We previously reported that expression of Nefs from HIV-1 subtype B strains (NL4.3, NA7 and LAI) in thymic progenitors impaired the development of T cells [24,25]. To assess whether this function is conserved in other primate lentiviral Nef proteins, we analyzed a panel of previously described HIV-1 group M (subtype B and C), HIV-1 group O, HIV-2, SIVcpz, SIVgor, SIVmac, SIVblu and SIVsmm nef alleles [28,29] (overview in Table 1). Human thymic CD34⁺ T-cell progenitor cells were transduced with a retroviral vector, co-expressing Nef and the enhanced green fluorescent protein (eGFP) marker from a single bicistronic mRNA, and assayed in vitro in FTOC. After 21 days of FTOC culture, CD34⁺ cells transduced with the control vector expressing eGFP, developed into double positive (CD4⁺CD8⁺), CD4⁺CD8^- and a few CD4^-CD8⁺ single positive cells with similar levels of surface marker expression (CD4, CD8β, CD3) compared to that of non-transduced eGFP^- cells (Figure 1A, B), as reported before [24,25]. In comparison, HIV-1 Nef expressing cells showed reduced CD4 and CD8β expression. As reported before [24,25], thymocytes were generated in reduced numbers and skewed to higher levels of cell surface CD3 expression compared to eGFP^- cells from the same culture (Figure 1B). Since transduced and non-transduced progenitors are not separated before FTOC, the non-transduced thymocytes generated serve as an internal control for the transduced population. Nearly all of the Nef proteins tested reduced thymic output, some to such an extent that hardly any eGFP⁺ and thus Nef expressing cells could be detected (representative dot plots shown in Figure 1C). To measure Nef induced thymic depletion, we calculated the thymocyte generation ratio (TGR) defined as the ratio of the percentage of eGFP⁺ thymocytes harvested to the percentage of eGFP⁺ progenitors put in FTOC (Figure 1D). This TGR parameter was shown before to be a robust measure of T-cell generation potency from transduced progenitors [25]. All HIV-1 group M subtype B and C, as well as two out of three group O Nef proteins showed a pronounced effect on thymopoiesis (Figure 1D). This was due to a developmental defect and not due to toxicity, as Nef expressing thymocytes grown in suspension culture showed comparable survival to control eGFP transduced cells ([24] and data not shown). Remarkably, the natural mutant O8 (HIV-1 group O) Nef protein failed to disturb T-cell development. Of note, this allele is mutated in the valine-glycine-phenylalanine (VGF) domain which is critical for many Nef functions (MHC-I and CXCR4 downregulation, PAK2 binding and cofilin hyperphosphorylation, inhibition of T-cell receptor (TCR) triggering-induced actin ring formation, targeting Lck to the trans-Golgi network and enhancement of infectivity and replication, but not for CD4 downregulation) [29,30]. For HIV-2, EHO Nef reduced T-cell generation (TGR 0.1) much more potently than 171 Nef (TGR 0.6) in comparison to control eGFP transduced cultures (TGR 1.4) (Figure 1D and Table 1). The latter protein was shown to be defective for CXCR4 downregulation, and reduced in its ability to downregulate MHC-I [29] and CD3 (Table 1). All of the SIV Nef proteins tested reduced T-cell generation, albeit with different potency. SIVcpz was the least potent, while SIVmac blocked T-cell development almost completely. To accomplish this, the SIV derived Nefs apparently can interact with the relevant human proteins in the developing thymocytes. Overall, these results demonstrate that the capacity to reduction of thymic output is a conserved feature of highly divergent lentiviral Nef proteins.

The Nef protein of the O8 strain is deleted in the VGF domain and in the first proline of the neighboring polyproline domain (PxxP). Although the importance of the latter domain for deregulation of T-cell development was shown before [25], the functional relevance of the VGF motif in this respect is unknown. To investigate the relevance of this VGF domain for reduced thymic output by Nef alleles from various group M HIV-1 strains, we mutated the VGF motif to an alanine triplet (VGF-AAA) in subtype B (NA7 and B2), and subtype C (C1422) Nefs as described before [29]. Table 2 shows an overview of all mutants used in this study, their ability to downregulate the cell surface markers assessed in primary CD4+ T cells (PBLs iso primary CD4+ T cells), to bind PAK2 and their effect on thymic output as measured by TGRs. When the VGF mutants were expressed in thymocytes, cell surface CD4 and CD3 levels were similar to that of thymocytes expressing wild-type Nefs (Figure 2A). However, the number of thymocytes generated was significantly increased (Figure 2B). This is in line with our previous observations, indicating that domains important in the signaling properties of Nef, such as binding to SH3 domains or PAK2, are of importance for reduced thymic output in the presence of Nef [25].

To investigate this further, we used HIV-1 subtype B Nef single point mutants known to be important for PAK2 binding while not affecting other known Nef activities. These proteins are mutated in the most C-terminal phenylalanine residues F191 and F195 for HIV-1 NA7 and SF2 strains respectively, which are important residues of the HIV-1 Nef PAK2 activating structural motif (PASM) [20,36,41]. As described before [35], the NA7 F191H mutant has a reduced affinity for PAK2 binding, while the NA7 F191R mutant does not bind PAK2 at detectable levels. Similarly, F195 in SF2 Nef was described to be important for PAK2 binding [20,36,37,42]. To demonstrate AU1-tagged SF2 Nef proteins (wild-type, F195I and F195A) differed in binding capacity for PAK2, they were transfected in 293 T cells to measure activated PAK2 binding as described before [43,44]. While in this setting residual binding of PAK2 to SF2 Nef F195I was detected, binding to SF2 Nef F195A was undetectable (Figure 3). When assayed for their effect on T-cell development, reduced PAK2 binding did not affect the CD3 or CD4 cell surface staining profiles (Figure 4A) but did diminish the effect on T-cell development (Figure 4B) in comparison to wild-type Nef. However, while NA7 F191R did not affect T-cell generation at all, SF2 F195A was still partially active (TGR 0.7) despite the fact that no detectable PAK2 activity was bound (Figure 3). This suggests that in the case of SF2 either a minute but undetectable PAK2 activity bound to Nef is sufficient to affect T-cell development, or that other Nef functions contribute to reduced thymic output.

Since PAK2 interaction with HIV-1 Nefs seems to play an important role in reduced thymic output, we investigated whether this mechanism is conserved in other lentiviral Nef proteins. First, we analyzed the effect of mutations in prolines P104 and P107 of the polyproline stretch (PxxP) of SIVmac239 Nef. The mutation of P104 was described to marginally affect PAK2 binding while P107 was shown to be essential for PAK2 binding [45]. Surprisingly, we found that P107A and the P104A/P107A double mutant (PxxA and AxxA) could still bind PAK2 albeit with strongly reduced efficacy (Figure 5A). We therefore sought other mutants that were selectively defective for PAK2 binding. As mentioned above, we previously described a PASM in HIV-1 Nef comprised of amino acids L85, H89, R188 and F191 [36,46]. When we aligned SIVmac239 protein sequence with a consensus subtype B Nef sequence [36], SIVmac239 Nef residues I117, H121, T218 and Y221 were revealed as potential candidates involved in the binding of PAK2 due to their positional homology to PASM residues of HIV Nef (Figure 5B). To determine whether the four candidate SIV Nef amino acids are indeed involved in PAK2 binding, we mutated each one to three different substitutions and measured PAK2 binding. While substitution at all positions could reduce PAK2 binding (Additional file 1: Figure S1), our results showed that the H121R and Y221R mutant proteins expressed well and were strongly reduced in PAK2 binding. These identified residues therefore constitute the PASM of SIVmac, homologous to that of HIV-1 Nef. We also generated a double mutant that was not able to bind active PAK2 at all (Figure 5A). The H121R, Y221R single and double mutants were capable of downregulating the expression of cell surface CD4, CD3, CD28 and to a lesser extent MHC-I to similar levels as SIVmac239 Nef in cell lines and infected human peripheral mononuclear cells (Additional file 1: Figure S2) and thymocytes (Figure 6A). In addition, the mutations of H121R/Y221R did not affect downregulation of MCH class II and upregulation of invariant chain Ii (Additional file 1: Figure S2D), nor hyperactivation of NFAT (data not shown) by Nef. By contrast, downregulation of CXCR4 was compromised (Table 2).

We found that individual or combined mutations in the PxxP motif or in residues H121 and Y221 of the SIVmac239 disrupted PAK2 interaction, but had little if any effect on reduction of thymic output (Figure 6B). This shows that PAK2 binding is not essential, or that residual but undetectable PAK2 binding is sufficient, for the reduction of thymic output by SIVmac239 Nef. In contrast to HIV-1 Nef, SIVmac Nef downregulates CD3 in CD4⁺ lymphocytes [28]. Since CD3 is important for TCR signal transduction and T-cell development critically depends on TCR signaling [47], we hypothesized that this function may explain why loss of PAK2 binding did impair the effect of SIVmac239 Nef on thymic output. To examine this we utilized two previously described SIVmac Nef deletion mutants: nef_∆153 and nef_∆183[48]. The Nef_∆153 protein does not downregulate CD4, CD28 and MHC-I or enhance viral infectivity and replication but still reduces CD3 surface expression. By contrast, the nef_∆183 allele has also lost the latter function due to an additional deletion of 10 amino acids. These phenotypical characteristics were confirmed in transduced thymocytes that developed in FTOC (Figure 6A) and in CD4⁺ PBLs (Table 2). As expected, both proteins were unable to bind PAK2 (Figure 5). However, Nef_∆153 still reduced thymic output, whereas Nef_∆183 was inactive (Figure 6B). Thus CD3 downregulation by SIVmac Nef is sufficient to block T-cell development. The experiments reported above cannot exclude other SIVmac239 Nef functions to contribute to this effect. However, no SIVmac mutant Nef protein was available that selectively lost PAK2 and CD3 downregulating properties.

Finally, we also tested the Nef allele of SIVblu infecting blue monkeys (Cercopithecus mitis) and a mutant thereof (RR-AA) containing amino acid changes R129A and R130A. In comparison to the wild-type protein, the SIVblu RR-AA mutant lost its ability to downregulate CD3 in thymocytes and PBLs (Figure 6A and Table 2) and bind PAK2 (Figure 5A). Nonetheless, the protein still impaired T-cell generation, albeit to a lower extent when compared to SIVblu wild-type Nef (TGR 0.5 and 0.2 respectively, Figure 6B). As CXCR4 signaling is important for T-cell development [49,50], the downregulation of CXCR4, conserved in SIVblu RR-AA possibly contributes to this residual effect.

HIV-1 group M and group O isolates were obtained from the AIDS Research and Reference Reagent Program, while the HIV-2 strains were previously isolated from patients attending the AIDS clinic at the Institute of Tropical Medicine in Antwerp, Belgium, with the approval of the ethical committee after written informed consent. The nef genes from these isolates were available in the LZRS-IRES-eGFP retroviral vector constructed before [29]. Similarly, nef sequences from SIVcpz (GAB2), SIVmac (239), SIVsmm, SIVgor, SIVblu and derived mutants [33,39,40] were amplified and cloned into the same vector. In some cases tagged proteins and new mutants were created by overlap extension PCR or site-directed mutagenesis (primers in Additional file 1: Table S1). Double mutant SIVmac 239 H121R/Y221R was constructed starting from cloned single mutant PCR amplicons, taking advantage of the PsiI restriction site that is internal relative to the 121 and 221 residues. Single and double mutant nef were further subcloned by introducing the BglII/NdeI SIV nef mutated fragments into a construct comprised of a ClaI/EcoRI fragment from SIVmac239 FL SPX (kindly provided by Dr. R. Desrosiers, Harvard Medical School, Boston). These were used to clone by PCR the mutant nef sequences in expression vectors and retroviral vector plasmids, and in replication competent HIV-1 reporter virus (see Additional file 1: Methods). For in vitro kinase assays (IVKA), nef sequences were amplified with primers containing an AU1 tag sequence for C-terminal tagging and cloned in the pCG expression plasmid as described before [39]. The integrity of the constructs and the nef genes was confirmed by Sanger sequencing, protein expression was evident from Western blot and/or biological activity.

The Phoenix-Amphotropic packaging cell line was transfected with LZRS retroviral vector plasmids to produce nef-IRES-eGFP bicistronic mRNA encoding retroviral vectors as previously described [24]. Isolation, culture and transduction of CD34⁺ thymus cells was performed as described before [24,25]. Briefly, CD34⁺ cells were seeded on RetroNectin (Takara Biomedicals, Otsu Shiga, Japan) coated culture plates with half of the medium volume replaced by retroviral supernatants, supplemented with SCF (10 ng/mL) and IL-7 (10 ng/mL). Transduced CD34⁺ cells were used after 24 hours for fetal thymus organ cultures (FTOC). The excess of transduced progenitor cells, which were not used in FTOC, were kept in culture for 72 hours, to determine the transduction efficiency that varied between 10% and 20%. Child thymus tissue, removed during cardiac surgery, was obtained and used following the guidelines of the Medical Ethical Commission of Ghent University Hospital. Written informed consent was provided according to the Declaration of Helsinki.

Isolation of thymic lobes from fetal nonobese diabetic (NOD)-SCID mice and subsequent murine FTOC were performed and analyzed as described previously [25,59]. Mice were treated and used in agreement with the guidelines of the local ethical committee. To asses thymic depletion due to Nef expression we calculated the thymocyte generation ratio (TGR), which is defined by the ratio of the percentage of eGFP⁺ thymocytes harvested to the percentage of CD34⁺ eGFP⁺ progenitors that were put in FTOC [25,59]. To asses cell surface modulation by the different Nef proteins, we isolated CD4⁺ cells (PBLs) from buffy coat peripheral blood mononuclear cells (normal blood donors, Red Cross, Ghent, Belgium) by negative selection using paramagnetic beads (MACS; Miltenyi Biotec, Bergish Gladbach, Germany). After isolation, the cells were cultured for 3 days in RPMI medium supplemented with 2 mM L-glutamin, 10% heat-inactivated fetal calf serum, phytohemagglutinin (1 μg/mL; Thermo Fisher Scientific, Waltham, USA), 20 ng/mL IL-2 (Peprotech, Rocky Hill, USA), 100 U/mL penicillin, and 100 g/mL streptomycin. Thereafter, PBLs were transduced with retroviral vectors on Retronectin RetroNectin (Takara Biomedicals, Otsu Shiga, Japan) coated culture plates with half of the medium volume replaced by retroviral supernatants, supplemented with IL-2 to keep final cytokine concentrations constant. After another 2 days of culture, cells were harvested and stained for flow cytometry. Antibodies used were directly labeled mouse monoclonals anti-CD1 phycoerythrin [PE] and anti-HLA-A, -B,- C-PE (Becton Dickinson, Erembodegem, Belgium), anti-CD8β-PE (Coulter, Miami, FL) anti-CD3 allophycocyanin [APC], anti-CD4-APC and anti-CXCR4-PE-Cy7 (Miltenyi Biotec, Bergisch Gladbach, Germany), flow cytometers used were FACSCalibur (Becton Dickinson) and MACSQuant (Miltenyi Biotec).

Western blot and IVKA was performed as described before [43]. Briefly, HA-tagged PAK2, dominant-active Cdc42V12 and AU1-tagged Nef were transfected into 293 T cells. Lysates of transfected cells were in part subjected to immunoblot to quantify total amount of tagged proteins, and in part used for anti-AU1 Babco, Richmond, CA) immunoprecipitation using protein G-Sepharose beads. Immunoprecipitate was incubated to detect autophosphorylation activity of PAK2 in the presence of ³²P-ATP by autoradiography of SDS-PAGE.

Analyses were performed using the GraphPad Prism version 5.00 statistical software (GraphPad Software Inc., La Jolla, CA, USA), non-parametric tests used were Mann–Whitney U test and Kruskal-Wallis with Dunn’s correction for multiple testing.