Concentrative Uptake of Cyclic ADP-ribose Generated by BST-1 (cid:1) Stroma Stimulates Proliferation of Human Hematopoietic Progenitors*

Cyclic ADP-ribose (cADPR) is an intracellular calcium mobilizer generated from NAD (cid:1) by the ADP-ribo-syl cyclases CD38 and BST-1. cADPR, both exogenously added and paracrinally produced by a CD38 (cid:1) feeder layer, has recently been demonstrated to stimulate the in vitro proliferation of human hemopoietic progenitors (HP) and also the in vivo expansion of hemopoietic stem cells. The low density of BST-1 expression on bone marrow (BM) the mean (cid:4) S.D. from three experiments. The cADPR concentration in the medium removed from the cultures was determined by a sensitive cycling assay (20) and was found to range between 0.15 and 0.30 (cid:2) M in the “micromolar” samples and between 1 and 3 n M in the “nanomolar” samples. These decreases reflect the presence of a low cADPR hydrolase activity expressed by the CD38 (cid:1) cells within the CD34 (cid:1) subpopulation.

The self-renewal capacity of hemopoietic progenitors (HP) 1 in the bone marrow (BM) is a fundamental process in the physiology of hemopoiesis. The role of the BM stroma in providing HP with soluble factors essential to their proliferation and differentiation is well established (1,2). However, the nature of the signals and the mechanisms by which stromal cells regulate the behavior of HP remain largely to be defined. Particularly, it is unknown whether perturbations of the balance between growth-stimulatory and -inhibitory factors may influence the expansion and self-renewal capacity of the hemopoietic reservoir.
Recently, we have demonstrated that the potent intracellular calcium mobilizer cyclic ADP-ribose (cADPR) (3,4) features properties of a novel hemopoietic growth factor (5,6). Coinfusion of human HP with murine stromal cells transfected with the human ectocellular ADP-ribosyl cyclase CD38 improves hemopoietic stem cells engraftment into NOD/SCID mice (7). The paracrine interaction between stroma and HP has been investigated in a transwell co-culture setting where human HP were cultured over confluent monolayers of murine stromal cells transfected with human CD38: NAD ϩ efflux from stromal cells through connexin 43-formed hemichannels (8) provides CD38 with its substrate for ectocellular production of cADPR; subsequent influx of the cyclic nucleotide into HP (both committed and uncommitted progenitors) induces intracellular calcium ([Ca 2ϩ ] i ) mobilization and cell proliferation (6). In a similar co-culture setting, influx of cADPR into target 3T3 fibroblasts and dimethyl sulfoxide (Me 2 SO)-differentiated HL-60 cells has been demonstrated to occur across concentrative nucleoside transporter(s) (CNT) sensitive to dipyridamole and nitrobenzylthioinosine (NBMPR) (9,10) and to induce an increase of the [Ca 2ϩ ] i and of proliferation (11).
While paracrine production of cADPR by the CD38 ϩ stroma was beneficial to HP growth, the increase of [Ca 2ϩ ] i induced on stromal cells themselves by autocrinally generated cADPR proved to trigger interferon-␥ (IFN-␥) release. This inhibited colony output during long term culture (LTC) of human HP over CD38 ϩ feeders (6).
The fact that native hemopoietic stroma expresses the ADPribosyl cyclase BST-1 instead of CD38 may be advantageous in the BM microenvironment; CD38 is a transmembrane, oligomeric, catalytically active transporter of cADPR (12), while BST-1 is a glycosylphosphatidylinositol-anchored ectoenzyme, possibly incapable of cADPR transport, and with a significantly lower specific activity compared with CD38 (13)(14)(15)(16)(17). Thus, while co-expression of CD38 with connexin 43 (as occurs in the transduced stromal cell lines) results in a significant (2-3-fold) increase of the [Ca 2ϩ ] i of CD38 ϩ stromal cells (18), the same may not hold true for BST-1. Accordingly, expression of BST-1 on the hematopoietic stroma might represent the critical feature ensuring paracrine production of HP-expanding cADPR, while limiting the extent of the [Ca 2ϩ ] i perturbation induced on stromal cells by autocrine cADPR.
The aims of this study were: (i) to compare the effect of LTC over CD38 ϩ versus BST-1 ϩ stroma on the in vitro clonogenic capacity of human HP and (ii) to investigate the mechanism of uptake of paracrinally produced cADPR by human HP.
Two BST-1-expressing stromal cell lines were generated and their biochemical characterization demonstrated distinctive properties as compared with the corresponding CD38-expressing cells. LTC of cord blood-derived HP over these BST-1 ϩ feeders significantly increased colony output compared with controls, co-cultured over cyclase negative feeders, in sharp contrast with the inhibitory effect induced by LTC over CD38 ϩ stroma. A concentrative transport of paracrinally generated cADPR into HP was found to be mediated by a NBMPR-and dipyridamole-sensitive CNT. These results are consistent with a role of BST-1-generated cADPR in the expansion of human HP in the bone marrow microenvironment.

EXPERIMENTAL PROCEDURES
Samples-Cord blood (CB) samples were obtained from umbilical and placental tissues scheduled for discard. CB-derived mononuclear cells (CB MNC) were isolated by centrifugation of the blood on Ficoll Paque Plus (Amersham Bioscience). CD34 ϩ cells (the MNC enriched in HP) were separated by immunoaffinity on magnetic columns (Miltenyi Biotec, Bergisch Gladbach, Germany), following the manufacturer's instructions, as described (5). Human-derived cells were maintained in glutamine-containing Dulbecco's modified Eagle's medium, supplemented with 10% fetal calf serum, penicillin (100 units/ml), and streptomycin (100 g/ml), in a humidified 5% CO 2 atmosphere at 37°C.
Transfection of 3T3 Cells with BST-1-The complete coding sequence of the murine full-length BST-1 cDNA was recovered as EcoR1 fragment (from PUC19/mBST-1 kindly provided by Dr. Toshio Hirano, Osaka, Japan; Ref. 17) and cloned in the same restriction site of the pcDNA3.1 expression plasmid. The plasmid obtained was sequenced on both strands of the cDNA to verify the presence of the correct insert. The pcDNA3.1/BST-1 construct was transfected into 3T3 cells with Lipofectamine Plus (Invitrogen, Milan, Italy), according to manufacturer's instructions. One day after gene transfer, Geneticin (1 mg/ml) was added to the culture medium, and transfected cells were cultured in antibiotic selection.
FACS Selection of COS-7 Cells for BST-1 Expression-The simian epithelial cells COS-7 naturally express BST-1 (see "Results"). Two cell populations, expressing low (BST-1 dim ) and high (BST-1 bright ) levels of BST-1 were selected from wild-type COS (COS wt ) by repeated FACS sorting, using a commercial anti-human BST-1 (CD157) antibody (see above) and a FITC-labeled anti-mouse IgG antibody (Sigma, Milan, Italy). Both monoclonal antibodies were applied under previously determined saturating conditions using the Indirect Immunofluorescence assay. Briefly, 2 l of anti-BST-1 monoclonal antibody were added to 1 ϫ 10 6 COS-7 cells in 100 l of phosphate-buffered saline (PBS) containing 0.1% bovine serum albumin (PBS-BSA) and cells were incubated for 30 min on ice, in the dark. Cells were then washed twice in PBS, and 150 l of the second monoclonal antibody (at a 1:75 dilution in PBS-BSA) was then added and cells were further incubated for 30 min. Finally, cells were washed twice and resuspended in PBS at a concentration of 1 ϫ 10 6 cells/ml. Deoxyribonuclease I (from bovine pancreas, Calbiochem Biochemicals, Nottingham, UK) and propidium iodide at 1 g/ml (Sigma Aldrich Fine Chemicals, Milan, Italy) were added to the cells, which were then filtered on a 70-m cell strainer (Falcon-BD) immediately before sorting. Control cells were exposed only to the secondary antibody.
Flow cytometric analysis and cell sorting were performed by using a FACS Vantage SE flow sorter (BD Biosciences Immunocytometry Systems, Palo Alto, CA) equipped with an air-cooled 50-megawatt argon ion laser. BST-1 ϩ COS-7 cells were gated within a viable, propidium iodidenegative, population using a Forward Scatter versus FL-3 logarithmic amplified channel dotplot. In each experiment, sorting gates were set to sort both the dimmest and the brightest BST-1 ϩ populations in the same two-way sorting experiment. A total of four sorting experiments were performed: between each round, the COS dim and COS bright populations were expanded in vitro.
Assay of ADP-ribosyl Cyclase Activity-Ectocellular ADP-ribosyl cyclase activity was measured by incubating 3 ϫ 10 6 cells in a total volume of 400 l of PBS-glucose (10 mM) with 0.1 mM NAD ϩ . At different times (0, 1, and 4 h), 100-l aliquots were withdrawn. Reactions were stopped by addition of 220 l of 0.9 M perchloric acid to each aliquot. After deproteinization, acid was removed and cADPR content was measured in each aliquot according to the enzymatic cycling assay (20). Protein determination was performed on an aliquot of the incubation (21).
Determination of the Intracellular Concentration of NAD ϩ and cADPR-BST-1 ϩ and control 3T3 cells (5 ϫ 10 6 ) were lysed at 4°C in 0.5 ml of 0.6 M perchloric acid. An aliquot was diluted 200-fold in PBS for NAD ϩ levels determination. The remaining volume was subjected to deproteinization, acid extraction, and NAD ϩ degradation, and cADPR content was determined by the enzymatic cycling assay (20). Intracellular NAD ϩ and cADPR content was expressed as picomole/mg of protein.
Determination of the [Ca 2ϩ ] i -Basal intracellular free calcium concentration ([Ca 2ϩ ] i ) was determined as described (19) on FURA 2AMloaded 3T3 fibroblasts and COS cells. For determination of the [Ca 2ϩ ] i on CD34 ϩ cells, aliquots of 2 ϫ 10 5 cells were incubated in a total volume of 1.0 ml, in the presence of 50 M cADPR, without or with 10 nM NBMPR. After 24 h, 10 M FURA 2-AM was added, and cells were further incubated for 60 min. Thereafter, cells were washed twice by centrifugation, resuspended in 0.3 ml of glucose-supplemented standard saline (5), and calcium measurements were performed at room temperature in a 0.5-ml cuvette. The basal [Ca 2ϩ ] i was determined as described previously (19).
Co-culture Experiments-All cell lines used as feeder layers were cultured in Dulbecco's modified Eagle's medium, as described (19). Transfected cells were maintained under Geneticin (1 mg/ml) selection in the same medium. After irradiation (1500 Gy), feeder cells were seeded into 24-well plates in complete medium, without Geneticin, and used for the co-culture experiments. Freshly separated CB MNC were seeded in transwells (Costar, Milan, Italy) at 2 ϫ 10 6 /well in LTC medium (Myelocult, Stem Cell Technologies, Vancouver, Canada) over irradiated, confluent (ϳ2 ϫ 10 6 cells) feeder layers. At weekly intervals, transwells were transferred over freshly irradiated feeders, cells were recovered from individual wells and counted, and aliquots were seeded in growth factor-supplemented semisolid medium to determine the content in clonogenic precursors (see below).
Semisolid Colony Growth Assay-After liquid culture, aliquots of 2 ϫ 10 4 MNC were seeded in growth factors-supplemented methylcellulose medium (Methocult, Stem Cell Technologies, Vancouver, British Columbia, Canada) and cultured at 37°C in a humidified, 5% CO 2 atmosphere, as described (5). After 14 days, colonies were scored and identified according to standard criteria. Colonies grown from cells preincubated for 1-7 days in liquid culture in the absence of added cytokine identify the committed HP, i.e. the colony-forming cells (CFC) (22). Conversely, colonies that develop from cells after 5 weeks of liquid culture identify the uncommitted HP, i.e. long term culture-initiating cells (LTC-IC) (23).

Development and Biochemical Characterization of BST-1 ϩ
Stromal Cell Lines-The fact that native hemopoietic stroma expresses BST-1 instead of CD38 prompted us to develop two different BST-1 ϩ stromal cell lines, to be used as feeders in co-culture experiments, aimed at establishing the effect of LTC over BST-1 ϩ stroma on HP growth. Murine 3T3 fibroblasts were transfected with the sense cDNA for murine BST-1, while simian COS-7 epithelial cells, which constitutively express BST-1 (Table I), were sorted into high (COS bright ) and low (COS dim ) expressing BST-1 ϩ subpopulations by FACS, using a commercially available monoclonal antibody against human BST-1 (see "Experimental Procedures").
Stable levels of BST-1 expression, as detected by enzymatic analysis, were obtained on 3T3 cells after five cycles of immunomagnetic selection. A clear-cut separation of BST-1 dim and BST-1 bright COS subpopulations was observed after four cycles of FACS selection (Fig. 1) and routinely checked thereafter, during the course of the co-culture experiments.
The limited increase of the [cADPR] i of BST-1 ϩ cells was not paralleled by any significant increase of the basal [Ca 2ϩ ] i compared with cyclase negative controls, neither in BST-1-transfected 3T3 fibroblasts nor in COS bright cells: conversely, the [Ca 2ϩ ] i of CD38 ϩ 3T3 was twice as high that of control values (Table I), as reported previously (11,18,19).
Expression of BST-1 did not induce any modification of the intracellular pyridine dinucleotide concentration ([NADϩ NADH] i ) of BST-1 ϩ cells (neither 3T3 nor COS), compared with the respective controls (3T3 cells transfected with the empty vector and unsorted COS cells, respectively) ( Table I). On the contrary, expression of CD38 in 3T3 fibroblasts reduced the [NADϩNADH] i to ϳ50% of control values (Table I), similarly to earlier results (19).
The cADPR concentration in the medium conditioned for 3 days by CD38 ϩ 3T3 was found to be in the subnanomolar range (0.51 Ϯ 0.14 nM, n ϭ 6), as detected by enzymatic cycling assay (20). In the medium conditioned by BST-1 ϩ 3T3 the [cADPR] was even lower (0.18 Ϯ 0.09 nM, n ϭ 5) and near the sensitivity threshold of the assay.
Previous observations had shown an increased, [Ca 2ϩ ] i -related IFN-␥ concentration in the medium conditioned by CD38 ϩ 3T3 feeders compared with controls (6); this prompted us to measure the [IFN-␥] in the medium of confluent monolayers of BST-1 ϩ 3T3 and of COS bright cells. No significant differences were observed between these cell lines and the corresponding controls (3T3 transfected with the empty vector and COS wt ), as measured with a colorimetric immunoassay (R&D Systems, Minneapolis, MN) on 7-day medium from confluent, irradiated feeder layers (6) (data not shown).
Long Term Culture of CB MNC over BST-1 ϩ Stroma-Long term co-culture in transwells of CB MNC over undiluted, CD38-transfected 3T3 fibroblasts proved to remarkably decrease their clonogenic activity. This effect was attributed to the autocrine cADPR-[Ca 2ϩ ] i loop in the CD38 ϩ 3T3, which induced overproduction of hemopoiesis-inhibiting IFN-␥ (6). To compare the effects of stromal BST-1 and CD38 on HP expansion, CB MNC were co-cultured in transwells over either BST-1 ϩ or CD38 ϩ 3T3 feeders and their clonogenic capacity was assayed weekly for up to 5 weeks (see "Experimental Procedures"). Controls were cultured over 3T3 fibroblasts transfected with the empty vector pcDNA 3.1.
As shown in Table II, the clonogenic activity, measured both as CFC frequency and total colony number, was significantly higher in CB MNC cultured over BST-1 ϩ feeders compared with controls, starting from the 2nd week of culture. Con-

BST-1 and CNT cADPR Influx Expand Hemopoietic Progenitors
versely, co-culture over CD38 ϩ feeders strongly inhibited colony production from the 1 st week of culture (Table II). Colony output of CB MNC cultured over 3T3 cells transfected with antisense CD38 was not significantly different from controls (data not shown). In previous experiments, addition of NAD ϩ glycohydrolase (NAD-ase) to the medium prevented the stimulatory effect of a short term co-culture over CD38 ϩ stroma on CFC output (6). This result was taken as an indication that release of NAD ϩ from the feeder cells provides the substrate for the extracellular generation of cADPR. Here we tested the effect of NAD-ase on the long term co-culture of CB MNC over BST-1 ϩ 3T3. Addition of NAD-ase (2 units/ml, twice weekly for 2 weeks) to the culture medium prevented the stimulation of colony output by the BST-1 ϩ feeder: colony frequencies (CFC/ 10 5 MNC seeded) were 22 Ϯ 3 versus 7 Ϯ 1 for BST-1 ϩ and control feeder, respectively, and 7 Ϯ 2 versus 6 Ϯ 1 for BST-1 ϩ and control feeder in the presence of NAD-ase (mean Ϯ S.D. from three experiments). The effect of long term co-culture over a BST-1 ϩ stroma on colony output was also investigated with the different COS feeders, COS wt (controls), COS bright and COS dim , expressing distinctively different levels of ectocyclase activity (Table I). COS bright feeders showed an improved long term supporting capacity over controls; colony output was higher from the 3rd week of co-culture compared with COS wt . Conversely, the clonogenic capacity of CB MNC cultured over COS dim feeders was markedly reduced compared with controls, again from the 3rd week onwards (Table IIIA). These results were confirmed by analysis of the cell expansion factor, calculated as the ratio between the number of cells seeded in semisolid medium and the number of cells harvested from the grown colonies (Table IIIB).
Nucleoside Transport Inhibitors Prevent the Stimulatory Effect of cADPR on Clonogenic HP-In 3T3 fibroblasts and in Me 2 SO-differentiated HL-60 cells (a human promyelocytic leukemia cell line), influx of extracellular cADPR is mediated by nucleoside transporters (NT), both equilibrative and concentrative (ENT and CNT, respectively) (9, 10). To establish whether these transporters are responsible for cADPR entry into human HP, CB MNC, i.e. the cells containing a significant fraction of HP, were exposed to either of the NT inhibitors dipyridamole and nitrobenzylthioinosine (NBMPR) prior to incubation with exogenously added cADPR; the schedule of cADPR treatment of CB MNC, a 24-h priming with 100 M cyclic nucleotide, was known to stimulate CFC output (5). Thus, dipyridamole and NBMPR were added to CB MNC in liquid culture at a final concentration of 100 nM and 10 nM, respectively, for 30 min.

TABLE III
Time course of colony output of CB MNC cultured over BST-1 ϩ COS cells CB MNC (2 ϫ 10 6 cells/well) were cultured in transwells over confluent monolayers of irradiated COS-7 cells selected by FACS for the level of BST-1 expression; wild-type (unselected) COS were used as a control. At weekly intervals, as indicated, cells from individual wells were harvested and seeded in growth factors-supplemented semisolid medium for 2 weeks, to allow for colony growth. Numbers represent median values for total colonies grown per well (A) and for the cell expansion factor (B), calculated as the ratio between the number of grown cells and the number of seeded cells (n ϭ 5). Bold numbers indicate a statistically significant difference (p Ͻ 0.02) compared to control (COS wt ). A, total colony output. B, cell expansion factor.  Thereafter, cADPR (100 M) was added, and the cells were cultured for 24 h. Aliquots of cells were then seeded in semisolid medium for 14 days, as described under "Experimental Procedures," for estimation of colony growth. Colony output was increased by 100 M cADPR (2200 Ϯ 190 CFC/10 6 MNC) compared with untreated, control cultures (1000 Ϯ 280 CFC/10 6 MNC; p Ͻ 0.04), and both NT inhibitors prevented the stimulatory effect of cADPR (800 Ϯ 200 and 1,100 Ϯ 230 CFC/10 6 MNC for 100 nM dipyridamole and 10 nM NBMPR, respectively; mean Ϯ S.D. from four experiments). These results indicated involvement of a dipyridamole, and NBMPR-sensitive NT, in the mechanism of entry of cADPR into the clonogenic HP (CFC); however, sensitivity to both inhibitors is a property shared by the equilibrative NT ENT-1 (not competent for cADPR transport) (9) and by the concentrative NTs cs (concentrative, sensitive) (26) and csg (concentrative, sensitive and guanosine-preferring) (27,28).
To investigate the possible role of concentrative NTs in mediating influx of cADPR into CFC, the effect of dipyridamole and NBMPR was assayed on a co-culture system, where CB MNC were overlaid in a transwell mode on stromal feeders, engineered to produce extracellular cADPR, i.e. 3T3 fibroblasts transfected with human CD38, sense (CD38 ϩ 3T3), and antisense (CD38 Ϫ 3T3) as negative control (19). The cADPR concentration in the medium conditioned by CD38 ϩ 3T3 was found to be in the subnanomolar range (see above), i.e. several orders of magnitude below the known K m of the equilibrative NT for cADPR, ENT2 (9, 10); this rules out any role of ENT2 in the influx of cADPR into the CFC, in close agreement with data obtained with Me 2 SO-differentiated HL-60 cells (10). After 24-h co-culture of CB MNC over CD38 ϩ/Ϫ 3T3 in the presence or absence of either dipyridamole or NBMPR, cells were seeded in semisolid medium to measure colony growth. As shown in Fig. 2, CFC output was ϳ2-fold higher for cells cultured for 24 h over CD38 ϩ feeders compared with controls, grown over CD38 Ϫ 3T3 (1,600 Ϯ 200 CFC/10 6 MNC versus 900 Ϯ 150, mean Ϯ S.D. from five experiments; p Ͻ 0.04). The presence of dipyridamole or NBMPR completely prevented the stimulatory effect of the co-culture over CD38 ϩ 3T3 (Fig. 2). Similar results were obtained using BST-1 ϩ 3T3 as feeder; CB MNC were co-cultured for 2 weeks over BST-1 ϩ or BST-1 Ϫ (control) 3T3, in the presence or absence of 10 nM NBMPR (added twice weekly) and then seeded in semisolid medium for colony growth. CFC frequency (CFC/10 6 MNC) increased after co-culture over BST-1 ϩ 3T3 as compared with controls (220 Ϯ 36 versus 70 Ϯ 18, mean Ϯ S.D. from three experiments; p Ͻ 0.06). In the presence of 10 nM NBMPR the stimulatory effect of the BST-1 ϩ feeder was abrogated (50 Ϯ 11 versus 58 Ϯ 15). Thus, the effect of paracrinally produced cADPR on CFC appears to be mediated by a concentrative NT, sensitive to nanomolar concentrations of the NT inhibitors dipyridamole and NBMPR, suggesting involvement of cs and/or csg (9,10).
A role of accessory mononuclear cells (e.g. via an increased production of cytokines) in the stimulatory effect of exogenously added cADPR on HP has been already ruled out, because it was observed on total MNC as well as on the CD34 ϩ subpopulation, which is markedly enriched in both the committed and early HP (5). To determine whether the uptake of cADPR by purified CD34 ϩ cells was also mediated by a dipyridamole and NBMPR-sensitive CNT, we isolated CD34 ϩ cells from CB MNC and incubated the purified subpopulation with exogenously added cADPR in the presence or absence of NBMPR. Steady presence of nanomolar cADPR, as measured following daily replacement of the medium with fresh medium supplemented with 20 nM cyclic nucleotide, proved as effective as low micromolar cADPR in stimulating colony output from CD34 ϩ cells (Fig. 3). The increase of colony number relative to control cultures was more evident at longer incubation times (1-2 weeks) than observed with 100 M cADPR, which caused doubling of colony output after 24 h incubation (5). As the long term culture (Ն1 week) of CD34 ϩ cells with the nucleoside transport inhibitor NBMPR proved cytotoxic, we decided to use a cADPR concentration in the order of tens of micromolar to observe a significant increase of colony output after 24 h incubation. Thus, cADPR priming of CD34 ϩ cells with 50 M cyclic nucleotide for 24 h induced a 2-fold increase of CFC (Fig. 4A) and a 3-fold increase of the [Ca 2ϩ ] i , compared with controls (Fig. 4B). Addition of 10 nM NBMPR prevented both effects of cADPR, reducing colony output and [Ca 2ϩ ] i to control values. NBMPR alone did not significantly modify colony growth or [Ca 2ϩ ] i in CD34 ϩ cells (Fig. 4, A and B). DISCUSSION Native hemopoietic stroma expresses the ADP-ribosyl cyclase BST-1 (13). Autocrine production of cADPR in BST-1 ϩ cADPR. Two-thirds of the medium were replaced daily with fresh medium. At the times indicated, aliquots of cells were seeded in duplicate in semisolid medium to evaluate colony growth. Results shown are the mean Ϯ S.D. from three experiments. The cADPR concentration in the medium removed from the cultures was determined by a sensitive cycling assay (20) and was found to range between 0.15 and 0.30 M in the "micromolar" samples and between 1 and 3 nM in the "nanomolar" samples. These decreases reflect the presence of a low cADPR hydrolase activity expressed by the CD38 ϩ cells within the CD34 ϩ subpopulation. cells should be significantly lower than in CD38 ϩ cells due to the low specific activity of BST-1 compared with CD38. Indeed, the limited increase of the [cADPR] i in BST-1 ϩ stromal cell lines compared with the corresponding controls did not induce significant modifications of the [Ca 2ϩ ] i (Table I) and consequently of the IFN-␥ production (see "Results"). During LTC of HP over CD38 ϩ stroma, growth inhibition by IFN-␥ overcomes the stimulatory effect of cADPR, resulting in a marked reduction of colony output compared with control co-cultures, set over CD38 Ϫ stroma (Table II). Conversely, LTC over BST-1 ϩ 3T3 significantly increases colony output, starting from the 2nd week of co-culture (Table II). The difference in colony output between the two co-culture settings (over BST-1 ϩ versus control stroma) increases during the course of LTC. This observation is confirmed by similar results obtained with the other BST-1 ϩ stromal cell line used in this study, COS cells (Table  III). Thus, it appears that the extracellular cADPR concentration generated by BST-1 ϩ feeders, although in a subnanomolar range, is sufficient to exert a stimulatory effect on CFC proliferation during long term culture. This conclusion prompted us to investigate the mechanism of cADPR uptake by CFC.
Dipyridamole and NBMPR inhibit the stimulatory effect of cADPR on CFC at nanomolar concentrations (Fig. 2); since ENT-1 has been shown to be incompetent for cADPR transport (9), this points to the NBMPR/dipyridamole-sensitive CNTs cs and csg as responsible for cADPR entry into CFC. A further consideration suggests a causal role for concentrative transporter(s) in the influx of cADPR into these hemopoietic progenitors; the extremely low cADPR concentration in the medium conditioned by both CD38 ϩ and BST-1 ϩ 3T3, which is in the sub-nanomolar range, is several orders of magnitude below the known K m of the equilibrative transporter (ENT2) for cADPR, while they are still lower than, but closer to, the K m of the CNTs for cADPR (9,10). The intracellular cADPR concentration ([cADPR] i ) in CD38 Ϫ 3T3 co-cultured over a CD38 ϩ feeder is 10 3 times higher compared with the extracellular level, indicating that a concentrative transport occurs (11). The finding that NBMPR prevents the stimulatory effect of cADPR on colony output and on [Ca 2ϩ ] i increase on CD34 ϩ cells confirms the involvement of a concentrative NT in cADPR transport into primitive HP (Fig. 4).
The fact that a concentrative NT mediates influx of cADPR into CFC is in line with the presence, in the bone marrow microenvironment, of a low extracellular cyclase activity, as that displayed by BST-1. Two considerations suggest that the [cADPR] e in the BM microenvironment may be even lower than that recorded in the medium conditioned by CD38 ϩ feeder layers: (i) expression of BST-1, instead of CD38, on stromal mesenchymal cells (MSC) and (ii) dilution of BST-1 ϩ MSC with other cyclase-negative cell types. The hormone-like concentration of cADPR provided by the BST-1 ϩ feeder might cooperate with other intracellular calcium-releasing signal molecules (inositol 1,4,5-triphosphate and NAADP ϩ ) (31-33) to induce growth-stimulatory Ca 2ϩ signals in HP. In CD38 ϩ 3T3 fibroblasts, cADPR was found to cooperate with inositol 1,4,5triphosphate in the generation and propagation of Ca 2ϩ signals following stimulation of P2Y purinergic receptors (34).
In conclusion, these results indicate that the continuous production of hormone-like concentrations of extracellular cADPR by a BST-1-positive stroma stimulates LTC-IC proliferation in vitro. Doubling of colony output can also be obtained by the continuous, 2-week-long exposure to nanomolar, exogenously added cADPR (Fig. 3), or to micromolar pulse-added cADPR (Fig. 4A). This seems to be a recurring motif in cytokine function; the continuous supply of low concentrations of stimulatory cytokines reduces their effective concentration by 2-3 logs, compared with that required to produce similar effects by pulse addition (35). Similarly, the steady release of growthinhibitory IFN-␥ by a transduced feeder has been shown to substantially reduce the effective concentration of the cytokine as compared with its direct addition to the medium (36).
In this study, the concentrative transport of cADPR into HP is apparently responsible for the efficacy of subnanomolar extracellular cADPR concentrations in eliciting a biological effect on HP. Expression of BST-1, instead of CD38, on stromal cells is advantageous, as it reduces the autocrine effects of cADPR on stroma, preventing the production of IFN-␥ while producing enough extracellular cADPR to elicit a stimulatory effect on HP growth.