCCS data acquisition from ER, GOLGI/TGN and PM loci. Transfections were done at 26 o C for 4 h using 0.5 µg each of HLA-A2 or CD4 and Nef fusion plasmids per 10 5 cells on each of 2-well Titer-Tek cuture chambers. ER was visualized by staining with Bodipy TR conjugated glibencamide, GOLGI/TGN with Bodipy TR conjugated ceramide and PM with TR conjugated WGA, all of which were excited at 840 nm, while HLA-I A2-V and Nef-C were detected by excitation at 920 nm. ACF profiles for Nef-C in green, A2-V in blue and Cross Correlation (CC) in red are shown on the right panels of each row (100 µm x 100 µm). B) Auto-and cross-correlation plots of Alexa 647 conjugated W632 (left) or BB7.2 (right) mAb specific for native HLA-I or the plasmid expressed A2 allele respectively in the context of Nef-C expression in the respective cells. C) ACF (CD4 in blue and Nef in green) and CCF (red) profiles in cells co-expressing wt or LL/AA CD4-V and Nef-C are shown in the left and middle panels respectively. Changes in diffusion rates of wt or LL/AA CD4 in the presence of Nef were deduced by comparing the CD4 ACF profile of LL/AA CD4 only (blue), LL/AA CD4 & Nef (green) and wt CD4 & Nef (red) Right panel). Note that the diffusion coefficient changes leftward from something very large to somewhat smaller.  The FRAP data were normalized to the slow component of the FCCS data. If only one component was identified in the FCCS analysis, then FRAP data were normalized to this D t . Immobile fraction has been plotted with a D t of 1e -5 due to the log scale constraint. Arrow (B1) and curly brackets (B2) denote the FRAP/ FCCS coefficient(s) that were normalized and have the same "height". They could be thought as a single component.  Figure 9. Cholesterol depletion disrupts the association of Nef with HLA-I A2 (left) or LL/AA CD4 (right) at the PM. The ACFs (receptor in blue and Nef in red) and CCFs (red) data points were obtained at single points on the membrane before or after MβCD extraction.