Neutropenia with impaired host defense against microbial infection in mice lacking androgen receptor

Because male infertility is observed in testicular feminization mutant (Tfm) mice, which carry an unstable truncated AR mutant (33), it is impossible to generate female homologous AR deletion mice by conventional gene disruption methods. To solve this difficulty, we used the Cre–loxP technique to generate a mouse model with the targeted disruption of AR (ARKO) (12). We successfully generated ARKO mice of both genders and WT littermates, and further used these mice to investigate if loss of AR influences neutrophil production. In ARKO mice, the deletion of exon 2 in AR (Fig. 1 A) results in a splice between exon 1 and exon 3, creating two sequential nonsense mutations at new amino acid positions 533 and 534, leading to premature termination of the AR protein. RT-PCR analyses of total RNA extracted from isolated bone marrow neutrophils (Fig. 1 B) showed that the mRNA from ARKO neutrophils lacked the 305-bp fragment from exon 2 of AR (Fig. 1 C), evidence that the intact AR is not expressed in ARKO neutrophils and precursors. Intracellular analysis of endogenous AR protein by flow cytometry further confirmed that AR protein expression occurs in WT, but not ARKO, bone marrow neutrophils (Fig. 1 D).

We then examined circulating neutrophils in ARKO and WT mice by FACS analyses. On the basis of forward scatter (FSC) and side scatter (SSC), FACS analyses of peripheral blood leukocytes showed that a high-SSC population (i.e., a granulocyte population) was barely detectable in ARKO mice (Fig. 1 E). Using FACS analysis with neutrophil-specific monoclonal antibodies (clone 7/4) and anti-CD11b antibodies, we found that only 3% of leukocytes were neutrophils in ARKO mice, in contrast to 26% neutrophils in WT mice (Fig. 1 F). Neutrophils were rarely found in cytospin smears of blood leukocytes from ARKO mice (Fig. 1 G). As shown by examination of blood smears with Wright-Giemsa staining, male and female ARKO mice have normal white blood cell counts, but very low percentages of neutrophils in peripheral blood (Fig. 1 H). Examining blood smears from Tfm mice (Fig. 1 H), we confirmed that loss of AR results in a significantly reduced neutrophil abundance in peripheral leukocytes. Automated blood analysis revealed that the numbers of total white blood cells, red blood cells, and platelets remained comparable in ARKO mice (Table I A). The mean differential ratio of peripheral lymphocytes in ARKO mice was increased from 75 to 94% without obvious alterations in monocyte and eosinophil numbers (Table I B). However, there was a significant reduction of neutrophils in blood compared with WT mice (Table I B). These results suggest that the AR might play a critical role in sustaining a normal number of circulating neutrophils in vivo.

We then sought to determine if the neutropenia in ARKO mice was caused by a defect in granulopoiesis in the bone marrow. We found that the high SSC granulocyte population in bone marrow was markedly decreased in ARKO mice (Fig. 2 A). Further analysis of the granulocyte population with clone 7/4 and CD11b antibodies demonstrated a 50% decrease of neutrophils in bone marrow of ARKO mice compared with those in WT littermates (Fig. 2 B). These results demonstrated that loss of AR results in significant reduction of neutrophils in bone marrow.

To learn if both androgen and AR are required for neutrophil production, we examined the effects of castration and androgen supplementation on neutrophil counts in WT, ARKO, and Tfm mice. We found that castration of normal mice only results in a relatively moderate reduction of neutrophils as compared with neutropenia in ARKO and Tfm mice (Fig. 2, C–E), and supplement of exogenous dihydrotestosterone (DHT) restores neutrophils in castrated mice, but not in ARKO and Tfm mice (Fig. 2, C–E). These results suggest that AR, but not androgen, is required for neutrophil homeostasis.

We also compiled blood cell counts from 33 advanced prostate cancer patients before and after the surgical castration therapy. Those patients had no prior history of chemotherapy and radiation therapy, and had no sign of infection (WBC counts were <10,000 per μl). The results showed that their neutrophil differential counts only moderately decreased after surgical castration therapy (Fig. 2 F). The decrease of neutrophil counts in those patients with surgical castration was not as significant as that observed in ARKO mice, further supporting that androgen is less important than AR in neutrophil homeostasis.

We next asked which stages during granulopoiesis in bone marrow are responsible for neutropenia in ARKO mice. Using colony-forming assays, we discovered that adding G-CSF, GM-CSF, macrophage-colony stimulating factor (M-CSF), IL-3, and IL-6 induced similar levels of colony formation in ARKO bone marrow cells and WT cells in low-androgen methylcellulose medium (Fig. 3 A). Adding DHT did not influence colony numbers induced by G-CSF (Fig. 3 B). These results suggest that the populations of early granulocytic and monocytic progenitor cells are comparable in WT and ARKO bone marrow cells. Thus, we conclude that neutrophil lineage commitments in ARKO mice must be similar to those in WT mice. In addition, using FACS analyses, we found that the abundance of both common myeloid progenitors (CMPs) and granulocyte/macrophage progenitors (GMPs) in bone marrow is also comparable between WT and ARKO mice (Fig. 3 C). Using Wright-Giemsa staining of bone marrow smears, we found that fewer isolated bone marrow neutrophils from ARKO mice than from WT mice had segmented nuclei (Fig. 3 D). Differential counting of bone marrow cells showed that ARKO mice had a decrease in myelocytes/metamyelocytes and a marked decrease in mature neutrophils as compared with bone marrow cells from WT mice (Table I C). Although we observed that cancellous bone volumes of ARKO male mice are reduced compared with WT littermates (12), there is no difference in bone marrow cellularity (Table I C). These results suggest that the defect in granulopoiesis of ARKO mice occurs during the transition between proliferation of precursors (myeloblasts, promyelocytes, and myelocytes) and maturation of neutrophils (metamyelocytes, band forms, and neutrophils), leading to a defect in terminal differentiation of neutrophils.

To see if restoration of AR rescues neutrophil differentiation, we infected FACS-sorted ARKO GMPs with retroviruses carrying the cDNA for AR and GFP or retroviruses carrying GFP alone, and then seeded infected cells into methylcellulose medium in the presence of G-CSF. Fluorescent colonies were harvested after 7 d. ARKO colonies infected with retroviruses carrying GFP alone exhibited fewer mature neutrophils (Fig. 3 E). In contrast, colonies from ARKO GMPs infected with retroviruses expressing functional AR consisted of more mature neutrophils, suggesting that the defects in neutrophil maturation resulting from ablation of AR can be rescued by AR overexpression. Again, we found that adding androgen did not result in an additional rescue on neutrophil maturation (Fig. 3 E). Also, we infected WT GMPs with retroviruses carrying AR-small interfering (si)RNA to see if knockdown of endogenous AR retards neutrophil differentiation. We confirmed the AR-siRNA knockdown efficiency with reporter assays (Fig. S1) and immunoblot analyses of AR protein expression in cells infected with retroviruses carrying AR-siRNA and scramble siRNA control (Fig. S2). Colonies from sorted WT GMPs infected with retroviruses carrying AR-siRNA and GFP showed fewer mature neutrophils than those infected with retroviruses carrying scrambled siRNA and GFP or GFP alone (Fig. 3 E). These results are consistent with the ARKO mouse data, suggesting that knockdown of endogenous AR by siRNA in GMPs could retard neutrophil maturation. Results from these complementary methods of AR manipulation in cultured cells are consistent with our in vivo ARKO mice data showing that loss of AR results in retardation of neutrophil differentiation.

We next examined granulocytic proliferation via analysis of the cell cycling status of granulocytes within bone marrow, by labeling cells with anti–Gr-1 antibodies and propidium iodide. The proportion of cells in S and G2/M phase was lower in Gr-1⁺ bone marrow cells from ARKO mice (Fig. 4 A), suggesting that the rate of granulocytic proliferation in bone marrow from ARKO mice is lower than in WT mice.

To learn more about the dynamics of neutrophil production in WT and ARKO mice, we labeled mitotic progenitors/precursors in vivo by intraperitoneal injection of BrdU, which only labels proliferating cells. Cells derived from the neutrophil precursor pool can be labeled with BrdU because these cells are proliferating and have matured into granulocytes, which are then released from the bone marrow into the circulation. At 96 h after the injection, the BrdU-labeled granulocytes reached a maximum in the peripheral blood (Fig. 4 B). It is known that BrdU labeling intensity is diluted by half with every cell division after BrdU incorporation. Consequently, by tracking the changes in relative BrdU intensity, we can determine how many division cycles the neutrophils from precusor cells have undergone. Cells in the BrdU^bright region correspond to granulocytes that incorporated BrdU in their last mitosis, whereas those in the BrdU^dim region have gone through one or more divisions after BrdU incorporation. As shown in Fig. 4 C, 40% of neutrophils were in the BrdU^dim pool in WT mice, whereas only 22% of neutrophils from ARKO mice were BrdU^dim, suggesting a markedly reduced mitotic cell division of neutrophil precursors in bone marrow of ARKO mice.

To determine if neutropenia in ARKO mice is partly caused by neutrophil apoptosis, we examined the bone marrow cell apoptosis using a FACS-based assay with Annexin V and 7-amino-actinomycin-D (AAD) staining. We did not observe a significant number of apoptotic Gr-1⁺ cells in fresh bone marrow from either WT or ARKO mice (Fig. 5 A), which might be caused by rapid clearing of apoptotic cells by macrophages in vivo. We then examined survival of mature bone marrow neutrophils in vitro and found that 25% of ARKO neutrophils were apoptotic cells as compared with 19% in WT cells in culture without G-CSF treatment, indicating that ARKO neutrophils are moderately more susceptible to spontaneous apoptosis than WT cells (Fig. 5 B). On the other hand, with G-CSF treatment for 48 h, only 10% of WT neutrophils, but 20% of ARKO neutrophils, remained apoptotic (Fig. 5 B). These results suggest that G-CSF is unable to delay apoptosis in ARKO neutrophils.

Together, the proliferation and apoptosis assays suggest that neutropenia resulting from loss of AR might be caused by decreased proliferation of neutrophil precursors and increased apoptosis of neutrophils in bone marrow.

G-CSF–mediated apoptosis protection might be less effective in ARKO granulocytes because of impaired response to G-CSF. To investigate the granulopoietic response to G-CSF in WT and ARKO littermates, we monitored blood neutrophil counts with daily injections of G-CSF for 9 d (Fig. 5 C). We observed that ARKO mice had lower increasing rate of blood neutrophils in response to G-CSF as compared with WT mice. These results suggest that there might be an inhibition to G-CSF signaling in AR-deficit myeloid cells. Using real-time quantitative PCR (Q-PCR), we measured the mRNA expression of G-CSF–downstream target genes, such as SOCS3, flotillin, chemokine (C-C motif) receptor-like 2, glycerol kinase, and complement component 3a receptor 1, and found their mRNA expressions were decreased in ARKO granulocytes (Fig. 5 D). Protein expression of SOCS3, a major downstream target of G-CSF, was also significantly reduced (Fig. 5 E) in isolated bone marrow granulocytes from ARKO mice compared with those from WT littermates. These results suggest that G-CSF signaling could be suppressed in ARKO granulocytes. We then examined the expression of G-CSFR using semiquantitative RT-PCR and a flow cytometric–based G-CSF binding assay. We saw no difference in mRNA and surface protein expression of G-CSFR between granulocytes from ARKO and WT mice (Fig. 5, F and G).

To investigate if downstream mediators of G-CSF signaling are suppressed by AR deficiency, we examined the expression of Stat3, a proximal mediator of the G-CSF signaling (34, 35). We found that the expression of both total Stat3 and the tyrosine-phosphorylated form of Stat3 (pY-Stat3) were similar between isolated bone marrow neutrophils from WT and those from ARKO mice in the absence or presence of G-CSF stimulation (Fig. 6 A). Similar levels of Stat3 and pY-Stat3 proteins were also observed in bone marrow neutrophils from castrated mice and those from WT and ARKO in the presence of G-CSF (Fig. 6 B). These results suggest that loss of AR does not influence the expression or the function of G-CSFR. Interestingly, we found that endogenous AR of granulocytes from castrated mice is less abundant (Fig. 6 B), which could be caused by the decrease of AR protein stability in the prolonged absence of androgen (36, 37). This is further supported by the finding that castration resulted in a significant reduction of AR expression and that exogenous DHT administration restored the reduced AR expression (38).

Because early studies have demonstrated that PIAS3 could inhibit Stat3 signaling via blocking Stat3 DNA binding (39), we sought to explore if AR interacts with PIAS3 and affects its inhibition on Stat3 activity. The glutathione S-transferase (GST) pull-down data shows that there is a direct protein–protein interaction between PIAS3 and AR (Fig. 6 C). The PIAS3 interaction with AR possibly occurs through the PIAS3 N terminus, which contains a LXXLL motif, an α-helical protein interaction module that mediates interactions between AR coregulators and AR. Using a coimmunoprecipitation assay, we found that anti-PIAS3 antibodies can precipitate more pY-Stat3 in granulocytes from ARKO, followed by that from castrated mice, which is more than that from WT mice (Fig. 6 D). This result suggests that loss of AR in ARKO granulocytes results in increased PIAS3-mediated sequestering of pY-Stat3, leading to the suppression of G-CSF signaling. A reporter assay further showed that suppression of constitutively active Stat3 (CA-Stat3) transactivation by PIAS3 was relieved by AR in a dose-dependent manner (Fig. 6 E). Notably, the major induction of Stat3 reporter is through increasing AR and is only slightly enhanced by adding androgen (Fig. 6 E), suggesting that androgen is not required for the AR's rescue of stat3 activity from the inhibition by PIAS3.

Stat3 activation has been demonstrated to promote differentiation of myeloid cell lines (34, 35, 40–43). To further investigate if increased activated Stat3 can rescue differentiation of ARKO myeloid progenitors, we infected primary bone marrow GMPs with retroviruses carrying CA-Stat3, dominant-negative Stat3 (DN-Stat3), or retroviruses carrying pMSCV vector alone. We seeded infected cells into methylcellulose medium in the presence of G-CSF, and harvested colonies after 7 d. WT and ARKO colonies infected with retroviruses carrying DN-Stat3 exhibited fewer mature neutrophils (Fig. 6 F). In contrast, WT and ARKO colonies infected with retroviruses expressing CA-Stat3 contained more mature neutrophils (Fig. 6 F). These results suggest that blocking endogenous Stat3 function by overexpression of DN-Stat3 could retard neutrophil differentiation, whereas increase of CA-Stat3 could rescue neutrophil differentiation.

To further explore the relationship between AR and PIAS3 in granulocytic differentiation, we examined differential counts of neutrophils in colonies infected with PIAS3 and PIAS3-siRNA. We infected primary WT and ARKO GMPs with retroviruses carrying PIAS3, PIAS3-siRNA, and controls. We then seeded infected cells into methylcellulose medium with G-CSF for 7 d, and then examined differential counts of fluorescent colony cells. We found PIAS3 overexpression reduced mature neutrophils, but increased AR could rescue the reduction (Fig. 6 G). We also found that PIAS3 knockdown by PIAS3-siRNA could restore mature neutrophil numbers in ARKO colonies (Fig. 6 H).

The binding of G-CSF to its receptor activates not only the Jak–Stat3 pathway (44) but also the Ras–Raf–ERK signaling cascade (45). No ERK1/2 activation was found in WT and ARKO neutrophils before G-CSF treatment (Fig. 7 A). We observed that ERK1/2 phosphorylation was significantly decreased in ARKO bone marrow granulocytes with G-CSF treatment (Fig. 7 A). We then found that in the presence of G-CSF, there is a decrease in phosphorylated ERK1/2 in WT bone marrow cells carrying AR-siRNA, whereas exogenous AR promotes ERK1/2 activation in ARKO bone marrow cells (Fig. 7 B). Next, we examined the effects of AR deficiency on the G-CSF–induced proliferation of primary bone marrow cells and the human myeloblastic cell line KG-1. Compared with ARKO cells, WT bone marrow cells showed a significantly higher proliferative activity in response to G-CSF (Fig. 7 C). Similarly, under G-CSF treatment, WT KG-1 cells also showed higher cell proliferation than the KG-1 clone carrying hAR-siRNA (Fig. 7 D). Furthermore, we found that U0126 (MEK kinase inhibitor) abolished a large part of G-CSF-induced cell proliferation for those cells containing AR (Fig. 7 C and D). All of these results suggest that AR is essential for the G-CSF–induced ERK activation and subsequent granulocytic proliferation. It has been reported that AR is important for the activation of c-Src–Ras–Raf–ERK signaling cascade through the interaction between the proline-rich motif of AR and the SH3 domain of c-Src (11). Also, our previous studies have demonstrated that AR contributes to ERK activation in an androgen-independent manner (13).

Together, these results indicate that AR not only releases the PIAS3 inhibition of the Stat3 but also promotes ERK activation in G-CSF signaling (Fig. 7 E).

We then asked if any functions of neutrophils derived from ARKO mice were impaired as compared with WT controls. First, we examined the oxidative burst and phagocytic activities of granulocytes derived from ARKO and WT mice by incubating granulocytes with fluorescein-labeled Escherichia coli at 37°C (Fig. 8 A). In addition, we assessed the oxidative burst in response to PMA, fMLP, or anaphylatoxin C5a stimulation by monitoring the production of H₂O₂, which converts 2′,7′-dichlorofluorescein acetate, a nonfluorescent precursor, to 2′,7′-dichlorofluorescein. The fluorescence intensity of 2′,7′-dichlorofluorescein was comparable in the WT and ARKO neutrophils as stimulated by PMA (Fig. 8 B), fMLP (Fig. S3), or C5a (Fig. S3), showing that the oxidative burst capacity of the two genotypes was similar. These results suggest that both phagocytosis and the oxidative burst are normal in granulocytes from ARKO mice.

Chemokines and cytokines are important for leukocyte migration and inflammatory response. To assess the expression of chemokines and proinflammatory cytokines in the current study, we used Q-PCR to measure the mRNA levels in WT and ARKO granulocytes. ARKO cells had lower levels of IL-1β, IL-6, TNF-α mRNA, and several chemokines than WT cells (Fig. 8 C), suggesting that reduced production of chemokines and proinflammatory cytokines might influence the functions of neutrophils or other leukocytes through autocrine or paracrine pathways.

The chemokine receptor CXCR2 plays an important role in neutrophil migration (46). Neutrophil chemoattractants macrophage-inflammatory protein-2 (MIP-2) and KC/CXCL1 can specifically bind to CXCR2, whereas fMLP signals through a different G protein–coupled receptor (47–50). To test the migration capability of ARKO neutrophils, we examined the chemotactic response toward MIP-2, KC, and fMLP. As shown in Fig. 8 D, ARKO neutrophils demonstrated reduced migration activities toward the chemoattractants MIP-2 and KC compared with WT cells; however, ARKO and WT neutrophils had similar responses to fMLP. It has been shown that lack of Stat3 can impair CXCR2-dependent neutrophil migration (51), and our results are correlated with that finding because the Stat3 activity is mostly inhibited in ARKO neutrophils.

We then examined the release of β-glucuronidase and lactoferrin as the markers of primary and secondary granules from isolated mature neutrophils after stimulation. We found that WT and ARKO neutrophils showed no differences in the release of β-glucuronidase or lactoferrin induced by fMLP, C5a, and PMA (Fig. 8 E).

To explore if neutropenic ARKO mice have impaired host resistance to acute bacterial infection, we injected 9-wk-old ARKO mice and age-matched WT controls (10 animals per group) intraperitoneally with a virulent strain of E. coli. Survival of the mice was monitored over 60 h. At 12 h after injection, all 10 ARKO mice showed signs of severe infection, including diarrhea, hunchback, imbalance, ruffled fur, and eye infection, but only 3 out of 10 WT mice exhibited signs of infection (Fig. 8 F). Moreover, all 10 ARKO mice were dead within 36 h, and only 1 WT mouse died within that time frame (Fig. 8 G). All remaining WT mice were still alive at 60 h. The inability of the ARKO mice to survive the E. coli challenge shows that these animals are more susceptible than WT controls to bacterial infection.

It is interesting that neutropenia was observed in ARKO and Tfm mice, whereas castrated male mice only have moderate reduction of neutrophil counts in both blood and bone marrow (Fig. 2, C–E), suggesting that AR, but not androgen, is critical for neutrophil homeostasis. Other evidences to support our hypothesis include the following: (a) Exogenous androgen can restore neutrophil counts in castrated WT mice, but not ARKO mice (Fig. 2, C–E). (2) Female ARKO mice have neutropenia, but WT female mice, with 10-fold less androgen levels compared with male, still maintain normal neutrophil counts (Fig. 1 H). (3) Our mechanistic studies indicate that AR reduces the inhibition of PIAS3 to Stat3 transcriptional activity, and adding androgen only mildly enhances the AR effect on Stat3 signaling (Fig. 6 E). (4) Without adding androgen, AR overexpression still increases granulocytic differentiation in colony formation assays (Fig. 3 E). With these observations, we believe that AR confers a direct and more profound effect than androgen. We note with interest that neutropenia resulting from AR ablation contrasts with the effect of knocking out the estrogen receptor β. The estrogen receptor β KO mice show an overproduction of mature granulocytes (52). These opposing effects on granulopoiesis suggest that both AR and ER are required to maintain neutrophil homeostasis.

In vitro and transgenic mouse studies have demonstrated that Stat3 plays an essential role in G-CSF–dependent granulopoiesis (35, 51, 53–56). Moreover, several reports have shown that SOCS3 KO mice, in which Stat3 activity is persistent, developed severe neutrophilia (57–59). Paradoxically, mice with targeted deletion of the Stat3 gene in hematopoietic cells were also found to develop neutrophilia (60–63). Nevertheless, ERK1/2 hyperphosphorylation was found in bone marrow cells of Stat3 conditional KO mice, and it has been shown to contribute to neutrophilia in those animals (63). It is possible that the ERK1/2 hyperactivation results from SHP–Grb2–Src binding to available Stat3 docking sites on G-CSF receptor and overactivating the Ras–Raf–ERK signaling cascade. This possible signal crossover has been suggested in some reports showing that G-CSFR truncated mutants lacking the receptor C terminus, which contains tyrosine Y764 (the default docking site for SHP-2/Grb2), still have the ability to activate p21Ras via the SHP/Grb2 route linked to Y704 (the stat3 docking site) of the G-CSFR (41, 64). Clearly, further investigation is still required to determine the precise role of Stat3 in neutrophil development.

Several KO models have been established for genes involved in the regulation of myelopoiesis. C/EBPα^−/− mice have circulating myeloblast-like cells but lack mature neutrophils in vivo at birth. Although it was found that a lower PU.1:C/EBPα ratio directs granulocytic differentiation over macrophage differentiation in myeloid cell fate determination, a PU.1^−/− mice study demonstrated that PU.1 is not essential for neutrophil lineage commitment, but is necessary for normal development, maturation, and function of neutrophils. Gfi^−/− mice are severely neutropenic and accumulate immature monocytic cells in blood and bone marrow. We found the mRNA levels of PU.1, C/EBPα, and Gfi in ARKO bone marrow neutrophils are comparable to those detected in WT neutrophils (unpublished data), indicating that C/EBPα, PU.1, and Gfi might not account for the granulopoietic defect observed in ARKO mice. Notably, a few mature neutrophils can still be observed in G-CSF^−/−, G-CSFR^−/−, and G-CSF^−/−GM-CSF^−/−M-CSF^−/− mice (5, 65, 66), suggesting that a myeloid-CSF independent mechanism exists in basal granulopoiesis. It might be interesting to further study whether AR also modulates a myeloid–CSF-independent pathway.

Androgens can stimulate erythropoiesis and have been used clinically for treatment of aplastic anemia (26, 27). However, our data (Table I A) indicate that red blood cell numbers and hemoglobin levels in WT and ARKO mice are similar, suggesting that AR is not required for steady-state erythrocyte production under physiological conditions. These results, however, are at odds with early reports showing that androgens could stimulate erythropoiesis (67). The diverse effect on erythropoiesis by different androgen derivatives is indicative of the complexity of androgen control in erythropoiesis. 5α-DHT binds to the AR and might stimulate erythropoiesis indirectly by increasing levels of erythropoietin, whereas 5β epimers (5β-DHT and etiocholanolone) do not bind to the AR, yet effectively increase the number of erythropoietic progenitor cells, thereby also enhancing erythropoiesis (68–73). It will be interesting to study if androgen stimulates erythropoiesis through an AR-independent pathway.

In summary, the data presented here indicate AR is essential for steady-state granulopoiesis. Further studies that explore potential cross talk between AR and other myeloid regulators in granulopoiesis might help reveal a better understanding of how cell signaling networks tightly control neutrophil homeostasis.

Construction of targeting vectors and generation of the chimera founder mice have been previously described (12). The original strain background of the mosaic founder mice was C57BL6/129sv. We have generated floxed AR mice with a nearly pure C57BL/6 background by backcrossing the original C57BL6/129sv background floxed AR mice with pure C57BL/6 background mice for more than eight generations. β-Actin is constitutively and ubiquitously expressed; thus, the β-actin promoter-driven Cre (ACTB-Cre, FVB background; The Jackson Laboratory) will express and delete floxed AR fragments in all the tissues. WT and ARKO mice were genotyped by PCR, as described previously (12). Male Tfm and WT control mice (C57BL/6J-A-Ta[6J]^+/Y) were purchased from The Jackson Laboratory. All animal experiments were approved by the University Committee on Animal Resources of the University of Rochester.

Castration was performed under anesthesia using sterile instruments and gloves. A transverse scrotal incision was made, the testicles were exposed, the vas deferens and spermatic blood vessels were ligated, and the testicles were removed. A sham operation was performed for noncastrated controls in which the testis and epididymis were only pulled out and then returned to the scrotum. The scrotal incision was closed using a single skin clip. Animals were placed in a clean cage and monitored until fully recovered from anesthesia.

Androgen supplementation in castrated WT mice and ARKO littermates was achieved by subcutaneous implantation of 60-d time-release pellets (Innovative Research of America) containing DHT (5 mg per pellet); mice were sacrificed 14 d after androgen supplementation. The levels of plasma DHT were determined by a DHT ELISA Kit (Alpha Diagnostic International) according to the manufacturer's protocol.

Between 6 a.m. and 8 a.m., we collected peripheral blood through cardiac puncture using a heparinized 3-ml syringe with a 30-gauge needle on mice under anesthesia. We dissected the femur free from the muscle, and then cut both at the knee and hip for removal. We removed bone marrow from each femur by flushing with 1 ml RPMI 1640 (using a 26-gauge needle and a 1-ml syringe) into a single well of a 6-well culture plate.

Bone marrow smears, blood smears, and cytospin preparations were stained with Wright-Giemsa for morphological assessment and manual differential counting. Hematologic parameters from blood samples obtained from 8–12-wk-old mice were measured using an Abbot CELL-DYN 4000 system.

We isolated neutrophils from bone marrow or peripheral blood by using the EasySep selection kit (StemCell Technologies) according to the manufacturer's suggested procedures. Mature bone marrow neutrophils were isolated by discontinuous Percoll density gradient centrifugation.

1–5 × 10⁵ cells were incubated on ice for 30 min with saturating amounts of antibody. Cells were washed with ice-cold PBS containing 2% fetal bovine serum, and flow cytometric analyses were performed using a dual-laser FACSCalibur flow cytometer (BD). For analysis of white blood cell populations, we set gates to include those cells that were CD45⁺ and viable by FSC and SSC scan profiles. Neutrophils were stained by fluorochrome-conjugated, neutrophil-specific monoclonal antibodies (clone 7/4; Invitrogen) or anti–Gr-1 antibodies (clone RB6-8C5; eBioscience). For intracellular AR staining, after cells were incubated with fixative buffer (2% paraformaldehyde) and permeabilized with permeabilization solution (0.1% saponin in PBS containing 2% fetal bovine serum), intracellular AR was stained by anti-AR antibody (clone C-19; Santa Cruz Biotechnology, Inc.), followed by fluorochrome-conjugated anti–rabbit IgG antibodies. Data were collected and analyzed from at least 10,000 gated events using CellQuest software (BD). To assess surface G-CSFR expression, G-CSF was biotinylated. Bone marrow Gr-1⁺ cells were incubated at 48°C for 1 h with biotinylated G-CSF (25 ng per 10⁶ cells) in the presence or absence of a 100-fold molar excess of nonlabeled G-CSF, followed by incubation with FITC-conjugated streptavidin. All cells were analyzed on a FACSCalibur flow cytometer.

We collaborated with the Department of Urology at the Sir Run Run Shaw Hospital to collect blood analysis data from each patient before and after surgical castration (0–2 wk before castration and 2–4 wk after castration). Appropriate data excluded any patients receiving radiation therapy and/or chemotherapy. Diurnal rhythms of neutrophil counts and possible increased neutrophil counts from smoking and infections were also prevented, and the results are statistically analyzed. We requested coded data from chart reviews that include only patient diagnosis, treatments, time of blood drawn, and blood analysis data.

The immunofluorescent staining of intracellular BrdU incorporation has been previously described (74). To study cycling status of granulocytes, we first stained bone marrow cells with FITC-Gr-1 antibody, fixed cells in 70% ethanol and removed residual RNA by RNase A, and then further stained them with propidium iodide (final concentration 5 µg/ml) for analysis by a FACSCalibur flow cytometer. Data were further analyzed using CellQuest Software (BD) for DNA histogram analysis of Gr-1⁺ cells. For neutrophil apoptosis assays, bone marrow cells were stained with Annexin-V and 7-AAD. High-SSC cell population was then gated for apoptosis analysis.

Single-cell suspensions of bone marrow (2.5 × 10⁴ cells) in methylcellulose medium (MethoCult M3234, Stem Cell Technologies) were plated in 35-mm Petri dishes in triplicate and incubated at 37°C. Purified recombinant murine (rm)G-CSF (10 ng/ml), rmGM-CSF (10 ng/ml), rmM-CSF (10 ng/ml), rmIL-3 (10 ng/ml), and rmIL-6 (400 ng/ml) (R&D Systems) were used.

Neutrophils were isolated by EasySep selection kit (Stem Cell Technologies) and total RNA was extracted from cells using TRIzol reagent (Invitrogen). We performed reverse transcription using Superscript II reverse transcription and oligo-dT primers (Invitrogen) on 250 ng of total RNA. In semiquantitative RT-PCR, cDNA was diluted 5, 25, 50, and 100 times, and an aliquot (2 µl) of the reverse transcription reaction was subjected to PCR in a thermocycler (MyCycler; BioRad Laboratories). β-Actin was used as an internal control. Q-PCR was performed in a final volume of 25 µl containing 1 µg cDNA, 12.5 µl SYBR Green Mix (2X), 6 pmol of each primer set. The primer sequences for Q-PCR are listed in Table S1. The reaction mixtures were incubated on a iCycler (BioRad Laboratories) real-time PCR amplifier (BioRad Laboratories) programmed at 50°C, 2 min; 95°C 10 min; then 40 cycles for three steps at 95°C, 15 s to 60°C, 30 s to 72°C, 30 s; and finally at 70°C for 10 min. HPRT mRNA was used as the internal control. Real-time Q-PCR was performed by determining the cycle at which the abundance of the accumulated PCR product crosses a specific threshold, the threshold cycle (CT) in each reaction. The difference in average CT values between the internal control and a specific gene was calculated for each individual and termed ΔCT, which is comparable to the log-transformed, normalized mRNA abundance. For each gene, at least three samples with triplicate amplifications were performed and mean ± SD taken for data analysis.

The retroviral vectors pBabe-AR and pBabe-PIAS3 were constructed by ligating digested PCR products of AR and PIAS3 cDNA into a pBabe vector containing the SV40 promoter-driven GFP. To generate vectors for siRNA expression, the pSuperior-GFP-neo vector (OligoEngine) was digested and the annealed oligonucleotides were ligated into the vector. The nucleotide sequences for targeting mouse AR, PIAS3, and Stat3 were designed as 5′-TCTAGCCTCAATGAGCTTG-3′, 5′-GCCACTGCCCTTCTATGAA-3′, and 5′-AGCCGATCTAGGCAGATG-3′, respectively. Scrambled control shRNA unrelated to AR, PIAS3, and Stat3 sequences were used as negative controls. hAR-siRNA was described previously (13).

The myeloid-committed precursor populations from murine bone marrow were isolated by first depleting lineage marker–positive cells by means of immunomagnetic beads. The remaining cells were then stained with goat anti–rat immunoglobulin Texas red and anti–c-kit–APC and then enrichment sorted for Lin⁻ c-kit⁺ cells. The enrichment-sorted cells were then stained with anti-FcRII/III (CD16/32)–FITC, anti–c-kit–APC, anti–IL-7R–Alexa Fluor 594, anti–Sca-1–Alexa Fluor 594, and anti–CD34-biotin, followed by PE-avidin. The GMP precursor population was sorted as Lin⁻ CD16/32⁺ CD34⁺ c-kit⁺ Sca-1⁻ IL-7R⁻ cells.

Phoenix ectrophic cells were grown in DME supplemented with 10% FCS. Cells were typically seeded in 60-mm-diam plates at a density of 2.4 × 10⁶ cells per plate and allowed to settle for 18 to 24 h. Cells were transfected with 5–10 µg of plasmid DNA using calcium phosphate transfection. The media containing the plasmid DNA precipitate were replaced with fresh media 16 h after transfection. Retroviral supernatants were collected 48 h after transfection from Phoenix cells transfected with the retroviral vector–based constructs and used to infect freshly sorted GMP precursor cells. Infections were performed with 1 ml of undiluted supernatant in 2 ml of DME plus 400 µg of Polybrene per ml at 37°C. After 4 h of incubation, 2 ml of DME with 10% FCS were added and incubated for an additional 20 h.

After retroviral infection, single-cell suspensions of GMP precursor cells in methylcellulose medium (MethoCult M3234; StemCell Technologies Inc.) were plated in 35-mm culture dishes in triplicate and incubated at 37°C in the presence of rm G-CSF (10 ng/ml) in a humidified incubator maintained at 5% CO₂. After 7 d in culture, the numbers of green fluorescent colonies in each dish were counted, and green fluorescent colony cells were collected in suspension for cell counting. Cells were cytocentrifuged onto a microscope slide and stained with Wright-Giemsa for morphological assessment and 200-count manual differential counting.

We injected three pairs of WT and ARKO littermates subcutaneously with rmG-CSF at 40 ng/g mice from day 0–8, and then examined blood neutrophil counts 4 h after G-CSF injection at the indicated days.

CV-1 cells were transfected with 4 × IRF-LUC (1.5 µg) together with CA-Stat3 (0.5 µg), flag-PIAS3 expression construct (0.5 or 1.0 µg), and increasing amounts of AR (0.5 or 1.0 µg) with plasmid amount balance. The media was changed 2 h after transfection, and cells were cultured in media containing 10% charcoal depleted-FBS for 24 h, followed by treatment with or without 10 nM DHT for another 16–18 h. 5 ng phRL-TK per well was used as internal control. Cells were harvested, and the Luc activity was analyzed using Dual-Luc Reporter Assay System (Promega).

We infected WT and ARKO GMPs or bone marrow cells with retroviruses carrying pBabe-AR, AR-siRNA, or controls. After retroviral infection, single-cell suspensions of bone marrow cells in methylcellulose medium were plated in culture dishes and incubated at 37°C in the presence of G-CSF (10 ng/ml). After 7 to 10 d in culture, colony cells were collected and further incubated with or without G-CSF (40 ng/ml) for 30 min, and cells were harvested for immunoblot analysis.

Cells were lysed in lysis buffer (50 mM Tris–HCl, pH 7.4, 0.3 M NaCl containing 1% NP-40, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and 10 µg/ml each of aprotinin, pepstatin, and leupeptin). Cell lysates were immunoprecipitated with anti-PIAS3 antibody as indicated. The immunoprecipitates from cell lysates were resolved on 5–20% SDS–PAGE and transferred to Immobilon membrane (Millipore). The membranes were then probed with each antibody as indicated. Immunoreactive proteins were visualized using an enhanced chemiluminescence detection system (GE Healthcare) and analyzed using Quantity One software (BioRad Laboratories).

GST-PIAS3 fusion protein and GST protein were obtained by transforming expressing plasmids into BL21 (DE3) pLysS strain-competent cells followed with 1 mM isopropyl-D-thiogalactoside induction. GST fusion proteins were then purified by glutathione–Sepharose 4B as instructed by the manufacturer (GE Healthcare). In vitro translated [³⁵S]methionine-labeled proteins (2 µl) generated with the TNT-coupled reticulocyte lysate system (Promega) were mixed with the glutathione–Sepharose-bound GST proteins at 4°C for 3 h to perform the pull-down assay. The bound proteins were separated on a 10% or 15% SDS-polyacrylamide gel and visualized by PhosphorImager (GE Healthcare).

In the presence or absence of G-CSF (10 ng/ml), bone marrow mononuclear cells (10⁶/ml) were incubated in RPMI with or without U0126 (20 µM). After 3 d, 0.5 µCi ³H-thymidine was added and the cells were cultured for an additional 8 h. Proliferative activity was then determined by ³H-thymidine incorporation.

Heparinized blood was incubated with opsonized fluorescein-labeled E. coli (Invitrogen) either at 37°C or at 0°C (control) for 10 min. After erythrocyte lysis, cells were washed twice and then stained with CyChrome–anti-Gr-1 antibodies. Oxidative burst in response to PMA was evaluated by detecting the intracellular conversion of dichlorofluorescein diacetate (DCFH-DA) to 2′,7′-dichlorofluorescein by flow cytometry. Leukocytes were stained with CyChrome–anti-Gr-1 antibodies and cells were mixed with 10 µM DCFH-DA and 10 nM PMA for 5 min at 37°C. We examined intracellular 2′,7′ dichlorofluorescein content in WT and ARKO Gr-1⁺ cells. 10,000 gated Gr-1⁺ cells were analyzed using CellQuest software. We isolated fresh mature bone marrow neutrophils by discontinuous Percoll density gradient centrifugation to perform oxidative burst assays in response to fMLP (1 µM) or C5a (1 µg/ml).

Fresh WT and ARKO mature bone marrow neutrophils were obtained by discontinuous Percoll density gradient centrifugation. Neutrophils (5 × 10⁵ cells) were then plated in 3 µm Transwells (Corning) with the indicated chemoattractant (200 ng/ml MIP-2, 1 µg/ml KC, or 0.1 mM fMLP) in the lower chamber. The number of cells migrated through the Transwell was determined by cell counting after incubation for 2 h at 37°C. Over 90% neutrophils remained viable at the end of the assay as judged by trypan blue exclusion.

Cells (10⁶/ml) were preincubated with 5 µg/ml cytochalasin B for 10 min and then stimulated with 1 µM fMLP, 0.5 µg/ml C5a, or 200 nM PMA for 20 min at 37°C. Reactions were stopped by cooling on ice and cell suspensions were centrifuged. β-Glucuronidase activity in supernatants of activated cells was measured spectrophotometrically using 4-methylumbelliferyl-β-d-glucuronide as substrate (75). β-Glucuronidase release is presented as the percentage of total cellular content. The secondary granule marker lactoferrin was determined by ELISA using anti-lactoferrin antibody (Sigma-Aldrich).

10 WT and 10 ARKO mice were administered 5 × 10⁵ colony forming units of E. coli (Serotype O18ac:K1:H7; American Type Culture Collection) by intraperitoneal injection and were followed for survival over 60 h.

Data are presented as the mean ± SD from at least three experiments, and statistical analysis was performed by either two-tailed Student's t test or one-way analysis of variance. The statistical significance of differences was set at P < 0.05.

Fig. S1 shows AR activity is knocked down by AR siRNA and restored by AR cDNA using retroviral infection in CV-1 cells. Fig. S2 shows AR-targeted disruption in colony cells derived from GMPs infected with retrovirus carrying AR-siRNA. Fig. S3 demonstrates that the oxidative burst induced by fMLP or C5a in WT and ARKO neutrophils was comparable. Table S1 shows primer sequences for Q-PCR analysis.

The authors are grateful to Karen Wolf and Leroy Hanchett for manuscript preparation and thank Marshall Lichtman, James Palis, Stephen Welle, and Alexandra Livingstone for critical review. We thank Ke Shuai and Daniel Link for sharing PIAS3, 4 × IRF-Luc, pMSCV-Stat3C (CA-Stat3), and pMSCV-Stat3D (DN-Stat3) plasmids. This work was supported by National Institutes of Health Grant DK073414 and the George Whipple Professorship Endorsement.

The authors have no conflicting financial interests.