INTRODUCTION

Healthy bone results from balanced ongoing bone formation and resorption. Normal bone resorption is carried out by osteoclasts, which are terminally differentiated cells of the monocyte-macrophage lineage. This resorption of bone is absolutely dependent on the osteoclasts' ability to generate an acid compartment on the surface of bone (1, 2). The low pH is essential for the solubilization of the alkaline salts of bone mineral (3) and the digestion of organic bone matrix by acid hydrolases, which osteoclasts secrete (4).

A model of osteoclast ion transport activities involved in the acidification of bone surface is shown in Fig. 1 (1, 2, 5-7). Osteoclasts attach to bone and form a circumferential sealing zone that isolates the bone resorption compartment from the extracellular space. The observation that NH₄Cl reversibly inhibits bone resorption by osteoclasts indicates that the resorption compartment is acidic and that the sealing zone is impermeant to H⁺ and NH₄⁺ (2, 8, 9). The osteoclast cytoplasm is rich in carbonic anhydrase (10), providing a continuous supply of protons and bicarbonate. Osteoclast plasma membrane within the sealing zone develops into the ruffled border. Protons are transported across this membrane into the bone resorption compartment by a V-type H⁺-ATPase (11-18). Chloride ions passively follow the protons through conductive anion channels (19, 20). The combined activities of the proton pump and chloride channel acidify the resorption compartment and alkalinize the cytoplasm. Bicarbonate exits the cell into the extracellular space in exchange for chloride via a basolateral electroneutral anion exchanger, correcting the cytoplasmic alkalinization and compensating for cytoplasmic chloride loss (21). The net result of these coordinated transport activities is the transcellular movement of HCl into the bone resorption compartment.

This model predicts that both the ruffled border proton pump and chloride channel play key roles in bone resorption. The proton pump provides the proton-motive force necessary to generate a pH gradient. However, the pump is electrogenic. The chloride channel short-circuits the electrogenic pump and allows maximal proton transport. It follows that limitation of the chloride conductance could inhibit acid transport independently of the intrinsic activity of the proton pump. Analogous to a current model for regulation of the pH of some intracellular organelles (22), regulation of the anion conductance rather than proton pump activity could be the key point at which the rate of osteoclast acid transport, and hence bone resorption, is governed. Thus, molecular characterization of the ruffled border chloride channel may provide insight into regulation of osteoclast bone resorption and could define a pharmacologic target for the treatment of metabolic bone disease.

We sought to determine whether the ruffled border chloride channel activity controls osteoclast acid transport. We have used the differentiation of bone marrow monocytes in culture as a model system. Bone marrow mononuclear cells derived from calcium-deprived hens have been shown to differentiate in culture spontaneously into multinucleated cells expressing markers of the osteoclast phenotype, including tartrate-resistant acid phosphate, carbonic anhydrase, and the _v3 integrin (23-25). However, these cells poorly resorb bone until cultured with devitalized bone for at least 24 h, consistent with prior observations on the importance of bone in osteoclast development (10, 26-29). Thus, it appears that in a final step, exposure to bone induces a transition from morphologically differentiated but inactive osteoclasts into bone-resorbing cells.

In this study, we assess the expression of the chloride channel and proton pump as bone marrow mononuclear cells differentiate in culture into bone-resorbing osteoclasts. We demonstrate that induction of the chloride channel rather than the proton pump is the critical step in development of acid transport when these cells are exposed to bone. Furthermore, we show that the ruffled border chloride channel is related to the bovine microsomal chloride channel, p64, and that novel p64-related protein and mRNA are induced as the differentiating cells are exposed to bone. Finally, we characterize the single channel properties of the ruffled border chloride channel incorporated in planar lipid bilayers.

EXPERIMENTAL PROCEDURES

Osteoclasts and bone marrow cells were isolated from the long bones of calcium-deprived laying hens (Gallus domesticus) (30). Membrane vesicles from mature osteoclasts were prepared as described (20). These vesicles contain all the known subunits of the proton pump as well as the chloride channel required for acidification (2, 18). The cultured cells were scraped from dishes and suspended in buffer at 4 °C as described (20), disrupted in a tight-fitting glass Dounce homogenizer, and centrifuged at 2000 rpm to remove unbroken cells and bone particles. This supernatant was then spun at 100,000 × g for 60 min to sediment all the cellular membranes. The supernatant was removed, and the membrane pellet was frozen at 70 °C for later analysis.

Membranes prepared as described above were suspended in 300 µl of 140 mM KCl and 10 mM HEPES (pH 7.0) and allowed to re-form vesicles for 30 min on ice prior to assay. The protein concentration of this suspension was determined using the BCA protein assay method (Pierce) to correct initial rates for protein. To assay acidification, 30 µl of the vesicle suspension was added to 2.0 ml of 140 mM KCl and 20 mM HEPES (pH 7.0) containing 3.3 µM acridine orange, 2.5 mM ATP, and other reagents as indicated. The reaction was started by adding 10 mM MgCl₂ to the reaction mixture at 37 °C, and fluorescence was followed in an SPF500 spectrofluorometer at an excitation of 460 nm and an emission of 520 nm (20). This assay provides a linear increase in initial rate over a 40-fold change in added membrane protein and can be used to compare specific activities. When desired, 0.1 µM valinomycin was added 20 s prior to the addition of MgCl₂.

A chicken genomic library (Stratagene, La Jolla, CA) was screened by hybridization with a probe containing the entire coding region of bovine p64 (31). Isolated genomic clones were sequenced using the Sequenase system (U. S. Biochemical Corp.). An exon was identified encoding a region homologous to the C-terminal hydrophilic domain of bovine p64. The chicken exon sequence corresponds to nucleotides 1314-1469 in the bovine p64 sequence with 84% identity. The amino acid sequence predicted by the exon corresponds to positions 387-437 (C terminus) in the p64 amino acid sequence and is followed by a stop codon. The chicken and bovine sequences show 88% identity and 94% similarity. An 18-amino acid peptide from this region, ^NCAADKEIEQAYADVAKRL (sites of nonidentity between the chicken and bovine sequences are underlined), was synthesized and covalently linked to rabbit serum albumin by the Peptide Synthesis Facility in the laboratory of Dr. Robert Meechum (Department of Cell Biology, Washington University). This peptide was used to immunize a rabbit (Cocalico Biological, Reamstown, PA).

A high titer antiserum was obtained. The antiserum recognized both bovine p64 and the chicken osteoclast ruffled border chloride channel, p62. For Western blotting, the antiserum was affinity-purified on LacZ-p64 fusion protein (31) immobilized on an Amino-link column (Pierce). The antiserum against the 70-kDa H⁺-ATPase subunit was prepared in the same way against a peptide previously shown to be a consensus sequence of this subunit in vertebrates (18, 32).

Freshly isolated avian medullary bone was fixed in 0.1% glutaraldehyde, 4% formaldehyde, and 100 mM cacodylate in phosphate-buffered saline (PBS)¹ (pH 7.2) (11). Tissue samples (0.5 cm) were fixed at 4 °C for 18 h, post-fixed in 1% OsO₄, dehydrated, and embedded in epoxy (London Resin White, EBTEC, Agawam, MA). Silver-gray ultrathin sections were cut, collected on uncoated 83-µm pitch nickel grids, and dried for 18 h at 25 °C. Sections were hydrated for 30 min in PBS-TB (PBS with 0.05% Tween 20 and 0.1% bovine serum albumin) and reacted with antibody or preimmune rabbit serum diluted 1:20 in PBS-TB for 18 h in humidified air at 25 °C. After three rinses, antibody was visualized by reaction for 8 h at 25 °C with protein A-conjugated 9-nm gold beads (Sigma) at A_520 nm = 0.2 in PBS-TB, followed by washing in PBS with 0.05% Tween 20, rinsing in distilled water, and counterstaining with lead citrate and uranyl acetate.

Poly(A)⁺ RNA was prepared, separated, and blotted as described previously (31). The PstI fragment containing the entire coding region from cDNA clone H2B (31) was used as a p64 probe. Proton pump probes were generated from bovine kidney total RNA using reverse transcription-polymerase chain reaction (33), amplifying nucleotides 707-1224 in the 70-kDa sequence (34) and nucleotides 114-598 in the 58-kDa (brain) isoform sequence (35). Polymerase chain reaction products of the expected size were generated and cloned by ligation with the PCR II vector (Invitrogen). The identity of clones was confirmed by DNA sequencing. Probes were labeled with ³²P by random priming, and hybridization was carried out by standard methods (36) in 40% formamide at 37 °C. The final wash was at moderate stringency in 30 mM NaCl, 3 mM sodium citrate (pH 7.0), and 0.1% sodium dodecyl sulfate at 55 °C for 15 min.

To prepare samples for Western analysis with antibody 656 (raised against the 16-mer polypeptide derived from the osteoclast ruffled border chloride channel gene), osteoclast membranes prepared as described above were suspended in 140 mM KCl and 10 mM HEPES and brought to 1 M NaSCN by the addition of 0.25 volume of 5 M NaSCN. The suspension was mixed and incubated on ice for 30 min, and membranes were collected by centrifugation at 100,000 × g for 1 h.

Proteins were separated by SDS-PAGE; electroblotted onto polyvinylidine difluoride membrane by standard methods (37); stained with Ponceau S to reveal the positions of molecular mass standards; and then blocked with TNT (50 mM Tris (pH 7.5), 200 mM NaCl, and 0.1% Tween 20) with 5% nonfat dried milk for 1 h. The primary antibodies were diluted into TNT with 1% milk and incubated with the blot for 1 h. Blots were washed four times in TNT with 1% milk for 10 min each and then incubated with alkaline phosphate-conjugated goat anti-rabbit IgG (Sigma) diluted 1:5000 in TNT with 1% milk for 1 h. The blot was washed as described above and developed with bromochloroindolyl phosphate and nitro blue tetrazolium (37).

Planar lipid bilayers were prepared by spreading 30 mg/ml asolectin in n-decane across the 0.25-mm orifice of a polystyrene cuvette (Warner Instrument Corp., Hamden, CT). This orifice communicated between the cis-solution (1.0 ml) and the interior of the cuvette (trans-solution; 0.5 ml). The chambers were connected to silver chloride wires by 3 M KCl and 3% agar bridges, and currents were monitored by a BCA-525 bilayer clamp and headstage (Warner Instrument Corp.). The data were digitized using Axotape software and a Digidata 1200 board (Axon Instruments, Inc.). This system was calibrated for response and accuracy using a bilayer model provided by Warner Instrument Corp. Records were archived on video tape using a Neurocorder DR-484 (Neuro Data Instrument Corp., Water Town Gap, PA). Data were commonly filtered at 1000 Hz, and digitization was at 3.3 × 10³ points/s. Analysis of the data was done in part with Pclamp (Version 6.02; Axon Instruments, Inc.). Digitized data were histogramed and fitted using the Levenberg-Marquardt method. Data were plotted using Origin (Microcal, Inc).

Membranes were allowed to thin to a capacitance of 400 nanofarads, at which point the bilayer typically had a leak current of <2 pA at 100 mV of applied potential. Vesicles were added to the cis-chamber with stirring, and the bilayer current was monitored for fusion and channel activity. Each current-voltage plot (see Figs. 6 and 7) was repeated (n = 3) in multiple experiments, yielding the limiting slope conductance and reversal potential indicated below.

RESULTS

Four distinct types of cells derived from the bones of calcium-deprived chickens were used as sources of material in these experiments. 1) "Mature osteoclasts" were isolated from bone and used immediately in the preparation of membranes and RNA. 2) Mononuclear cells that adhered to culture dishes after 24 h are referred to as "day 1" cells and are committed to osteoclast differentiation. 3) Day 1 cells that are cultured for 5 days acquire much of the osteoclast phenotype, except rapid bone resorption (23, 24), and are "5-day no-bone" cells in our experiments. 4) Mononuclear cells that are cultured for 5 days with 12 µg of devitalized bone particles/ml develop into multinucleated cells that rapidly excavate bone surfaces and are referred to as "5-day plus-bone" cells. We examined acid transport and the expression of the proton pump and chloride channel in these four cell types.

Ruffled border membranes prepared from mature chicken osteoclasts contain a bafilomycin A₁-sensitive proton ATPase in parallel with a DNDS-inhibitable chloride conductance (19, 20). Acidification of these vesicles requires Mg²⁺-ATP and a permeable anion (i.e. chloride) and can be limited by inhibition of either the H⁺-ATPase or the anion channel. The chloride dependence of acidification is eliminated by collapsing the potential gradient generated by the proton pump with the potassium ionophore valinomycin, making acidification directly dependent upon H⁺-ATPase activity and independent of chloride channel activity. We assayed vesicle acidification in the presence and absence of valinomycin to determine whether it was limited by chloride channel activity. Activity in total cell membranes rather than subcellular fractions was studied to avoid potential complications from differences in polarization between cells grown with or without bone.

Membrane vesicles from each of the cell types described above were assayed for acidification by acridine orange quenching (Fig. 2) as described previously (2, 18, 20). In this assay, acidification of the vesicle interior leads to accumulation of the weak base, acridine orange. The fluorescence of acridine orange is quenched as the intravesicular concentration rises, and vesicle acidification is reflected as a fall in fluorescence. During the first 3-10 s of vesicle acidification, the rate of acridine orange quenching is linear, strictly dependent upon the amount of vesicles added, and demonstrates classic hyperbolic behavior as the substrate concentration is varied (20).

Fig. 2A shows the acid transport properties of vesicles prepared from 5-day plus-bone cells. These vesicles acidify rapidly on addition of MgSO₄. Very little increase in the rate of acidification is seen in the presence of valinomycin, indicating that the membrane potential is not limiting. Replacement of chloride with the impermeant anion sulfate drastically inhibits acidification. Bafilomycin A₁ completely blocks acidification. Thus, these vesicles contain both an active proton pump and an active chloride channel, and the chloride channel activity present is adequate to support full proton pump activity. This pattern of activity is essentially identical to the acid transport properties of ruffled border vesicles from mature osteoclasts as previously reported (20) and is plotted in Fig. 2A for comparison.

Fig. 2B shows the acidification of membrane vesicles prepared from 5-day no-bone cells. In sharp contrast, these vesicles acidify poorly in a standard reaction mixture, but there is a marked increase in acidification in the presence of valinomycin. As before, bafilomycin A₁ completely inhibits acidification. Therefore, membranes from 5-day no-bone cells contain an active proton pump whose activity is limited by the membrane potential. The acidification of these vesicles is limited by the low level of anion permeability, not the intrinsic activity of the proton pump.

The acidification of membrane vesicles from day 1 cells is shown in Fig. 2C. These membranes demonstrate minimal acidification, with slight enhancement by valinomycin and inhibition by bafilomycin A₁. Thus, they express minimal levels of both the proton pump and chloride channel.

The initial rates of acidification by each membrane preparation in the presence and absence of valinomycin are shown in Fig. 2D. Rates were normalized to that of the ruffled border vesicles prepared from fresh osteoclasts and corrected for protein concentration. In cells cultured without bone, ~90% of acidification is valinomycin-dependent. It is only in the presence of bone that the cells express both pump and channel activities at levels adequate to support acidification fully.

A 62-kDa protein (p62) was reported to be the ruffled border chloride channel based on DIDS affinity purification and reconstitution of chloride permeability into proteoliposomes (19). p64 is a chloride channel protein identified in bovine kidney microsomes by biochemical purification (31, 38). The similarity of the two proteins led us to investigate whether bovine p64 and avian p62 are related. Antibodies raised against bovine p64 (31) were used to probe osteoclast ruffled border proteins subjected to the DIDS affinity purification as described previously (14) and summarized in Table I (Fig. 3A). Lanes 1 and 2 show Coomassie Blue staining of the flow-through fraction and eluate from a DIDS column. p62 is prominent in the eluate lane. Lanes 3 and 4 show identical lanes that were blotted and probed with anti-bovine p64 antiserum (31). The antiserum weakly reacted with p62. To generate a more useful antiserum, sequences were obtained from chicken genomic clones encoding a homolog of p64 (see "Experimental Procedures"). A 16-amino acid peptide homologous to the extreme C terminus of p64 was used to raise a new antiserum, named 656. The 656 antiserum recognizes a LacZ-p64 fusion protein (Fig. 3B) and was affinity-purified by binding to this fusion protein immobilized on Sepharose (31).

Table I. Purification of the osteoclast ruffled membrane chloride channel Fractions were prepared, and 36Cl uptake assays were carried out as described (19). The activities represent passive, DIDS-inhibitable uptake. Ruffled membranes were assayed directly; other fractions were assayed following reconstitution in lipid vesicles as described (19). Fraction Specific activity Total activity Recovery Purification cpm/µg/min cpm/min % -fold Ruffled membranes 2.16 69,120 100 1 N-Octyl glucoside extract 14.8 65,120 94 6.9 DIDS column Flow-through 7 28,700 42 3.2 Eluate 2766 30,400 44 1281

Affinity-purified antibody 656 recognizes p62 both in crude ruffled border membranes and in DIDS affinity-purified material (Fig. 3C). In bovine kidney membranes, the antibody recognizes p64 as well as two smaller proteins of 33 and 31 kDa. The chicken p62 and bovine p64, p33, and p31 signals are all eliminated by preabsorption with bacterial lysate containing the LacZ-p64 fusion protein, but not with control lysate (Fig. 3D).

The 656 antiserum was used to probe chicken osteoclasts by immunogold electron microscopy as shown in Fig. 4. Fig. 4A is a low-power view showing an osteoclast attached to bone, with the cell body on the right and bone on the left, separated by the ruffled border. High-power views of an area of ruffled border from cells probed with the 656 antiserum (Fig. 4B) or preimmune serum (Fig. 4C) are shown. The highly folded nature of the ruffled border membrane combined with the low levels of p62 make the pattern of gold beads intermittent. Small clusters of two to four gold beads usually located on an oblique section of the membrane were found only with immune serum (arrowheads). These data were analyzed quantitatively by counting total and membrane-associated beads in 14 µm² of images, revealing a clear difference between immune and preimmune sera. In samples prepared with immune serum, 66 ± 8% (n = 3) of the beads were within 9 nm (one-bead diameter) of a membrane image, in contrast with 24 ± 7% (n = 3) in the preimmune serum-treated sample. This correlated with the 3-fold increase in beads/14 µm² in the immune serum-treated sample.

To determine whether the pattern of transport activities expressed by the differentiating bone marrow cells (Fig. 2) was reflected in different levels of expressed protein, membrane proteins from cells differentiated in vitro were separated by SDS-PAGE and probed with antibodies specific for the p64 chloride channel or the 70-kDa subunit of the H⁺-ATPase. p62 (Fig. 5A) is not detectable in day 1 cells or in 5-day no-bone cultures. p62 is only detectable in 5-day plus-bone cultures. In contrast, the 70-kDa ATPase subunit (Fig. 5B) is not detectable in day 1 precursor cells, but is easily detectable in 5-day cultures grown with or without bone. The apparent decrease in the 70-kDa pump subunit with the presence of bone may result from failure to release ruffled membranes completely from bone particles, which are discarded early in the preparation.

5

We also assessed expression of the chloride channel and of two subunits of the proton pump (the 70-kDa subunit and the "brain" isoform of the 58-kDa subunit) at the mRNA level by Northern blotting (Fig. 5, C-E). The chloride channel probe (Fig. 5C) reveals two major p64-related transcripts in the mature osteoclasts: one of 4.8 kb and one of 5.3 kb. The 4.8-kb transcript does not appear to be regulated during the terminal differentiation of osteoclasts: it is present in roughly equal abundance in mature osteoclasts, day 1 precursor cells, and 5-day cells grown in the presence or absence of bone. The 5.3-kb transcript does appear to be regulated during differentiation. It is absent in precursor cells, being replaced with a 5.6-kb transcript. By 5 days of culture, the precursor-specific 5.6-kb transcript has disappeared, and low levels of the 5.3-kb transcript are detectable. The 5.3-kb transcript is much more abundant in cells grown in the presence of bone. The in vitro differentiated cells express a second novel transcript of ~9 kb, which is also greatly enhanced by growth in the presence of bone. The probe to the 70-kDa proton pump subunit hybridizes to a single transcript of ~4 kb (Fig. 5D), and the probe to the bovine 56-kDa proton pump brain isoform hybridizes to a single major transcript of ~3.4 kb (Fig. 5E). Both these messages are present in day 1 cells and are markedly amplified as cells differentiate in culture in the absence of bone. The presence of bone has little effect on the level of expression.

To characterize the osteoclast chloride channel in more detail, we fused ruffled border membrane vesicles into a planar lipid bilayer. The resulting currents were studied using voltage-clamp methods. Typically, experiments were started with 140 mM CsCl, 5 mM KCl, and 10 mM HEPES (pH 7.0) in the trans-chamber and with 75 mM KCl, 75 mM potassium gluconate, and 10 mM HEPES (pH 7.0) in the cis-chamber, producing an E_K of 70 mV and an E_Cl of 17.5 mV. Introducing 6-30 µg (protein) of the ruffled border vesicle preparation into the cis-bilayer chamber with mixing resulted in the appearance of channel activity in the bilayer membrane. About 70% of the fusions produced anion channel activity, although other currents were also observed. Fig. 6A is a portion of a 40-s record with at least five distinct current transitions, with the prevalence of each transition shown. The transition having a conductance consistent with the ruffled border chloride channel (see below) is indicated. The chloride currents were enhanced by increasing the trans-solution to 280 mM CsCl, shifting the E_Cl to 35 mV. Fig. 6B shows that under these conditions, adding 1 mM BaCl₂ to the cis-chamber eliminates the large currents that reverse below 44 mV, leaving smaller currents that reverse between 22 and 44 mV, close to E_Cl. When Cs⁺ replaced K⁺ completely (cis-solution: 140 mM KCl; and trans-solution: 280 mM CsCl), potassium currents were eliminated, and the reversal potential equaled E_Cl (17.5 mV). The current at +50 or 50 mV was inhibited by the chloride channel inhibitor DNDS (Fig. 6C).

The anion currents were further characterized in preparations in which only single chloride channel transitions were present. A rectifying Cl current dominated these preparations. Fig. 6 (D-E) presents its current-voltage relationship recorded with 145 mM KCl in the cis-chamber and 140 mM CsCl + 5 mM KCl in the trans-chamber, producing an E_K of 70 mV and an E_Cl of 0. The current is strongly rectifying and has a reversal potential of 0 and a slope conductance of 25 picosiemens at positive potentials. At negative potentials, the current is small, and the channel is almost continuously open. Half-substitution of the chloride with gluconate or cyclamate in the cis-chamber varied the reversal potential to approximately 17.5 mV, as predicted for a chloride-selective channel, whereas substitution of sodium for potassium had no effect on the reversal potential, rectification, or slope conductance (data not shown). The vesicles used in these experiments have their cytoplasmic surface on the outside of the vesicle (20). After fusion with the bilayer, the cytoplasmic face will be on the cis-side of the membrane. Thus, the polarity of rectification is outward. In addition to the predominant anion channel characterized above, we also found Ba²⁺-sensitive K⁺ channels and a nonrectifying anion channel similar to activities previously reported to be in osteoclast plasma membrane (39, 40).

The chloride channel was purified from osteoclast ruffled border membranes using DIDS affinity chromatography as described previously (19) and outlined in Table I. The final preparation is dominated by a 62-kDa protein, although smaller amounts of other proteins are present (see Fig. 3A). This material was reconstituted into asolectin vesicles by detergent dialysis at a ratio of ~0.1-0.2 µg of protein/100 mg of lipid.

p62 phospholipid vesicles were fused into a planar bilayer, producing the currents shown in Fig. 7. Fig. 7A shows the voltage-driven current from a bilayer in symmetric 140 mM KCl at 40 mV after insertion of many channels. The multichannel current is completely inhibited by 200 µM DNDS applied in the trans-chamber. Preparations with single channel transitions were used to characterize this current further. Only one type of channel was observed using vesicles containing purified protein. Fig. 7B shows current recordings at a series of holding potentials for this channel, and Fig. 7C shows the current-voltage relationship averaged from three sets of data.

The activity associated with the purified protein is an anion-selective, DNDS-inhibitable, outwardly rectifying channel similar to the predominant activity in crude ruffled border vesicles. The slope conductance of the purified material is slightly lower than the channel activity in the crude membrane vesicles, being close to 15 picosiemens. The kinetics of purified channel closing appears much more rapid than that in the crude vesicles.

DISCUSSION

We have investigated the osteoclast ruffled border chloride channel and have demonstrated the following. 1) Expression of chloride channel activity is the critical step as osteoclasts acquire acid transport upon exposure to bone. 2) The ruffled border chloride channel is a homolog of the bovine kidney microsomal chloride channel, p64. 3) The appearance of ruffled border chloride channel activity is associated with the expression of novel osteoclast p64-related mRNA and protein. 4) The ruffled border chloride channel is a chloride-selective, DNDS-inhibitable, outwardly rectifying channel that copurifies on DIDS affinity chromatography with the 62-kDa protein previously identified as the ruffled border chloride channel. These observations define critical steps in osteoclast differentiation and have important implications for understanding osteoclast action.

Dissolution of the alkaline bone mineral hydroxylapatite requires low pH and consumes protons (3). To generate the acidic environment and the continued supply of protons necessary for bone mineral dissolution, the osteoclast creates an isolated compartment on the surface of bone and actively pumps HCl into this compartment (Fig. 1). The current model of osteoclast acid transport includes a chloride channel in the ruffled border membrane whose role is to short-circuit the electrical potential generated by the electrogenic proton pump.

The studies presented here further support the critical nature of the anion conductance to bone resorption. The chicken bone marrow mononuclear cells from calcium-deprived hens used in these studies, cultured in the absence of bone, differentiate into large multinucleated cells expressing osteoclast markers, but initially having a limited bone-resorbing capacity (23, 24). Upon exposure to devitalized bone, the bone-resorbing capacity of these cells rapidly increases. We have studied and compared the acid transport of vesicles isolated from mature osteoclasts and three types of bone marrow cells differentiating in culture (Fig. 2). Vesicles derived from day 1 precursor cells fail to acidify both in the absence and presence of valinomycin, indicating that they express low levels of the active proton pump. Vesicles from 5-day no-bone cells show valinomycin-dependent acidification, indicating the presence of an active proton pump without adequate active chloride channel to short-circuit the electrogenic pump. Vesicles from 5-day plus-bone cultures, like mature osteoclasts, can support valinomycin-independent acidification, indicating that these cells express high levels of the proton pump and adequate chloride channel to short-circuit the pump.

These functional studies are supported by observations on the expression of protein and mRNA specific to the proton pump and osteoclast chloride channel (Fig. 5). The day 1 precursor cells express undetectable levels of both the proton pump and chloride channel and produce poorly acidifying vesicles. The 5-day no-bone cells have increased expression of the 70- and 58-kDa proton pump subunits, but not the 62-kDa chloride channel, and the acidification of their vesicles is valinomycin-dependent. The 5-day plus-bone cells express both the proton pump and chloride channel, and the acidification of their vesicles is valinomycin-independent, like those from mature osteoclasts. Thus, it appears that expression of the chloride channel is a final step that confers bone-resorbing capacity to differentiating osteoclast precursors and that expression of this channel is induced by exposure to bone.

Expression of the ruffled border chloride channel is coincident with the induction of a novel osteoclast-specific p64-related transcript that may encode the ruffled border, p62. In contrast, expanded expression of at least the 70- and 58-kDa subunits of the proton pump does not involve production of new osteoclast-specific transcripts, but instead is the result of increased levels of transcripts that are present at low levels in the precursor cells. These data suggest that the ruffled border proton pump may be identical to the vacuolar proton pump present in most cells, whereas the ruffled border chloride channel is an osteoclast-specific protein expressed as a final step in developing the osteoclast phenotype and may be a useful target for drug therapy aimed at selectively inhibiting acid transport activity in osteoclasts.

These results demonstrate that the 62-kDa protein identified as the ruffled border chloride channel is antigenically related to the bovine microsomal chloride channel, p64. Antibodies raised against bovine p64 recognize p62. A second antiserum raised against a C-terminal epitope conserved in both the bovine and chicken sequences recognizes the same osteoclast protein localized to the ruffled border by immunogold electron microscopy (Figs. 3 and 4). The ruffled border membrane has proton transport functions similar to those of the lysosomal and endosomal membranes of many cell types, and it is not surprising that the ruffled border channel is a homolog of the intracellular chloride channel. It is clear that p62 is not simply the chicken version of bovine p64, but a fuller understanding of the relationship between osteoclast p62 and bovine p64 must await complete molecular cloning of the osteoclast transcript.

The above data support our hypothesis that bone resorption requires a specific ruffled border chloride channel activity and is limited by its absence. We propose that regulation of chloride conductance may be important in the modulation of acid transport as osteoclasts go through cycles of initiation and termination of active bone resorption.

In planar lipid bilayer studies of ruffled border vesicles, we characterized the single channel behavior of the most prevalent chloride channel present. We also studied channel activity of affinity-purified p62 reconstituted into vesicles. In both preparations, we found anion-selective (P_Cl/P_K > 10), rectifying channels of moderate conductance that are inhibited by DNDS. The substantial similarity between the crude and purified activities provides evidence that p62 in fact does account for the ruffled border chloride channel. However, there are subtle differences between the crude and purified channel activities, including more rapid channel closing and a lower single channel conductance in the purified channel. Whether these differences result from the removal of other unidentified subunits during the purification, from reconstitution in artificial lipids versus the natural lipid composition of ruffled border membranes, or from some other effect is unknown.

Outward rectification of a chloride channel means that the single channel conductance is greater for chloride entering the cell than for chloride exiting the cell. Since the proposed function of this channel is to allow chloride to exit the cell in parallel with protons, outward rectification may seem counter-intuitive. However, if one focuses on the single channel current at physiologic membrane potentials, the utility of outward rectification becomes apparent.

At membrane potentials below E_Cl, when Cl is moving out of the cell, the transport of protons and chloride is electrically coupled, and the chloride current across the ruffled border represents HCl transport (20). In situations where the chloride conductance might be limiting (Fig. 3), proton pump activity hyperpolarizes the ruffled border membrane, increasing the inside negative potential in proportion to the pump activity. If the ruffled border chloride channel were not rectifying, the proton pump-driven hyperpolarization would increase the chloride current and hence HCl transport. However, since the chloride channel is outwardly rectifying, changes in the membrane potential (below E_Cl) have a small effect upon the chloride current (see Figs. 6E and 7C) and acid transport. Therefore, during bone resorption, when the electrogenic proton pump hyperpolarizes the ruffled border membrane, the number of active chloride channels in that membrane will directly determine and limit HCl transport.