Features of Macrophage Differentiation Induced by p19^INK4d, a Specific Inhibitor of Cyclin D–Dependent Kinases

MAINTENANCE OF CELLS in a proliferative state requires the activities of cyclin-dependent kinases (CDKs) that regulate progression through distinct cell cycle transitions. Best characterized is the oscillatory activation of the holoenzyme composed of cyclin B (the regulatory subunit) and CDK1 (the catalytic subunit, also known as CDC2 in the yeast Schizosaccharomyces pombe and CDC28 in Saccharomyces cerevisiae ), whose abrupt activation and subsequent degradation are universally required for mitotic entry and exit, respectively, in all eukaryotes.1 In mammalian cells, progression through the first gap phase (G₁ ) into the DNA synthetic (S) phase requires the activity of at least two distinct classes of G₁ cyclins (types D and E) and their associated catalytic partners. As quiescent cells enter the cycle, different D-type cyclins (D1, D2, and D3) are induced by various mitogens in a lineage-dependent manner and assemble under growth factor control with their catalytic subunits, CDK4 and CDK6.2 The activity of cyclin D–dependent kinases increases throughout G₁ phase as long as mitogen stimulation continues, and in mid to late G₁ , these enzymes catalyze the phosphorylation of the retinoblastoma protein (Rb), thereby helping to reverse its growth suppressive function.3,4 One consequence of Rb phosphorylation is the activation of a class of transcription factors (the E2Fs) that coordinately regulate genes whose expression is necessary for DNA synthesis.5,6 Among these is cyclin E, which, when complexed to CDK2, completes the inactivation of Rb and also triggers the onset of S phase by phosphorylating additional substrates.7-9 The timing of Rb phosphorylation in late G₁ phase occurs as cells pass a restriction point where they no longer require mitogens for subsequent progression through the cell cycle and, conversely, can no longer be inhibited by certain antiproliferative cytokines.10 11 The irreversibility of the restriction point transition is in part caused by the shift in the control of Rb phosphorylation from a mitogen (D-type)-dependent kinase to a mitogen-independent (E-type) one.

Inhibition of cyclin D–dependent kinases prevents cells from passing the restriction point, so although cells can progress through S, G₂ , and M phases without the activity of these enzymes, they accumulate in G₁ phase and may then exit the cycle into a quiescent (G₀ ) state.12-14 The simplest example of this process occurs upon mitogen withdrawal.15 Because the D-type cyclins are intrinsically unstable proteins, cells deprived of growth factors quickly lose cyclin D–dependent kinase activity and arrest in G₁ phase, usually within a single cycle. But even in the face of persistent mitogen stimulation, G₁ arrest can be induced by specific polypeptide inhibitors of CDK4 and CDK6, the so-called INK4 proteins of which four have now been identified.16-20 Importantly, cells that lack Rb function, either as a result of Rb mutation or loss, are refractory to the effects of INK4 overexpression,18,21-24 thereby underscoring the existence of a biochemical pathway in which INK4 proteins act “upstream” of cyclin D–dependent kinases to negatively regulate Rb phosphorylation. In contradistinction to the effects of mitogens on the D-type cyclins, we might well imagine that INK4 proteins would be induced by antiproliferative signals, and this is the case for p15^INK4b, which is synthesized in response to stimulation by transforming growth factor-β.17 25 

Although the activity of D-type CDKs is nonessential for G₁ progression in cells lacking Rb, this does not preclude the possibility that these enzymes phosphorylate other substrates. For example, the activation of D-type CDKs as cells enter the cycle may be important in canceling the activity of inhibitors that hold cells in a quiescent state. Alternatively, cyclin D–dependent kinases could act during G₁ phase to interfere with differentiation-specific programs that might otherwise be executed in noncycling cells. In agreement with this concept, it was previously shown that the enforced overexpression of cyclins D2 or D3 in immature IL-3–dependent 32Dcl3 myeloid cells shortened their G₁ phase26 and prevented their ability to differentiate to neutrophils in response to G-CSF.27 

To further explore the possibility that D-type CDKs might affect differentiation-specific or apoptotic programs, we have now studied the effects of inducible INK4 protein expression on interleukin-3 (IL-3)-dependent proliferation and granulocyte colony-stimulating factor (G-CSF )-induced differentiation in 32Dcl3 cells. These cells, and others of the myeloid lineage, normally express two INK4 family members, p18^INK4c and p19^INK4d.18,19 Their synthesis oscillates throughout the cell cycle with the highest levels observed in the S and G₂ phases and the lowest in G₁ phase, the latter being the only interval where elevated INK4 protein expression can inhibit cell proliferation.19 We now show that enforced overexpression of p19^INK4d and concomitant inhibition of cyclin D–dependent kinase activity in 32D cells not only blocks G₁ phase progression but also prolongs cell survival in the absence of supportive growth factors and, surprisingly, induces certain features of macrophage differentiation in the presence of IL-3. The p19 overexpressors exhibited accelerated differentiation to neutrophils in response to G-CSF, but even under these conditions, some cells acquired macrophage-like properties. Therefore, apart from their role in facilitating G₁ progression, the cyclin D–dependent kinases may negatively influence the execution of differentiation programs.

Cell lines and culture conditions.IL-3–dependent 32Dcl3 cells (32D cells)28 29 were provided by Dr J.N. Ihle (St Jude Children's Research Hospital, Memphis, TN) and were maintained in RPMI 1640 medium (Mediatech, Inc, Herndon, VA) supplemented with murine IL-3 (25 U/mL), and 10% fetal bovine serum (FBS; BioWhittaker, Walkersville, MD). Cultures were maintained in a humidified atmosphere of 5% CO₂ at 37°C. Growth rates of parental 32D cells and their derivatives (discussed later) were determined by seeding cells at 1 × 10⁵ cells per mL in 5 mL of complete media in 60-mm diameter tissue culture dishes and enumerating them daily after suspension in medium containing trypan blue dye. Unless otherwise stated in the figure legends, p19 expression was induced by adding 75 μmol/L ZnCl₂ to the culture medium. The time courses for zinc induction and G-CSF treatment are indicated in the figures and their legends.

DNA constructs and electroporation.A Not I-BamHI fragment containing mouse cDNA encoding p19^INK4d 19 was inserted unidirectionally into Not I-BamHI cloning sites downstream of the sheep metallothionein promoter in the eukaryotic expression vector, pMT-CB6⁺ (a gift from F. Rauscher, Wistar Institute, Philadelphia, PA) as used by others.30 The resulting pMT-CB6⁺p19 vector is bicistronic and contains a neomycin resistance gene (neo ) regulated independently by the Simian Virus 40 promoter. Circular vector plasmid DNA purified through cesium chloride gradients was linearized by restriction with Pvu I, and 50 μg was electroporated27 31 into 5 × 10⁶ 32D cells using a gene pulsar apparatus (Bio-Rad, Richmond, CA) at 960 μF and 290 V. Electroporated cells were cultured for 24 hours in complete medium and then selected for 10 days in medium containing 750 μg/mL G418 (GIBCO-BRL, Gaithersburg, MD). Clonal transformants were selected from single cells by limiting dilution, and six p19-inducible clones were chosen for further study. Control 32D cell lines were generated in parallel by introducing the naked pMT-CB6⁺ vector and selecting the cells in G418.

Protein analyses.Direct immunoblotting was performed using total cell lysates from 3 × 10⁵ cells per gel lane. After disruption of the cells in cell lysis buffer (50 mmol/L Tris HCl, pH 8.0, 120 mmol/L NaCl, 0.5% NP-40), boiled lysates were resolved by electrophoresis on denaturing polyacrylamide gels containing sodium dodecyl sulfate (SDS) and transferred to nitrocellulose. Immunoblotting was performed32 using a goat anti-mouse p19 polyclonal antibody (Santa Cruz Biotechnology Inc, Santa Cruz, CA) followed by incubation with a secondary donkey antibody directed to goat immunoglobulin G (IgG) conjugated to horseradish peroxidase (Santa Cruz Biotechnology Inc). Proteins were detected by enhanced chemiluminescence (ECL, Amersham, Arlington Heights, IL) according to the manufacturer's instructions. The levels of D cyclins and CDK proteins in total cell lysates from equivalent numbers of 32D cells were estimated by comparing the intensity of blot signals obtained using antibodies with the individual components versus those generated with defined amounts of the recombinant proteins run in adjacent lanes of the same gel. Recombinant cyclin D2, cyclin D3, CDK4, and CDK6 were prepared using a baculovirus expression system as described33 and were detected as above using monoclonal antibodies to the cyclins34,35 or antisera to the CDK4 (serum R_Z ) and CDK6 (serum R_JJ ) carboxyltermini.35 36 

In experiments designed to measure p19-CDK complexes formed in vivo, p19^INK4d was recovered by immunoprecipitation after detergent lysis of 5 × 10⁶ cells in buffer containing protease inhibitors.37 Washed immune precipitates generated with polyclonal rabbit antisera to p19 and recovered on protein A-Sepharose19 were separated by electrophoresis on 15% denaturing polyacrylamide gels, transferred to nitrocellulose, and immunoblotted as above using the same antisera to p19, or antisera directed to the CDK4 or CDK6 carboxyltermini. In analogous experiments designed to detect complexes containing cyclin D3, washed precipitates generated with monoclonal antibody to mouse cyclin D3³⁴ were separated on gels and blotted with the same antibody or with antisera directed to p19 or CDK4. Recombinant proteins produced under baculoviral vector control in insect Sf9 cells were used as positive controls as indicated in the legends of Figs 1-3.

Cyclin D–dependent protein kinase activity in immune complexes was measured as described35 after precipitation of lysates from 2 × 10⁷ cells with rat monoclonal antibodies to cyclins D2 or D3³⁴ or with rabbit antisera to the CDK4 or CDK6 carboxyltermini.35,36 Bacterially produced GST-Rb substrate was prepared as described33 38 and, after phosphorylation in vitro, was resolved by electrophoresis on denaturing 7.5% polyacrylamide gels which were dried and subjected to autoradiography.

To detect receptors for colony-stimulating factor-1 (CSF-1R), 32D cells and control mouse BAC1.2F5 macrophages39 were metabolically labeled for 1.5 hours with 200 μCi/mL L-[³⁵S]methionine (specific activity, 1,000 Ci/mmol; ICN, Irvine, CA) in methionine-free medium containing dialyzed FBS. 32D clones were labeled in the presence of IL-3. Lysates were precipitated with antisera to the recombinant feline v-fms40 41  or human c-fms31 42  gene products, both of which crossreact strongly with murine CSF-1R. Precipitated proteins were resolved on denaturing polyacrylamide gels and detected by autoradiography as described.43 

Coupled reverse transcription and polymerase chain reaction (RT-PCR) assay.Polyadenylated RNA was isolated using a mRNA isolation kit from Pharmacia (Piscataway, NJ), and cDNA was synthesized from 1 μg of mRNA templates according to the manufacturer's instructions (Strata-Script RT-PCR kit; Stratagene, La Jolla, CA). After amplification with primers specific for mouse CSF-1R44 (sense primer, 5′-CCAGAACTGGTTGTAGAGCC-3′; antisense, 5′-CAGCTTGCTAGGCTCC-AATT-3′) with 35 cycles of denaturation (94°C, 1 minute), annealing (55°C, 2 minutes) and extension (72°C, 3 minutes), products were electrophoresed on 1.5% agarose gels and visualized by ethidium bromide staining.

Cell morphology and immunohistochemistry.Cells collected with a Cytospin-3 cytocentrifuge (Shandon, Pittsburgh, PA) were processed by May-Grünwald-Giemsa staining (Sigma Chemicals, St Louis, MO) or for immunocytochemistry and enzyme cytochemistry. Cytocentrifuge preparations were air-dried and fixed for 10 minutes at 4°C, washed with phosphate-buffered saline (PBS), and incubated for 50 minutes at room temperature with antibodies diluted in PBS and containing 1% bovine serum albumin (PBS⁺). Cells were reacted either with monoclonal antibodies directed to murine CSF-1R (Upstate Biotechnology Inc, Lake Placid NY) or to Mac-3 (clone M3/84.6.34; American Type Culture Collection, Rockville, MD). After washing twice in PBS⁺, the preparations were incubated for 45 minutes with a mouse anti-rat horse radish peroxidase-conjugated immunoglobulin (Jackson ImmunoResearch Laboratories Inc, West Grove, PA) diluted in PBS⁺ with 10% vol/vol normal mouse serum. Peroxidase activity was shown with 3,3′-diaminobenzidine tetrahydrochloride as substrate (Sigma). Isotype-matched control immunoglobulins (PharMingen, San Diego, CA) were used in place of anti–CSF-1R or anti–Mac-3 as negative controls. Acid phosphatase activity was shown45 with naphthol AS-BI phosphate (Sigma) as substrate and hexazotized pararosaniline as diazonium salt (90 minutes at 37°C).

Fc-receptors and phagocytosis.Fc-receptor (FcR)-positive cells were detected by rosette formation with IgG-coated sheep red blood cells (SRBC).46 A 1.6% vol/vol suspension of washed SRBC was made in PBS⁺, and mouse anti-SRBC IgG was added at a subhemagglutinating concentration. After a 30-minute incubation at 37°C, the SRBC were washed twice in PBS⁺ and resuspended in PBS⁺ to a 1.6% vol/vol suspension (SRBC solution). 32D cells (2 × 10⁵ cells) were incubated for 1 hour at 4°C with 200 μL SRBC solution, and the percentage of cells binding three or more erythrocytes per cell was determined by examination of the suspension under a light microscope. Immunophagocytosis was assayed by incubating 2 × 10⁵ cells in 200 μL SRBC solution for 1 hour at 37°C. After osmotic lysis of extracellular SRBC using 0.83 mol/L NH₄Cl₂ solution,46 32D cells were cytocentrifuged and processed for May-Grünwald-Giemsa staining. The percentage of phagocytosing cells was microscopically determined. In both assays, uncoated SRBC were used as negative controls.

Flow cytometric analyses.Cell suspensions screened for expression of specific cell surface antigens were incubated for 20 minutes at 4°C with PBS⁺ containing 10% vol/vol normal mouse serum, followed by a 30-minute incubation with optimal concentrations of the indicated monoclonal antibodies diluted in the same solution. Monoclonal antibodies used for staining included phycoerythrin (PE)-conjugated anti–Mac-1 (PharMingen), anti–Mac-3 (PharMingen), and anti–F4/80 antibody (Caltag Lab, San Francisco, CA), and fluorescein isothiocyanate (FITC)-conjugated anti-CD14 (PharMingen), together with isotype-matched control antibodies (PharMingen). Cells were analyzed on a FACScan flow cytometer (Becton Dickinson, Mountain View, CA) as described.47 Where indicated, the DNA content of cells was measured by flow-cytometric analysis of propidium iodide-stained nuclei, as previously described.15 

Inducible overexpression of p19^INK4d in 32Dcl3 myeloid cells.32Dcl3 cells were transfected with an expression vector encoding p19^INK4d under the control of the sheep metallothionein promoter, and stable clones were generated that overexpressed the p19 protein in response to zinc-treatment. Immunoblotting (Fig 1A) revealed that asynchronously proliferating, parental 32D cells (lane 1) and neomycin-resistant derivatives transformed with the naked expression vector (lane 2) expressed low levels of endogenous p19 when propagated in medium containing IL-3 (see also Fig 3A, lanes 2 and 3). Because of “leakiness” of the expression vector, six representative clones (lanes 4 to 15) expressed somewhat higher levels of p19 when grown under noninducing conditions (even-numbered lanes). Despite this constitutive increase in basal p19 synthesis, these levels of the inhibitor were insufficient to block cells in G₁ phase, as observed in previous studies with rodent fibroblasts.19 Addition of 75 μmol/L zinc to the media resulted in marked increases in p19 accumulation within 4 hours in each clone (odd-numbered lanes). Induction of p19 in response to heavy metals was dose-dependent and rapid, being maximal by 5 hours after treatment of cells with 75 μmol/L zinc, and high levels of the protein persisted as long as cells were maintained in medium containing IL-3 and zinc (Fig 1B). Despite braking effects on their proliferative rate, and changes in their differentiation status (discussed later), cells maintained in both IL-3 and zinc remained fully viable. Induction of p19 was reversible, and basal levels of p19 expression were again achieved within 3 to 4 days after removal of zinc from the medium (data not shown).

p19-mediated inhibition of cyclin D-dependent kinases.Like many immature myeloid cells, 32D cells express cyclins D2 and D3, but not D1, together with the two major cyclin D–dependent catalytic subunits, CDK4 and CDK6.26,27 The expression of the D-type cyclins and their associated kinases is highly IL-3–dependent. When 32D cells are switched to medium containing G-CSF, cyclin D2 is degraded within 24 hours and cyclin D3 by 72 hours after IL-3 withdrawal, resulting in the complete loss of cyclin D–dependent kinase activity and G₁ phase arrest.27 The arrested cells then terminally differentiate to granulocytes over the next 7 to 10 days and ultimately die.29 

Because combinatorial assembly generates four distinct cyclin D-CDK holoenzymes in cells grown in IL-3, we estimated the relative expression of cyclins D2 and D3, and CDK4 and CDK6 in parental 32D cells and in representative p19-transfected subclones maintained in IL-3, both in the presence or absence of zinc. Using specific, well-characterized antibodies directed to the individual components (see Materials and Methods), the relative concentration of each protein in lysates from equivalent numbers of cells was quantitated by comparing their immunoblot signals with those generated by known quantities of purified, recombinant proteins run side by side on the same gels (Fig 2). The calculated quantity (ng) of each cyclin (panels A and B) or CDK (panels C and D) is denoted beneath lanes 2 to 4 of each panel. This experimental approach negates any requirement for using antisera of similar titers or for matching the autoradiographic exposure times of different panels.

In proliferating parental 32D cells, cyclin D3 was expressed at 10- to 15-fold higher levels than cyclin D2 (compare Fig 2A and B), whereas the level of CDK6 was about half that of CDK4 (Figs 2C v D). Therefore, cyclin D3–dependent kinases should predominate in these cells. In addition, the levels of D-type cyclins and their associated CDKs were not greatly different (within a factor of 2) in parental versus p19-transfected cells, regardless of whether p19 was induced by zinc or not (for exemplary control data for D3 expression in zinc-treated parental cells, see Fig 3C). The steady-state concentrations of the CDKs were significantly greater than those of the D-cyclins (Fig 2C and D v A and B), consistent with data indicating that the D-cyclins are the rate-limiting components.13 48 

Using a monoclonal antibody to mouse cyclin D3, pRb kinase activity was readily detected in immune complexes precipitated from cells grown in IL-3 (Fig 1C; anti-D3, time = 0), whereas no significant kinase activity was obtained using an irrelevant, isotype-matched control antibody (lane C). Following treatment of the cells with zinc for different times, cyclin D3–dependent kinase activity recovered from the lysates decreased rapidly and approached background levels by 2.5 hours of zinc addition (anti-D3, remaining time points to 55 hours), correlating with the rapid kinetics of p19 induction (Fig 1B). When a recombinant glutathione S-transferase (GST)-p19 fusion protein19 was added to lysates of untreated cells before immunoprecipitation, pRb kinase activity was directly inhibited (Fig 1C, +GST-p19), whereas GST itself was without effect (+GST). Similar data were obtained using antisera directed to carboxylterminal epitopes of CDK4 (Fig 1D), where in matched exposures, the CDK4-dependent kinase activity approached that of the cyclin D3–dependent kinase (Fig 1D v C). The remaining D3-associated Rb kinase activity was caused by CDK6 (data not shown), consistent with the relative abundance of the two cyclin D–dependent catalytic subunits (Fig 2C and D). In contrast, negligible pRb kinase activity was detected under these conditions using antibodies to cyclin D2 (data not shown), in agreement with data indicating that cyclin D3 was present in more than 10-fold excess in these cells (Fig 2A and B).

As predicted, when p19 was induced, it moved into complexes with both CDK4 and CDK6 in vivo. The level of p19 expressed in a transfected clone grown in IL-3 was approximately twofold higher than that of parental 32D cells (Fig 3A, lane 4 v lane 2). Whereas zinc treatment for 5 hours did not affect p19 expression in control transfectants (Fig 3A, lane 3 v lane 2), it led to marked induction of the protein in the INK4d-transfected clone (lane 5). When lysates from an equivalent number of zinc-induced cells were immunoprecipitated with antiserum to p19 and the precipitated proteins were separated on denaturing gels, transferred to nitrocellulose, and blotted with antiserum to CDK4, CDK4 subunits could now be readily visualized in complexes with p19 (Fig 3B, lane 5). Similar results were obtained using antiserum to CDK6 (not shown). Cyclin D3 was expressed at the same levels in control and INK4d-transfected cells, whether or not they were treated for 5 hours with zinc (Fig 3C), but complexes between cyclin D3 and CDK4 were largely disrupted in cells induced to overexpress p19 (Fig 3D, lane 5). Notably, immunoprecipitates prepared with antibody to cyclin D3 contained p19 (Fig 3E, lane 5), even though p19 does not directly associate with D-type cyclins.19 Hence, some p19 must have entered into ternary complexes with residual cyclin D3-CDK4 (Fig 3D and E, lane 5), consistent with previous observations that such complexes can form in vitro and are enzymatically inactive.19 We can therefore conclude that induction of p19 (Fig 1A and B) and formation of p19-CDK complexes, either containing or lacking cyclin D (Fig 3), resulted in a loss of cyclin D–dependent kinase activity (Fig 1C and D), even though cyclins D2 and D3 and CDKs 4 and 6 were continuously expressed (Fig 2). Similar data to those shown with clone 2C were obtained with other individually derived p19-inducible subclones.

Induction of p19 induces cell cycle arrest and prolongs viability in the absence of IL-3.In each of six independent clones tested, the induction of p19 and the resulting inhibition of cyclin D-dependent kinase activity induced cell cycle arrest. When cells were transferred to medium containing IL-3 and zinc, the growth rate of the p19 overexpressors was retarded, as compared with transfected clones grown in IL-3 alone, which proliferated at rates indistinguishable from those of parental cells or of neo-resistant derivatives transfected with the naked vector (Fig 4A). Flow cytometric analysis of their DNA content showed that most cells expressing high levels of p19 accumulated rapidly in G₁ phase within a single cell cycle (Fig 4C). Although the percentage of cells in G₁ phase almost doubled within 10 hours of zinc addition to these cultures, no further increases in the G₁ fraction were observed at later times, implying that growth arrest was incomplete. This was consistent with the fact that some zinc-treated cells continued to proliferate at a slow rate (Fig 4A).

Like parental 32D cells, the viability of the p19-inducible subclones was strictly dependent on IL-3, but their rate of cell death after IL-3 withdrawal was significantly slower than that of parental 32D cells or of neo-resistant control transfectants (Fig 4B). When shifted into medium lacking both zinc and IL-3, the p19-inducible cells, which express higher basal levels of p19 than parental cells (Figs 1A and 3A), showed extended survival. Induction of p19 in response to zinc treatment further prolonged their viability (Fig 4B). Thus, induction of p19 protects against the rapid cell death that normally accompanies growth factor withdrawal.

Cells overexpressing p19 acquire macrophage characteristics.May-Grünwald-Giemsa staining of cytocentrifuged preparations showed that p19-transfected 32D clones exhibited a myeloblastic phenotype similar to that of parental 32D cells when grown in the presence of IL-3 (Fig 5A v B). However, after 4 days of zinc-treatment, the p19 inducible cells appeared larger, developed a higher cytoplasmic to nuclear ratio with conspicuous cytoplasmic vacuoles, and became somewhat adherent (Fig 5C). The cytoplasm of zinc treated cells also stained for acid phosphatase whereas untreated cells did not (data not shown), consistent with increased lysosomal activity following induction. No such effects were observed in zinc-treated control cells that had been transfected with the naked vector. Parental 32D cells and p19-inducible clones, whether treated with zinc or not, expressed equivalent numbers of FcRs, but FcR-dependent immunophagocytosis was also observed in approximately 6% of zinc-treated cells that overexpressed p19 (Fig 5D).

The expression of macrophage-specific antigens on the surfaces of induced and uninduced cells was examined using monoclonal antibodies against Mac-1,49 Mac-3,50 CD14,51 and F4/80.52 As shown in Fig 6, parental 32D cells stained positively in different degrees for Mac-1, Mac-3, and F4/80 antigens.53 54 Overexpression of p19 led to a pronounced and generally uniform increase in Mac-1 and Mac-3 fluorescence, with a lesser effect on CD14 expression, but with no detectable change in F4/80. Therefore, cells induced to express p19 acquired certain properties characteristically found in more differentiated cells of the monocyte/macrophage lineage.

Although expression of the CSF-1R is one of the hallmarks of mononuclear phagocyte maturation,55-58 inducible clones remained unresponsive to CSF-1, whether treated with zinc or not. When exposed to CSF-1 instead of IL-3, all clones failed to survive, suggesting that they expressed negligible or no CSF-1R (data not shown). Clones were metabolically labeled with [³⁵S]methionine both before and at various times after zinc treatment, and lysates were immunoprecipitated with two different high titer rabbit antisera reactive to murine CSF-1R. As a positive control, lysates of metabolically labeled BAC1.2F5 macrophages were analyzed in parallel. Again, no CSF-1R was detected in any of the 32D derivatives, although receptor expression was readily documented in control macrophages (Fig 7A). Cytohistochemical staining for murine CSF-1R confirmed that no p19-inducible 32D clones expressed receptors after zinc treatment, whereas BAC1.2F5 macrophages were uniformly positive (data not shown).

We therefore undertook RT-PCR analysis to detect CSF-1R mRNA, reasoning that this would be the most sensitive test for CSF-1R expression that we could apply. No PCR products were documented in parental or p19-transfected 32D cells, whether they were treated with zinc or not (Fig 7B, lanes 1 to 3). However, control BAC1.2F5 macrophages yielded the expected 554 base-pair CSF-1R PCR product, as did those that had been mixed with p19-inducible 32D cells at 5% or 1% levels (Fig 7B, lanes 4 to 6). Because the majority of zinc-treated p19-inducible 32D cells expressed increased Mac-1 antigen on their surface (Fig 6A), whereas a minimum of 6% were phagocytic (Fig 5D), we should have readily detected CSF-1R mRNA products had they been represented by at least 1% of cells in the population.

The discussed results raised the intriguing possibility that p19-induced 32D cells might not terminally differentiate because they lacked CSF-1R. Therefore, we transfected one of the p19-inducible clones with a vector encoding human CSF-1R. Transfection of immature, IL-3–dependent myeloid cell lines with CSF-1R cDNA can confer CSF-1 responsiveness on a small percentage of derived subclones, and a fraction of these surviving variants are able to differentiate toward macrophages.31,53,54 59 Nonetheless, in the experiments here, selected drug-resistant 32D cell lines uniformly expressing CSF-1R could not be maintained in medium containing human recombinant CSF-1, and they failed to upregulate their macrophage markers when treated with human CSF-1 plus zinc (negative data not shown).

When transferred from IL-3–containing medium to that containing G-CSF but lacking zinc, clones capable of expressing p19 retained their capacity to differentiate to neutrophils in response to G-CSF, and no macrophage-like cells were observed in the cultures. After only a 3-day exposure to G-CSF alone, p19-inducible clones already had begun to exhibit neutrophilic morphology, as marked by the formation of ring-shaped nuclei and band forms (Fig 5E). When such cells were treated with both G-CSF and zinc, the majority still formed neutrophils, but a minority (∼5%) acquired macrophage characteristics (Fig 5F ). Induction of p19 accelerated granulocyte formation in response to G-CSF by the majority of cells (Fig 5F v E). Hence, although cells differentiating in G-CSF alone lose cyclin D–dependent kinase activity because of cyclin degradation upon IL-3 withdrawal,27 more rapid p19-induced inhibition of these enzymes appeared to accelerate differentiation. In addition, the ability of zinc plus G-CSF to selectively support the acquisition of macrophage markers by a small fraction of the p19-inducible cells argues that p19, and not IL-3, triggers this alternative response. Cultures that were treated for 4 days with IL-3 and zinc and had acquired macrophage characteristics (Fig 5C) also retained the ability to differentiate to neutrophils when shifted to medium containing G-CSF alone, in agreement with the concept that the differentiation of these cells in the macrophage lineage is incomplete and reversible (see above).

Negative regulation of cell cycle progression may help reinforce the proper execution of differentiation programs and coordinate the number of mature cells that develop from uncommitted progenitors in response to a particular differentiation stimulus. Universal CDK inhibitors of the Cip1/Kip1 family as well as specific inhibitors of cyclin D–dependent kinases (the INK4 proteins) have been broadly implicated in such processes.60 Not surprisingly, many CDK inhibitors are specifically upregulated during differentiation of various cell types in vivo.61-68 However, their exact biological functions remain a mystery.

In normal tissue development, the functions of particular CDK inhibitors may well be redundant, so that the individual loss of one such regulator might be compensated by the presence of another. Nullizygous animals unable to synthesize p21^Cip1 or p16^INK4a develop entirely normally69-71 even though both proteins have the capability to govern differentiation-specific decisions in cultured cell lines (discussed below). On the other hand, mice lacking p27^Kip1 develop generalized organomegaly being 60% larger than their wild-type littermates, and their tissues are hyperplastic and contain cells that are unusually small in size.72-74 The latter results are in good agreement with observations made with cultured cell lines, in which cells overexpressing G₁ cyclins were reduced in size and had a decreased dependency on mitogens,13,75 and where inhibition of p27^Kip1 inhibitory function prevented the immediate cell cycle exit usually seen upon mitogen withdrawal.76 77 

Overexpression of cyclin D–dependent kinases in cultured myoblasts can cancel the ability of the MyoD transcription factor to promote muscle cell differentiation; conversely, enforced expression of CDK inhibitors, including p21^Cip1 or p16^INK4a, can potentiate muscle cell differentiation even in the presence of otherwise overriding mitogenic signals.64-66 Transient overexpression of p21^Cip1 triggers differentiation of U937 cells to macrophages, suggesting that G₁ arrest in this system might itself be sufficient to activate the program.78 Similarly, enforced expression of cyclins D2 and D3 in 32Dcl3 myeloid cells prevented their differentiation in G-CSF,27 whereas, as shown here, p19^INK4d overexpression and the resulting inhibition of cyclin D–dependent kinase activity induced differentiation, albeit in an unusual manner.

When cultured in IL-3, induction of p19 by addition of zinc to the medium led to the expression of macrophage-specific markers. The morphology of virtually all of the cells changed dramatically, and the vast majority expressed considerably higher levels of particular cell surface antigens, such as Mac-1. In contrast, a smaller fraction of zinc-treated cells expressed higher levels of Mac-3 or CD14, and only about 6% exhibited frank Fc-dependent phagocytic activity. The nonuniformity of the differentiative response, despite the fact that the cells were clonally derived, may reflect the variability in the magnitude of p19^INK4d induction or extent of cyclin D–dependent kinase inhibition in individual cells. Although most cells showed greatly reduced cyclin D3- and CDK4-dependent pRb kinase activity within 2.5 hours of zinc treatment and arrested in G₁ phase within 12 hours, a fraction continued to proliferate at a reduced rate, indicating that, for whatever reason, their response was not complete. Moreover, differentiation toward macrophage-like cells in each of six individually derived clones was not terminal, and the cells reverted to an immature phenotype within 3 to 4 days after zinc was removed from the cultures. Although we cannot exclude the possibility that this in part reflected the regrowth of less differentiated 32D precursors that remained in the zinc-induced cultures, the uniformity of the observed changes in cell morphology and Mac-1 expression following zinc addition and its subsequent withdrawal argues that the more differentiated cells were not committed to the macrophage lineage and were capable of reversion.

Although 32D cells do not normally differentiate toward macrophages, ectopic expression of the CSF-1 receptor54 or of the Egr-1 transcription factor in the presence of CSF-179 was previously shown to induce expression of macrophage-specific markers. In our case, however, differentiation within the mononuclear phagocyte lineage was completely independent of CSF-1, as the cells lacked any detectable CSF-1 receptors and did not respond to the exogenously added growth factor. Neither did the subsequent introduction and expression of the human CSF-1 receptor gene (FMS ) in the p19-inducible 32D cells enable them to respond further to recombinant human CSF-1.

Like parental 32D cells, most p19-overexpressing clones terminally differentiated to neutrophils in G-CSF, whether induced by zinc or not. The parental cells and the p19-inducible subclones synthesize cyclins D2 and D3 in response to IL-3, but they terminate cyclin D synthesis when transferred to medium containing G-CSF.26 27 Cyclin D2 is degraded very rapidly under these conditions, whereas cyclin D3, the major D-type cyclin expressed in this cell line, persists for 2 to 3 days. The cells arrest in G₁ within this 2- to 3-day time frame and then differentiate toward granulocytes. When p19-inducible clones were cultured in IL-3 and zinc, the vast majority arrested in G₁ phase within a single cell cycle, even though cyclin D was not yet degraded. By 5 hours after zinc treatment, p19 had disrupted most cyclin D-CDK complexes and had moved into inactive, ternary complexes with the remainder, effectively inhibiting all detectable cyclin D–dependent kinase activity. Similarly, p19-induced G₁ arrest occurred much more rapidly in cells transferred to G-CSF plus zinc than in those cultured in G-CSF alone. The fact that some zinc treated cells acquired macrophage markers when grown in either IL-3 or G-CSF must, therefore, reflect an activity of p19^INK4d itself. Clearly, cell cycle arrest per se cannot account for the observed macrophage differentiation, as G-CSF-treated cells also arrested in G₁ , but at a slower rate. Nor was IL-3 required, because p19-induced macrophage differentiation occurred as long as the viability of zinc-treated cells was ensured by either IL-3 or G-CSF. Indeed, induction of p19^INK4d extended the survival of these cells in the complete absence of hemopoietins.

Taken together, these results suggest that a specific inhibitor of cyclin D–dependent kinases can induce G₁ arrest, cell survival, and a partial differentiative response, possibly through independent biochemical pathways. The simplest interpretation is that cyclin D–dependent kinases can phosphorylate substrates other than Rb to influence gene expression. We suggest that one role of cyclin D–dependent kinases may be to cancel differentiation programs in cycling cells, thereby coupling their execution with cell cycle arrest. Attractive substrates for these kinases might include differentiation-inducing transcription factors whose activities might be negatively regulated through such a mechanism.80 

We thank Hiroshi Hirai and Dawn E. Quelle for constructive suggestions during the course of this work; Richard A. Ashmun for flow cytometric analyses of DNA content; and Shawn Hawkins, Carol Bockhold, and Joseph Watson for excellent technical assistance.