HIV-1 Nef Induces Hck/Lyn-Dependent Expansion of Myeloid-Derived Suppressor Cells Associated with Elevated Interleukin-17/G-CSF Levels

ABSTRACT Human immunodeficiency virus (HIV) or simian immunodeficiency virus (SIV) infection causes myelodysplasia, anemia, and accumulation of inflammatory monocytes (CD14+ CD16+) through largely unknown cellular and molecular pathways. The mouse cells thought to be equivalent to human CD14+ CD16+ cells are CD11b+ Gr1+ myeloid-derived suppressor cells (MDSC). We used HIV transgenic (Tg) mouse models to study MDSC, namely, CD4C/Nef Tg mice expressing nef in dendritic cells (DC), pDC, CD4+ T, and other mature and immature myeloid cells and CD11c/Nef Tg mice with a more restricted expression, mainly in DC and pDC. Both Tg strains showed expansion of granulocytic and CD11b+ Gr1low/int cells with MDSC characteristics. Fetal liver cell transplantation revealed that this expansion was stroma-independent and abrogated in mixed Tg/non-Tg 50% chimera. Tg bone marrow (BM) erythroid progenitors were decreased and myeloid precursors increased, suggesting an aberrant differentiation likely driving CD11b+ Gr1+ cell expansion, apparently cell autonomously in CD4C/Nef Tg mice and likely through a bystander effect in CD11c/Nef Tg mice. Hck was activated in Tg spleen, and Nef-mediated CD11b+ Gr1+ cell expansion was abrogated in Hck/Lyn-deficient Nef Tg mice, indicating a requirement of Hck/Lyn for this Nef function. IL-17 and granulocyte colony-stimulating factor (G-CSF) were elevated in Nef Tg mice. Increased G-CSF levels were normalized in Tg mice treated with anti-IL-17 antibodies. Therefore, Nef expression in myeloid precursors causes severe BM failure, apparently cell autonomously. More cell-restricted expression of Nef in DC and pDC appears sufficient to induce BM differentiation impairment, granulopoiesis, and expansion of MDSC at the expense of erythroid maturation, with IL-17→G-CSF as one likely bystander contributor. IMPORTANCE HIV-1 and SIV infection often lead to myelodysplasia, anemia, and accumulation of inflammatory monocytes (CD14+ CD16+), with the latter likely involved in neuroAIDS. We found that some transgenic (Tg) mouse models of AIDS also develop accumulation of mature and immature cells of the granulocytic lineage, decreased erythroid precursors, and expansion of MDSC (equivalent to human CD14+ CD16+ cells). We identified Nef as being responsible for these phenotypes, and its expression in mouse DC appears sufficient for their development through a bystander mechanism. Nef expression in myeloid progenitors may also favor myeloid cell expansion, likely in a cell-autonomous way. Hck/Lyn is required for the Nef-mediated accumulation of myeloid cells. Finally, we identified G-CSF under the control of IL-17 as one bystander mediator of MDSC expansion. Our findings provide a framework to determine whether the Nef>Hck/Lyn>IL-17>G-CSF pathway is involved in human AIDS and whether it represents a valid therapeutic target.

FIG 1 Accumulation of immature myeloid CD11b 1 Gr1 1 cells in CD11c/Nef Tg mice. (A) Increase of spleen (SP) weight (g) measured at 10 months of age. Non-Tg (n =6) and Tg (n = 6) mice were used. (B to D) Representative FACS profiles (B) and quantification of percentages (C) or absolute cell numbers (D) of CD11c-negative CD11b 1 Gr1 low/int and CD11b 1 Gr1 hi cells in spleen, BM, and blood of 4-to 6-month-old Tg (1) and control non-Tg (2) mice. (E and F) Wright-Giemsa staining of total spleen cells from Tg and non-Tg mice, total spleen cells (E) and Tg splenocyte Gr1 low/int or Gr1 hi subsets purified by cell sorting (F). Note abundant vacuoles in Tg cells and different morphology of Gr1 low/int and Gr1 hi cells. (G) Representative FACS profiles and quantification of percentage of Ly6C 11 CD43 11 spleen cells of non-Tg (n =5) and Tg (n = 5) mice stained for Ly6C and CD43 gated on CD11b 1 cells. The percentage of Ly6C 1 CD43 11 (R4) cells was significantly different (P # 0.01) in Tg versus non-Tg mice, but the two other Ly6C/CD43 cell subsets (R3 and R5) were not significantly different. Statistical analysis was performed using Student's t test. (H) Blood of Tg and non-Tg mice was separated on a Ficoll gradient, as described in Materials and Methods, and cells were collected in the high-or low-density region of the tube and stained for CD11b and Gr1. A representative FACS profile is shown. (I) The number of spleen CD11b 1 Gr1 low/int and CD11b 1 GR1 hi cells was measured in CD11c/Nef Tg (1) and non-Tg (2) mice of different ages. Data were obtained from FACS analysis. (J to L) T cells in spleen and peripheral LN (pLN) of CD11c/Nef Tg (1) and non-Tg (2) mice were assessed by FACS analysis after staining for CD4 and CD8 and expressed as percent (J) or absolute cell number (K and L). Note that the percentage of both CD4 1 and CD8 1 T cells tends to be decreased in both organs (J), but the absolute number of both T cell subsets is decreased in LN (K) but not in spleen (L). **, P # 0.01; ***, P # 0.001. addition to morphology, the lower mean fluorescence intensity and percentages of F4/80, CD80, CD86, major histocompatibility complex class I (MHC-I), and MHC-II ( Fig. 2A to C) also suggest an immature phenotype of these accumulating CD11b 1 Gr1 1 cells.
To determine whether accumulation of these myeloid cell populations was caused by enhanced proliferation, mice were fed with bromodeoxyuridine (BrdU). Incorporation of BrdU was increased in Tg CD11b 1 Gr1 1 cells from spleen, blood, and BM relative to non-Tg cells (Fig. 2D and E). We also evaluated survival of these cells. Both subsets of Tg CD11b 1 Gr1 1 (Gr1 low/int or Gr 1/hi [data not shown]) either among total spleen cells ex vivo (Fig. 2F) or purified and cultured in the presence of apoptotic stimuli (LPS or GM-CSF [data FIG 2 Analysis of maturation/activation markers, proliferation, and death/apoptosis of CD11b 1 Gr1 1 cell subsets from CD11c/Nef Tg mice. (A to C) Representative FACS profiles (A) and mean fluorescence intensity (MFI) (B) of the indicated cell surface markers on CD11b 1 Gr1 low and CD11b 1 Gr1 hi spleen cells from Tg (n = 15) and non-Tg (n = 15) mice and quantification (C) of the percentage spleen CD11b 1 GR1 low cells from Tg (1) and non-Tg (2) mice expressing the indicated cell surface markers. (D and E) Representative FACS profiles (D) and calculated ratios (Tg/non-Tg) of percentage (E) of purified BrdU-positive CD11c-negative CD11b 1 Gr1 1 cells in spleen, BM, and blood of Tg and non-Tg mice. Blood HD cells are from a high-density Ficoll gradient. A value of 1 was assigned to non-Tg cells. Results are from three independent experiments. (F and G) Representative FACS profiles of apoptotic/dead cells and quantification of percentage of spleen CD11b 1 Gr1 low/int cell subsets ex vivo (F) or of purified CD11b 1 Gr1 low spleen cells cultured in vitro for 24 h in the presence of LPS (G). Cells were stained for cell surface CD11c, CD11b, Gr1, 7-AAD, and annexin V and gated on CD11c-negative cells. Statistical analysis was performed using Student's t test (*, P # 0.05; **, P # 0.01; ***, P # 0.001).
not shown]) for 24 h (Fig. 2G) showed increased survival relative to non-Tg ones, as assessed by 7-aminoactinomycin D (7-AAD) and annexin V staining. These data suggest that accumulation of Tg CD11b 1 Gr1 1 cells results from their increased proliferation and resistance to cell apoptosis/death.
Inducible nitric oxide synthase (iNOS)-mediated T cell suppressive activity of CD11b + Gr1 low/int cells from CD11c/Nef Tg mice. CD11b 1 Gr1 low/int cells have been shown to have suppressive activity (4) against CD4 1 and CD8 1 T cells in models of mouse tumors or in other contexts (4). To determine whether Tg CD11b 1 Gr1 1 cells exhibit T cell-suppressive functions, purified CD11b 1 Gr1 1 cells were cocultured with sorted-purified CFSE-labeled, normal C3H ( Fig. 3A and B) or AD10 TcR Tg (Fig. 3C) T cells stimulated in vitro with anti-CD31 and CD28 or Pcc antigen, respectively. AD10 TcR mice were used to facilitate readout, as all their T cells harbor a single TcR recognizing a single antigen (Pcc). After 3 days in culture, CFSE fluorescence (reflecting the number of cell divisions) was analyzed by a fluorescence-activated cell sorter (FACS). In this assay, both Tg and non-Tg CD11b 1 GR1 low/int cells suppressed CD4 1 and CD8 1 T cell division to the same extent compared to division of T cells in the absence of suppressor cells ( Fig. 3A to C), indicating that Tg cells have retained, but not increased, suppressive activities on a per-cell basis. In contrast, in the same assay, Tg or non-Tg CD11b 1 Gr1 hi cells showed no suppressive activity (data not shown), as expected.
We next measured the in vivo suppressive functions of Tg and non-Tg CD11b 1 Gr1 low/int cells by their adoptive transfer into normal B10BR mice, along with CSFE-labeled Tg AD10 TcR CD4 1 T cells stimulated with their cognate antigen (Pcc) presented by DC. FACS analysis was performed by gating on AD10 TcR 1 CFSE 1 CD4 1 T cells. This analysis showed that both Tg and non-Tg CD11b 1 Gr1 low/int cells had the ability to modestly suppress proliferation of AD10 CD4 1 T cells in vivo (Fig. 3D). The lower levels of suppression observed in vivo, compared to in vitro, may reflect the purity of CD11b 1 Gr1 low/int cell preparations. Because the CD11b 1 Gr1 1 purification technique used for in vitro suppression experiments (cell sorting) was inadequate to obtain the high number of cells required for in vivo suppression in a sufficient number of mice, we had to use an alternative purification strategy (MACS anti-biotin beads). This led to preparations of CD11b 1 Gr1 low/int cells of lower purity (70% versus 95%). Our use of target CD4 1 T cells for in vivo experiments also may have contributed to lower suppression. It was reported that the same MDSC can suppress proliferation of CD4 1 T cells in vivo less efficiently than that of CD8 1 T cells (18). Finally, it is possible that Pcc-driven signaling for AD10 CD4 1 T cell proliferation inherently is less responsive to MDSC suppression in vivo. We already documented that the percentage of CD4 1 T cells stimulated by Pcc (m4 1 m5 CFSE divisions), in the absence of CD11b 1 Gr1 1 cells (control group), was substantially higher in vivo than in vitro (80% 6 3% versus 61% 6 7%).
NO production is one of the main pathways of suppression by monocytic CD11b 1 Gr1 low cells (4). We first assessed the involvement of iNOS in suppression activity by coculturing non-Tg CD4 1 or CD8 1 T cells with purified suppressive CD11b 1 Gr1 low/int cells in the presence or not of L-NMMA, an iNOS inhibitor. L-NMMA restored .90% of T cell division in such cocultures (Fig. 3E), strongly suggesting that NO is involved in the suppressive functions of Tg CD11b 1 Gr1 low/int cells.
To further evaluate the contribution of NO in suppressive activities of CD11b 1 Gr1 low/int cells, we next bred the CD11c/Nef Tg mice on the iNOS gene-deficient background and assayed the in vitro T cell-suppressive activity of CD11b 1 Gr1 low/int cells from these mice. CD11b 1 Gr1 low/int spleen cells from heterozygous (1/2) and homozygous (2/2) iNOS-deficient non-Tg or Nef Tg mice had decreased suppressive activity against CD4 1 and CD8 1 T cell proliferation relative to control CD11b 1 Gr1 low/int cells from iNOS wild-type (WT; 1/1) non-Tg or Nef Tg mice ( Fig. 3F and G). These results suggest that Tg CD11b 1 Gr1 low/int cells suppress T cell proliferation largely through the iNOS pathway.
Together, these data show that Tg CD11b 1 Gr1 low/int cells have retained NO as a major mechanism of suppression, similar to that found in normal non-Tg cells, on a per-cell basis. Thus, CD11b 1 Gr1 low/int cells, which accumulate in Nef Tg mice and have suppressive activities, appear to be MDSC. GR1 low/int cells from CD11c/Nef Tg mice show T cell suppressive activity mediated by iNOS. Sorted purified CD11b 1 Gr1 1 cells (10 5 ) from Tg or non-Tg mice were cocultured for 3 days with CFSE-labeled normal CD4 1 or CD8 1 T cells (10 5 ) purified by cell sorting and stimulated with anti-CD3 and anti-CD28 to induce their proliferation. (A and B) Representative FACS profiles (A) and quantification (B) of CFSE dilution in non-Tg CD4 1 and CD8 1 T cells cocultured (1) or not (2) with purified CD11b 1 Gr1 low/int cells from Tg or non-Tg mice. In panel B, data are expressed as a ratio of the percentage of stimulated T cells reaching 4 and 5 divisions (m4 1 m5) in the presence (1) of CD11b 1 Gr1 low/int cells over those cultured in their absence (2) (normalized at 1). Results from nine experiments are shown. (C) CFSE dilution of purified CD4 1 T cells from AD10 TcR Tg mice stimulated with Pcc and cocultured (1) or not (2) with CD11b 1 Gr1 low/int cells from non-Tg or Tg mice. Values are from three experiments and expressed as in panel B. (D) In vivo adoptive transfer. CFSE-labeled CD4 1 T cells (3 Â 10 6 to 4 Â 10 6 ) from AD10 TcR Tg mice were transferred by i.v. tail vein inoculation into normal recipient B10BR mice; 48 h later, mice were injected i.v. with non-Tg or Tg purified CD11b 1 Gr1 low/int cells (2 Â 10 6 to 3Â 10 6 ) and, within 1 h, inoculated s.c. with Pcc-loaded DC (3 Â 10 6 ). Values normalized at 1 in the absence of transfer are from 3 independent experiments (n = 5 Tg and 5 non-Tg). (E) Quantification of CFSE dilutions of non-Tg CD4 1 (left) or CD8 1 Accumulation of early myeloid progenitors in CD11c/Nef Tg mice. To explore the possible origin of MDSC and other cells of the granulocytic lineage, we next analyzed hematopoietic progenitors. Multipotent BM hematopoietic cells have been defined as the lin 2 Sca1 1 c-kit 1 (LSK) population. For myeloid cell production, LSK differentiate and produce common myeloid progenitors (CMP). CMP differentiate further to give rise to either granulocyte-macrophage progenitors (GMP), megakaryocyteerythrocyte progenitors (MEP), or monocyte-macrophage-DC progenitors (MDP). GMP can differentiate into both granulocytes (neutrophils) and monocytes (19). In addition, both GMP and MDP can yield LyC6 hi and LyC6 2 CD43 1 monocytes (20), whereas MDP is also the precursor of CD11c 1 DC (21).
When the same number (10 3 cells) of cell-sorted purified Tg or non-Tg progenitor cells (GMP, CMP, or LSK) were plated in methylcellulose supplemented with interleukin-3 (IL-3), IL-6, SCF, and erythropoietin to induce their differentiation, comparable numbers of total CFU were generated (data not shown), suggesting their similar potential for differentiation on a per-cell basis. However, when total cells from BM, spleen, and blood were assayed, the number of CFU generated in vitro was significantly higher in Tg than in non-Tg compartments (Fig. 4D), indicating a higher number of Tg hematopoietic precursors. The majority of more numerous Tg colonies were more condensed and bigger than non-Tg ones (Fig. 4E) and represented CFU-GM, CFU-M, and CFU-G (Fig. 4F), indicating the presence of higher numbers of Tg myeloid precursors and suggesting their increased proliferation. To test this possibility, CFU-GM were picked, stained with CFSE, and replated in methylcellulose. FACS analysis of CFSE dilution (reflecting the number of cell divisions) demonstrated higher proliferation of Tg than non-Tg CFU-GM (Fig. 4G). In addition, BM cells cultured in the presence of IL-4 and GM-CSF, to differentiate them into DC, showed lower percentages of Tg DC (CD11c 1 CD11b 1 ) and higher Tg CD11b 1 Gr1 1 cell production relative to non-Tg controls (Fig. 4H), suggesting skewed differentiation of DC precursors. Erythroid colonies, generated in Tg cultures of total BM cells or of purified CMP, were also decreased relative to those in control non-Tg cultures (Fig. 4I), consistent with decreased MEP observed ex vivo by FACS analysis ( Fig. 4B and C). Together, these results suggest that the higher number of Tg CD11b 1 Gr1 1 cells originates from a higher number of proliferating myeloid GMP.
To determine whether the expansion of myeloid progenitors in CD11c/Nef Tg mice could result from Nef expression in these cells, we measured cell surface CD11c expression on different myeloid cells by FACS, because the promoter of the CD11c gene is driving expression of nef in these Tg mice. We confirmed that early myeloid progenitors (GMP and CMP) express only low levels of CD11c on a small percentage of progenitors (Fig. 4J), as previously reported (22). However, Western blot analysis on the more abundant myeloid cell subsets showed that Nef is expressed in DC (CD11c 1 ) but not in CD11b 1 Gr1 1 neutrophils (Fig. 4K), as expected, since the latter are CD11c negative.  Therefore, it appears that the accumulation of CD11b 1 Gr1 1 cells in CD11c/Nef Tg mice is not a consequence of Nef expression in their precursors but is indirect and not cell autonomous.
The CD11b + Gr1 + cell accumulation in CD11c/Nef Tg mice is stroma independent. We used a transplantation strategy to study the role of the stroma in Tg myeloid cell accumulation. Donor fetal liver (FL) cells (containing large numbers of hematopoietic stem cells) were from Pep3b (CD45.1 1 ) C3H (CD45.2 1 ) Tg or non-Tg mice. These mice harbor the Pep3b marker to distinguish donor from host cells. FL cells were transplanted into lethally irradiated normal (CD45.2 1 ) C3H hosts. In both Tg and non-Tg reconstituted chimeras studied 3 to 5 months later, ;90 to 95% of the hematopoietic cells were of donor origin, i.e., CD45.1 1 CD45.2 1 . In chimeric mice reconstituted with Tg cells, an accumulation of donor-derived CD45.1 1 CD45.2 1 CD11b 1 Gr1 1 cells was observed in BM, spleen, and blood ( Fig. 5A and B). Among these, a lower percentage and mean fluorescent intensity (MFI) of maturity markers were observed ( Fig. 5C and D). CD4 1 and CD8 1 T cells also were decreased (data not shown). Proliferation of spleen CD11b 1 Gr1 1 cells of chimeric mice, measured by BrdU incorporation in vivo, was enhanced (Fig. 5E). Spontaneous (Fig. 5F) or lipopolysaccharide (LPS)-induced ( Fig.  5G) apoptosis/death of Tg relative to non-Tg CD11b 1 Gr1 1 donor cells was decreased. FACS analysis showed that the number of donor myeloid Tg GMP in BM was increased, whereas MEP were decreased (Fig. 5H).
These phenotypes are very similar to those described above for CD11b 1 Gr1 1 cells in CD11c/Nef Tg mice themselves, indicating that the accumulation of CD11b 1 Gr1 1 cells in CD11c/Nef Tg mice can be independent of the stroma.
The accumulation of CD11b + Gr1 + cells in CD11c/Nef Tg mice is abrogated in the presence of non-Tg hematopoietic cells. The apparent skewed differentiation of purified myeloid Tg progenitors in vitro ( Fig. 4H and I) suggests that they are impaired. To test this hypothesis, we compared their reconstitution potential in vivo with that of non-Tg progenitors in the same mouse. For this, we used a transplantation strategy similar to that described above, except that mixed (1:1) FL cells from Tg (CD45.1 1 CD45.2 1 ) and non-Tg (CD45.1 1 ) mice were used as donors. The control represented mixed (1:1) FL cells from CD45.1 1 non-Tg and CD45.1 1 CD45.2 1 non-Tg mice. Analysis was performed as described above 4 to 5 months posttransplantation by gating on donor CD45.1 1 or CD45.1 1 CD45.2 1 cells. No accumulation of donor Tg CD11b 1 Gr1 1 cells (relative to non-Tg ones) was observed in any of the organs tested ( Fig. 5I and data not shown), indicating that the CD11b 1 Gr1 1 cell accumulation in Tg mice can be abrogated by the presence of non-Tg hematopoietic cells. This result suggests that donor Tg myeloid precursors were outperformed by non-Tg ones in a context of transplantation.
The accumulation of CD11b + Gr1 + cells is also present in CD4C/Nef Tg mice expressing Nef through the regulatory elements of the human CD4 gene. To determine whether similar accumulation of CD11b 1 Gr1 1 cells would be observed in Tg mice expressing Nef in the same cells as those expressing CD4 in humans (myeloid cells, such as monocytes, macrophages, DC [23], and some of their progenitors [24][25][26]), we studied the CD4C/Nef Tg mice that express Nef under the regulatory elements of the human CD4 gene (14). We already reported that myeloid CD11c 1 CD11b hi DC accumulate in spleen of these older CD4C/Nef Tg mice (16), as they also do in CD11c/ Nef Tg mice (17). Analysis of C3H CD4C/Nef Tg mice showed that the percentage of CD11b 1 Gr1 1 cells was higher in blood, spleen, and BM than in their non-Tg littermates ( Fig. 6A and B). This was observed in Tg mice bred on five different backgrounds (C3H/ HeNHsd, C57BL/6, CBA, BALB/c, and FVB) (Fig. 6C) and was most evident in older Tg mice   (data not shown), as in CD11c/Nef Tg mice discussed above (Fig. 1I). However, myeloid cell accumulation appeared to be less extensive in CD4C/Nef than in CD11c/Nef Tg mice.
Myeloid cell accumulation has not been studied in CD4C/HIV Wt Tg mice expressing all HIV genes. We reported that the latter Tg mice develop numerous multisystem phenotypes undistinguishable from those of CD4C/Nef Tg mice (14,27). It remains unknown whether another HIV gene(s) collaborates with Nef for promoting myeloid cell accumulation.
Interestingly, however, such myeloid cell accumulation was not observed in CD4C/ HIV MutH Tg mice (data not shown) expressing all HIV genes except nef (14), indicating that Nef is necessary and sufficient to elicit this phenotype.
We next investigated the status of BM cells in CD4C/Nef Tg mice. We found that the number of more mature benzidine-positive erythroid BM cells was reduced in CD4C/Nef Tg mice (Fig. 6D), suggesting impaired erythroid cell differentiation. In addition, colony assay in methylcellulose was performed with these BM cells. Results showed an enhanced number of total BM colonies relative to those from non-Tg mice (122.5 6 16.3 versus 89.5 6 7.8). However, Tg colonies first grown in methylcellulose, followed by incubation in suspension with GM-CSF and then replated in methylcellulose with GM-CSF to further differentiate them, were not capable of full differentiation in the presence of GM-CSF and remained undifferentiated as CFU-GM (non-Tg, ,1; Tg, 53 6 2 colonies). Tg early erythroid BFU-E colonies were also severely depleted (Fig. 6E), apparently in a cell-autonomous way (Table 1), likely explaining the lower number of more mature benzidine-positive Tg BM erythroid cells observed (Fig. 6D). Finally, we evaluated the capacity of the BM from these Tg mice to recover from a severe insult, a method frequently used to assess the integrity of hematopoietic progenitors. We challenged these Tg mice with the drug fluorouracil (5-FU), toxic for BM, and analyzed their BM progenitors 10 days later. Normal non-Tg mice recovered relatively well from this treatment, showing a number of colonies similar to that from untreated non-Tg mice (Fig. 6F). However, a severe failure to recover was evident in BM of 5-FU-treated Tg mice, both morphologically (data not shown) and by the low number of BFU-E, GM-CFU, and GEMM-CFU colonies generated relative to those from 5-FU-treated non-Tg mice (Fig. 6F), strongly suggesting impairment of myeloid and erythroid progenitor differentiation.
We next investigated whether the expansion of myeloid progenitors and the severe BM failure observed in CD4C/Nef Tg mice result from Nef expression in these progenitor cells through the CD4C regulatory elements. We already reported that HIV or reporter human CD4 is expressed in mature myeloid cells (macrophages and DC) of CD4C/HIV or CD4C/hCD4 Tg mice (28). For this experiment, we used CD4C/hCD4 and CD4C/hCD4 1 CCR5 Tg mice and measured the cell surface expression of the reporter human CD4 protein (hCD4) by FACS and immunofluorescence analyses. In these two founder lines, the reporter was expressed in LSK and in early myeloid progenitors (LSK.CMP.GMP) (Fig. 6G and data not shown), but hCD4 was minimally expressed in CD11b 1 Gr1 1 cells (Fig. 6G), as expected. These results show that the regulatory elements of the human CD4 gene are active in mouse myeloid and early (LSK) cell progenitors, consistent with expression of CD4 in human myeloid progenitors (24,26).
Therefore, these results suggest that Nef expression in early and myeloid progenitors of CD4C/Nef Tg mice has direct (cell-autonomous) effects. It is likely responsible for the severe failure of BM recovery observed in these Tg mice following 5-FU challenge. Such Nef expression in BM myeloid precursors, alone or with Nef-expressing DC (16), may favor the accumulation of CD11b 1 Gr1 1 cells, at the expense of erythroid differentiation, in these CD4C/Nef Tg mice.
The accumulation of CD11b + Gr1 + cells in CD4C/Nef is transplantable but is abrogated in the presence of non-Tg hematopoietic cells. To determine whether the accumulation of CD11b 1 Gr1 1 cells in CD4C/Nef Tg mice was stroma dependent, we performed FL cell transplantation. FL cells from CD45.1 1 CD45.2 1 CD4C/Nef Tg mice or from control non-Tg littermates were transplanted into lethally irradiated normal C3H hosts and sacrificed 3 to 4 months posttransplantation. An accumulation of donor-derived CD11b 1 Gr1 1 cells was observed in spleen and blood (but not in BM) of Tg relative to non-Tg chimera (Fig. 6H), indicating that this phenotype is stroma independent, as found in CD11c/Nef Tg mice ( Fig. 5A and B).
We also performed a similar transplantation experiment with FL cells from non-Tg or Tg mice (CD45.1 1 CD45.2 1 ) mixed (1:1) with FL cells of non-Tg (CD45.1 1 ) mice as donors. No accumulation of donor CD11b 1 Gr1 1 cells was observed relative to non-Tg controls in spleen (Fig. 6I), BM, or blood (data not shown). This result indicates that the accumulation of CD11b 1 Gr1 1 Tg cells can be abrogated by non-Tg hematopoietic cells, as in CD11c/Nef Tg mice (Fig. 5I), consistent with the 5-FU data (Fig. 6F) showing that their progenitors are impaired.
Nef-induced CD11b + Gr1 + myeloid cell accumulation is associated with enhanced levels of G-CSF mediated by IL-17. The very low CD11c (Fig. 4J) and undetectable Nef (Fig. 4K) expression in BM progenitors of non-Tg mice and in CD11b 1 Gr1 1 cells of CD11c/Nef Tg mice suggested that the latter cells accumulate through a non-cell-autonomous mechanism, such as cytokine production by other cells. Among these, granulocyte colony-stimulating factor (G-CSF) is a well-known mediator of myeloid cell expansion (29,30), in some cases through IL-17 (31). We measured these two cytokines in serum of non-Tg and CD11c/Nef Tg mice and found both of them to be significantly increased in Tg mice ( Fig. 7A and B) and correlated (Fig. 7C). The levels of G-CSF also correlated with the proportion of CD11b 1 Gr1 1 cells in the blood (Fig. 7D).
To determine whether higher IL-17 levels cause the enhanced G-CSF production, we neutralized IL-17 in vivo by injecting anti-IL-17A antibody (Ab) into a group of CD11c/Nef Tg mice and measured G-CSF levels 7 days after the initiation of this treatment. The levels of G-CSF were significantly reduced in anti-IL-17RA-treated Tg mice but not in Tg mice treated with irrelevant IgG (Fig. 7E), indicating a positive control of IL-17 on G-CSF production and suggesting that G-CSF contributes to the expansion of CD11b 1 Gr1 1 myeloid cells in these mice.
Src deletion favors, but Hck/Lyn deficiency abrogates, the Nef-mediated expansion of CD11b + Gr1 + myeloid cells of CD4C/Nef Tg mice. Because Nef has been reported to bind to members of the Src family, in particular to Hck (32), and to activate them (33), we studied the involvement of these signaling molecules in the expansion of CD11b 1 Gr1 1 cells. First, Hck levels were measured by Western blot analysis (Fig. 8A), and Hck activity was evaluated by an in vitro kinase assay (IVKA) (Fig. 8B). Both were found to be slightly elevated in Tg cells, confirming the activating role of Nef for Src kinases reported in other cells (33,34).
Next, we bred Tg mice on Src family gene-deficient backgrounds. Intriguingly, in CD4C/Nef Tg mice harboring homozygous (2/2) deletion of Src, the expansion of CD11b 1 Gr1 1 cells was larger than that in control heterozygous (1/2) Src Nef Tg mice or than in non-Tg mice (Fig. 8C), suggesting that Src signaling inhibits Nef activity. This could occur either directly in myeloid cells (possibly through its binding to Src) or indirectly in other cells not expressing Nef but involved in the accumulation of CD11b 1 Gr1 1 cells. Finally, we bred HIV Tg mice on an Hck/Lyn double gene-deficient background. It is already known that mice deficient for Hck and Lyn show an accumulation of immature myeloid cells (7) very similar to that of the CD11c/Nef and CD4C/Nef Tg mice shown above. We found that deletion of Hck/Lyn induced expansion of CD11b 1 Gr1 1 cells in non-Tg double Hck/Lyn knockout (KO) mice (Fig. 8D), confirming an earlier report (7). However, the further HIV-mediated expansion of these cells was prevented in homozygous (2/2) double Hck/Lyn gene-deficient HIV Tg mice, but not in heterozygous (1/2) or single gene-deficient HIV Tg mice (Fig. 8D), strongly suggesting that Nef favors myeloid cell expansion through Hck/Lyn. Because CD4C/Nef Tg mice and Hck/Lyn double KO mice show similar accumulation of CD11b 1 Gr1 1 cells and because the loss of Hck/Lyn was reported to be associated with decreased phospho-SHIP (pSHIP) and enhanced p-Stat-5 (7), we measured these effectors. We found that myeloid cells from Nef Tg mice show decreased pSHIP (Fig.  8E) and increased p-Stat-5 ( Fig. 8F and G), as reported in Hck/Lyn KO mice (7), suggesting that loss of Hck/Lyn and Nef function share a common molecular pathway.

DISCUSSION
Our data show that HIV-1 Nef, expressed in Tg mice, can trigger the progressive stroma-independent accumulation of CD11b 1 Gr1 1 granulocytic cells and MDSC, most likely through their increased proliferation and resistance to death as well as expansion of their progenitors. The MDSC of Nef Tg mice exhibit the same characteristics as those generated in other diseases (5): they are immature and show iNOS-mediated immunosuppression against T cells. They may have contributed to the impaired control of oral C. albicans infection previously reported in CD11c/Nef Tg mice (17).
Apparent role of DC in Nef-induced CD11b + Gr1 + cell expansion. Two Tg mouse models were investigated here, CD11c/Nef and CD4C/Nef Tg mice. In CD11c/Nef Tg mice, we documented that Nef is mainly expressed in mature DC (17) (Fig. 4K), and this promoter is known to also be active in precursor pre-cDC and, to a lesser extent, in pDC (22). CD4C/Nef Tg mice express Nef in several cell subsets, such as CD4 1 thymocytes, mature CD4 1 T cells, and macrophages, but also in DC and pDC (16). Since DC from both Nef Tg strains express nef (16,17), Nef-expressing Tg DC or pDC are likely to be involved in the accumulation of CD11b 1 Gr1 1 cells in these mice. Indeed, a similar accumulation of CD11b 1 Gr1 1 cells was observed after depletion of CD11c-positive DC  or pDC in CD11c/DtA Tg mice (35)(36)(37). This cell accumulation was stroma-independent, abrogated in mixed chimera (35,36) (as in Nef Tg mice) and thought to result from depletion of DC (35) or pDC (37). Inhibition of DC differentiation by S100A9 also was found to induce accumulation of MDSC (38). A similar phenomenon as in Nef Tg mice was also observed in SIV-infected macaques. After experimental deletion of CD4 1 T cells, it was reported that a very large number of macrophages and of HLA-DM 1 antigen-presenting cells (APC) are infected (39,40) and that classical (CD14 1 CD16 2 ) and proinflammatory (CD14 1 CD16 1 ) monocytes accumulate (39). Interestingly, among these infected APC, some are likely to be DC known to harbor the HLA-DM marker (41,42), suggesting that infected DC contribute to myeloid cell expansion in this SIV model, as seems to be the case in Nef Tg mice.
Involvement of IL-17/G-CSF in Nef-induced CD11b + Gr1 + cell expansion. The accumulation of CD11b 1 Gr1 1 cells in CD11c/DtA Tg mice depleted of their DC or pDC was reported to be through enhanced expression of Flt3 ligand (35), MCP-1 (37), or elevated IL-17A-producing cells (36). Similarly, in Nef Tg mice, the presence of immature, poorly differentiated and/or functionally impaired (16, 17) Nef-expressing DC or pDC may cause non-cell-autonomous (bystander) accumulation of CD11b 1 Gr1 1 cells. Consistent with this notion, we documented enhanced IL-17 levels in serum of CD11c/ Nef Tg mice, as reported by some groups for SIV (43) and HIV (44) infection. Activated DC have been shown to stimulate the production of IL-17 by other cells (45,46). IL-17 is known to favor the induction of MDSC and of cells of the granulocytic lineage, both in mouse and human (47)(48)(49)(50), through G-CSF (31). Here, we showed that the high levels of IL-17 were responsible for the increased levels of G-CSF in Nef Tg mice, indicating that the IL-17/G-CSF axis (31) is maintained despite the presence of Nef. G-CSF is elevated in HIV-infected subjects (51,52) and is well known for its activity promoting granulopoiesis, erythroid suppression, and MDSC expansion (29,30,53), all observed in Nef Tg mice. Thus, IL-17/G-CSF are likely to be involved in the bystander accumulation of both subsets of CD11b 1 Gr1 1 cells in these Tg mice. However, we cannot ruled out that other pathways, not investigated here, such as a bystander effect of Nef secreted in exosomes, contribute to this phenotype.
Other cell subset(s) may be involved in Nef-induced CD11b + Gr1 + cell expansion in Tg mice. Additional cell subset(s) may contribute to myeloid cell accumulation in CD4C/Nef Tg mice, since Nef is expressed not only in DC or pDC but also in T cells and other cells of the myeloid lineage, such as macrophages, Kupffer cells (14,17), and osteoclasts (54), which are also targets of HIV or SIV infection. Indeed, we found here, using the human CD4 reporter Tg mice, that this CD4C promoter is also quite active in a fair percentage of early hematopoietic progenitors and more weakly active in a small proportion of myeloid progenitors (Fig. 6G), consistent with the reported low expression of CD4 in a small fraction of human BM progenitors (24,26). However, in human patients, it has remained controversial (reviewed in references 13, 55, and 56) whether HIV can infect hematopoietic progenitors (26,57) or not (58,59). Independent of this controversy, the aberrant growth/differentiation of CD4C/Nef Tg myeloid colonies in clonogenic assays and the documented activity of the CD4C promoter in some early and myeloid progenitors suggest that part of the MDSC accumulation in CD4C/Nef Tg mice results from expression of Nef in myeloid precursors through a cell-autonomous process.
Interestingly, in both Nef Tg strains, the accumulation of CD11b 1 Gr1 1 cells was abrogated in mixed Tg-non-Tg chimera, a phenomenon also reported in mixed CD11c/DtA/WT chimera (35,36). This result indicates that the biology of transplantable Nef-expressing Tg myeloid cells (possibly DC or pDC and their precursors) can be modified by the presence of transplantable non-Tg hematopoietic cells. The nature of this Tg/non-Tg cell interaction remains unclear. Tg early or myeloid precursors may be defective in critical myeloid cell differentiation signal(s) and may be outcompeted by non-Tg cells. This scenario is most likely occurring in CD4C/Nef Tg mice in which BM regeneration is severely impaired (Fig. 6E and F). Alternatively, non-Tg cells may inhibit Tg cell survival or function. The mechanisms behind this intriguing result may help to explain, at least partly, some features of human AIDS, notably the impressive success of BM transplantation in some HIV 1 patients and the late appearance of myeloid cell expansion and of other myeloid-related phenotypes, such as neuroAIDS. Indeed, CD163 1 perivascular macrophages infiltrating the CNS were reported to be derived from circulating CD16 1 monocytes (13). Early in AIDS and especially after BM transplantation, the high ratio of normal versus HIV-infected hematopoietic cells may prevent myeloid cell expansion, either directly as in non-Tg/Tg chimera or by preventing their infection, as shown in CD4 1 T cell-depleted SIV-infected macaques (39,40). When this ratio decreases, higher number of myeloid cells, especially DC (as documented for APC in CD4 1 T cell-depleted SIV-infected macaques [40]) and/or hematopoietic progenitors may get infected, and myeloid cell expansion may proceed.
Nef-induced CD11b + Gr1 + cell expansion mediated through Hck/Lyn. Nef binds to and activates members of the Src kinase family, preferentially Hck (33). We previously reported that Lck (another Src kinase) is activated by Nef in thymocytes of CD4C/ Nef Tg mice and that these cells were depleted (34). Similarly, in CD11c/Nef Tg mice, Hck activity was found to be increased in spleen cells (Fig. 8B), consistent with the activating effect of Nef on Hck in other cells (33) and on Lck in human T cells and in T cells of CD4C/Nef Tg mice (34). Moreover, enhanced Stat-5 phosphorylation was documented in Nef Tg mice ( Fig. 8F and G), possibly as a result of increased Hck activity, capable of phosphorylating Stat-5 (60). In addition, deletion of Hck/Lyn abrogates Nefmediated myeloid cell expansion in these double Hck/Lyn knockout Nef Tg mice (Fig.  8D), indicating that these effectors are required for Nef functions. Thus, it is likely that the observed constitutive expression of activated Hck (and likely that of Lyn) in Nef Tg mice is contributing to myeloid cell expansion in these Tg mice. The generation of Tg mice expressing activated Hck/Lyn through the same regulatory elements (CD11c and CD4C) as those used in Nef Tg mice would be helpful to validate this apparent gain-offunction mechanism.
It is also possible that Hck/Lyn behaves as a loss of function regarding Tg CD11b 1 Gr1 1 accumulation despite being activated by Nef. Indeed, Nef Tg mice show phenotypes very similar to those of double Hck/Lyn-deficient mice, i.e., a stroma-independent accumulation of CD11b 1 Gr1 1 cells and myeloid progenitors, decreased pSHIP, and increased p-Stat-5 levels in myeloid cells, as reported before (7) and confirmed here ( Fig. 8D to G).
Genetic Hck rescue experiments, as we did with Lck in T cells (34), may clarify which of these two possibilities is more likely to explain the phenotypes observed in Nef Tg mice.
Are these mouse phenotypes relevant to human disease? As models for human AIDS, these Nef Tg mice have obvious limitations. In contrast to HIV in humans, there is no virus replication or immune response against HIV. The nef gene also is expressed in a much larger number of DC (for CD11c/Nef and CD4C/Nef Tg) or of CD4 + T cells (for CD4C/Nef Tg). The latter characteristic does not necessarily invalidate their relevance. Indeed, the very high number of myeloid cells expressing HIV genes is not unique to these Nef Tg models and also occurs during SIV infection, especially in macrophages (39,61,62). In one SIV model, based on anti-CD4 antibody-mediated CD4 1 T cell depletion, ;80% of productively infected cells are macrophages (39). In addition, in the latter model, a large number of microglial cells (39) and of HLA-MD 1 APC become infected (40). Some of these APC are likely dendritic cells known to be positive for HLA-DM (41,42).
Interestingly, despite these limitations, important features or phenotypes of HIV or SIV infection are replicated in our Tg mouse models. First, CD14 1 CD16 1 or other subsets of monocytes, sharing many of the characteristics of mouse CD11b 1 Gr1 1 cells (3), are expanded during HIV or SIV infection (8)(9)(10)(11)(12), including in the CD4 1 T cell-depleted SIV model discussed above, expressing SIV in a very large percentage of myeloid cells (39), some of them likely DC (40), as in Nef Tg mice. Second, DC and pDC are infectable in vitro (63) and infected to some extent during HIV or SIV infection (64,65), including in humanized mice (66) and very likely in the CD4-depleted SIV model where they were identified as HLA-DM + (40). Our results with CD11c/Nef Tg mice suggest that infection of DC favors the accumulation of myeloid cells, as we documented in these Tg mice. Third, anemia is one of the best correlates of disease progression in HIV infection (67). MDSC accumulation during HIV or SIV infection is associated with impaired erythroid cell differentiation (68), as we found in both Nef Tg strains. Fourth, increased vacuolization of circulating monocytes and BM granulocytic cells, which is a striking characteristic of AIDS (69)(70)(71), is also a feature of CD11c/Nef Tg CD11b 1 Gr1 1 cells. Finally, almost all studies of BM histology in a large number of AIDS patients have reported frequent myelodysplasia, hyperplasia, and accumulation of cells of the immature granulocytic and eosinophil lineages (71)(72)(73)(74)(75)(76)(77)(78)(79), similar to that found in BM of patients with myelodysplasic syndrome (MDS), to which it was compared (71,75). This BM phenotype of human AIDS appears to show some similarities with the one described here in Nef Tg mice. Consistent with this, decreased numbers of erythroid colonies (BFU-E) were documented in both Nef Tg models, as in HIV (80)(81)(82) or SIV (83,84) infection. Intriguingly, however, the apparent and well-documented increased number of immature myeloid precursors observed microscopically in BM of HIV-or SIV-infected humans or animals is not reflected in colony assays. Rather, decreased numbers of human (80)(81)(82)85) or simian (83,84) myeloid progenitor colonies were measured in these in vitro assays. This apparent paradox is not understood and may reflect their poor survival ex vivo compared to those from Tg mice.
Together, our present data indicate that important phenotypes of human AIDS (BM granulocytic hyperplasia, decreased erythroid precursors, and accumulation of MDSC) can be reproduced in Nef Tg mice and are associated, in these mice, with a number of other characteristics typically found in human AIDS. This suggests that these mouse models, despite their limitations, are helpful in highlighting the critical role of Nef in the development of some AIDS phenotypes and for understanding at least some aspects of myeloid cell biology in human AIDS.
ELISA for G-CSF and IL-17 detection. Reagents to detect G-CSF (number 900-M103) or IL-17 (number 900-M392) by enzyme-linked immunosorbent assay (ELISA) were from Peprotech, Inc., and the assay was run as recommended by the supplier.
Purification of CD4 + and CD8 + T cells. Peripheral (axillary, inguinal, cervical, and brachial) lymph nodes (LN) were collected, and single-cell suspensions were stained with biotin-conjugated anti-mouse CD19, B220, DX5, TcRgd , Ter119, CD11c, CD11b, and Gr1 for 10 min, followed by incubation with anti-biotin-microbeads for 15 min. T cells were purified by negative selection with an autoMACS system. In some experiments, CD4 1 T cells were purified using a mouse CD4 1 T cell enrichment kit and EasySep immunomagnetic cell separation (without columns), from StemCell Technologies. The purity of T cells was 95 to 97%, as detected by FACS analysis.
In vitro cell proliferation by CFSC fluorescent dye labeling. CFSE labeling was performed as described previously (90) with a Vybrant CFDA-SE cell tracer kit (Molecular Probe).
In vitro suppression of T cell proliferation by Gr1 + cells. Purified T cells, stained with CFSE, were stimulated in anti-CD3-and anti-CD28 (3 mg/ml)-coated 96-well plates (10 5 cells/well). Stimulation of AD10 CD4 1 T cells was done with BM-differentiated DC and Pcc (100 mg/ml) at a ratio of 1:1. Purified CD11b 1 Gr1 1 cells (10 5 well) were added at the same time as T cells (1:1 ratio) in complete medium (250 ml). After 3 days, T cells were analyzed by flow cytometry.
In vivo suppression of T cell proliferation by CD11b + Gr1 + cells. CD4 1 T cells from donor AD10 TcR mice were purified by negative selection with lineage biotin-conjugated antibodies (mouse CD4 1 T cell enrichment kit; from Stem Cell Technologies) and stained with 1 mM CFSE for 15 min, washed in RPMI, and injected (3 Â 10 6 to 4 Â 10 6 cells per mouse) intravenously (i.v.) into B10BR female mice. Two days later, 3 Â 10 6 to 4 Â 10 6 CD11b 1 Gr1 1int/low cells purified from spleen of C3H Nef Tg or non-Tg mice, as described above, were injected i.v. Within 1 to 2 h after the latter injection, an equal number of Pcc-loaded DC were injected subcutaneously into recipient mice to stimulate donor AD10 CD4 1 T cells. The latter DC were prepared from BM maturated with GM-CSF and IL-4 and activated with LPS, as described previously (16), and further incubated with Pcc for 2 h (250 mg/ml). Recipient mice were sacrificed after 3 to 4 days, and proliferation of their T cells was analyzed by FACS.
FL cell transplantation. The FL procedure was previously described (34). Western blot analysis and IVKA. Immunoblot analysis was performed according to a modified procedure reported previously (90). Proteins were visualized with Alexa 680-coupled secondary Ab on the Odyssey imaging system (Li-Cor).
RT-PCR for HIV-1 Nef. The reverse transcription-PCR (RT-PCR) assays for quantitation of spliced transgene-specific HIV RNA was performed as described previously (90), using forward primer 377 (59GCACGGCAAGAGGCGAGGG) at nucleotide 715 of HIV-NL4-3 and reverse primer 1494 (CTAATCGA ATGGATCTGTCTCTG) at nucleotide 8460. For quantitative RT-PCR, the same HIV primers were used along with primers for S16 internal standard (forward primer 2432, 59AGGAGCGATTTGCTGGTGTGG; and reverse primer 2433, 59GCTACCAGGGCCTTTGAGATG). The reaction mixture contained 5 ml SYBR select master mix (Life Technologies), 0.5 ml of each primer, at 10 pmol/liter and 2 ml template DNA (cDNA) in 2.0 ml RNase-free H 2 O. Amplification was for 40 cycles (15 s at 95°C and 1 min at 60°C) with an AB Applied Biosystems ViiA7 instrument (Life Technologies).
Statistical analysis. Statistical analysis was performed in most cases using the unpaired two-tailed Student's t test. In cases of comparison of ratios, a one-sample t test was used. One or two-way analysis of variance (ANOVA) with Bonferroni's correction, where appropriate, was used for comparison of multiple samples. Spearman's test was used to evaluate correlation of IL-17/G-CSF and G-CSF/CD11b 1 Gr1 1 cells.

ACKNOWLEGMENTS
This work was supported by grants from the Canadian Institute of Health Research and the National Heart, Lung, and Blood Institute (NIH) (USA) to P.J.