Tumor Necrosis Factor α and Adiponectin in Bone Marrow Interstitial Fluid from Patients with Acute Myeloid Leukemia Inhibit Normal Hematopoiesis

Purpose: Locally residing cytokines may inhibit bone marrow hematopoiesis in acute myeloid leukemia (AML). Using a novel method to isolate bone marrow interstitial fluid, we examined if this fluid from 10 adult AML patients could affect normal bone marrow hematopoiesis. Experimental Design: Bone marrow interstitial fluid was isolated by centrifugation of bone marrow biopsies obtained at time of diagnosis and 2 to 4 weeks after start of induction therapy. The isolated fluid was added to normal bone marrow CD34 hematopoietic progenitor cells sampled from five healthy subjects. Results: Unlike plasma, AML-derived bone marrow interstitial fluid clearly repressed hematopoietic progenitor cell growth as determined by an in vitro colony assay, an effect that was lost after successful induction treatment. Antibodies against tumor necrosis factor α (TNFα) and adiponectin abolished growth inhibition by bone marrow interstitial fluid, suggesting a mechanistic role of these cytokines in impairing normal hematopoiesis in AML. The plasma levels of adiponectin and TNFα were unaffected by therapy whereas bone marrow interstitial fluid levels of both cytokines fell significantly in patients entering remission. Transcripts for TNFα, but not for adiponectin, were found in AML blast cells. Neither the plasma levels nor the bone marrow interstitial fluid levels of the proangiogenic factors vascular endothelial growth factor or basic fibroblast growth factor were appreciably elevated in the patients nor did they change with treatment. Conclusions: Specific analyses of bone marrow interstitial fluid may give novel information on normal and malignant hematopoietic activity and thus form the basis for mechanism-based therapy.

Acute myeloid leukemia (AML) is a hematologic malignant disorder that usually carries a poor prognosis in adults. Potentially curative treatment consists of intensive chemotherapy, and, in addition, selected patients are offered allogeneic stem cell transplantation. However, even in patients receiving such therapy, the prognosis remains dismal with the median 5-year survival being usually below 60% in most studies, irrespective of specific cytogenetic abnormalities (1,2).
In addition to the high mortality rate among AML patients, the intensive therapeutic regimens are associated with severe side effects, affecting particularly the bone marrow. Both the underlying leukemic disease as well as most cytotoxic drugs invariably exert inhibitory effects on normal bone marrow hematopoietic activity. Consequently, most AML patients are prone to anemia, infections, and bleeding.
Studies suggest that certain cytokines might have negative effects on normal hematopoietic cells; for example, the elevation of proinflammatory cytokine tumor necrosis factor a (TNFa) in certain leukemias and depression of normal bone marrow hematopoiesis (3 -5). Furthermore, recent data indicate that the adipocyte-derived cytokine adiponectin might selectively hamper growth of human myelomonocytic progenitor cells as well as inhibit murine B-lymphopoiesis. Moreover, in vitro studies suggest that adiponectin is highly expressed and secreted in the murine bone marrow (6 -8).
Several lines of evidence point to the possible role of enhanced bone marrow angiogenesis in promoting the leukemic process (9,10). Reportedly, bone marrow microvessel density was a predictor of poor prognosis, and conversely, the extent of newly formed bone marrow microvessels correlated with poor response to treatment (11). Moreover, plasma levels of the surrogate markers of angiogenesis, vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (b-FGF), were associated with poor outcome in AML patients (12 -14). However, the precise role of these cytokines within the bone marrow microenvironment during AML merits further attention.
A hitherto unresolved problem is whether the obtained plasma/serum concentrations of cytokines and other signaling molecules reliably reflect the biological relevant concentrations in the interstitial compartment surrounding the target tissues. For example, we recently showed marked differences in the concentrations of proinflammatory cytokines, like TNFa, in interstitial fluid of inflamed rat skin compared with those in the systemic circulation (15). Hence, to examine in more detail changes in the bone marrow microenvironment with special emphasis on changes in the levels of signaling molecules like cytokines, we have developed a new method for sampling of interstitial fluid from biopsies of human and rat bone marrows (16). We now examined if signaling substances found locally in bone marrow interstitial fluid (BM-IF) from AML patients might hamper normal bone marrow hematopoiesis. If so, we wanted to test specifically whether AML-derived BM-IF contained elevated levels of TNFa and/or adiponectin compared with normal BM-IF. Finally, we also asked if the proangiogenic factors VEGF and b-FGF in newly diagnosed AML patients could be found in their BM-IF, and whether changes in the levels of these cytokines, during the initial treatment period with intensive courses of induction chemotherapy, could be detected.

Materials and Methods
Study subjects and blood and biopsy sampling. The protocol was approved by the Regional Ethics Committee for Medical Research, Health Region II in Norway and conformed to the Declaration of Helsinki. Written consent was obtained from all study subjects. We studied 10 patients with newly diagnosed and previously untreated AML as defined according to the French-American-British classification (17). None of these patients had any previous malignancy or chronic illness, or used any medication. Relevant patient characteristics are outlined in Table 1.
From each study subject we obtained blood samples (5 mL) by puncture of an antecubital vein. One or two cylindrical bone marrow biopsies were sampled from one or both iliac crests after infiltration of the overlying skin and periost with local anesthesia (5 mg/mL; Xylocain, AstraZenica, London, United Kingdom). From the AML patients, blood and bone marrow specimen were collected at the time of diagnosis and 2 to 4 weeks after start of induction chemotherapy. Additional sampling was done after a second course of chemotherapy in those patients not reaching complete hematologic remission following the first induction course.
Treatment and remission evaluation. All 10 patients underwent an initial induction chemotherapy course consisting of a combination of cytarabine (200 mg/m 2 /d for 7 days) and daunorubicin (45 mg/m 2 /d for 3 days; ref. 18). Patients not entering complete hematologic remission after this first induction course received a combination of cytarabine (200 mg/m 2 /d for 5 days), amsacrine (150 mg/m 2 /d for 5 days), and etoposide (110 mg/m 2 /d for 5 days). No central nervous system prophylaxis was given.
When necessary, intercurrent infections were treated with various antibiotics, most frequently a combination of penicillin and an aminoglycoside. The patients routinely received erythrocyte or platelet transfusions if the hemoglobin concentration fell below 8 g/100 mL or if the platelet concentration decreased below 20 Â 10 9 /L, respectively.
The effect of therapy was evaluated after 2 to 4 weeks, and complete hematologic remission was defined as a platelet concentration > 100 Â 10 9 /L, a neutrophilic granulocyte concentration > 1 Â 10 9 /L, and a near-normocellular bone marrow with <5% blast cells.
Isolation of bone marrow interstitial fluid. We used our previously detailed centrifugation technique to obtain samples of interstitial fluid from the bone marrow biopsies (16,19,20). Briefly, the biopsy was wrapped in a nylon mesh (pore size, 25 Am) and placed in 2 mL centrifuge tubes. These samples were then centrifuged for 10 minutes at 1,500 rpm (239 Â g). Usually, 10 to 15 AL of BM-IF were obtained per biopsy.
Determination of colony numbers. We collected blood samples and bone marrow aspirates from five healthy adult males using the same procedures as outlined above. The mononuclear cells were isolated from the specimen with ammonium chloride -induced hemolysis followed by density centrifugation (Lymphoprep, Nycomed, Oslo, Norway). More than 97% of the sorted cells were viable when assayed with the trypan blue membrane exclusion test. For colony assays, hematopoietic progenitor cells were isolated from this cell suspension with a magnetic sorting procedure using an anti-CD34 monoclonal antibody (MACS, Miltenyi Biotech, Bergish Gladbach, Germany). We then added AML-derived BM-IF (7 AL) to the CD34 cells immediately before plating the colony cultures. The total number of short-term (14 days) formations of normal CD34-induced colonies was recorded after plating the cells on semisolid medium (MethoCult System, StemCell Technologies, Vancouver, BC, Canada), and the morphologic subtypes were established after supravital staining (21,22). To determine the total number of colonies originating from long-term (6-8 weeks) colony-initiating cells, the sorted CD34 cells were first grown in liquid medium, then in semisolid medium for the last 2 weeks (MyeloCult System, StemCell Technologies).
We measured the in vitro growth of the normal CD34 cells by culturing colonies in the presence or absence of TNFa, adiponectin, an anti-TNFa monoclonal antibody, or an antiadiponectin polyclonal antibody (R&D Systems, Minneapolis, MN; ref. 23).
Measurement of cytokines. Due to the limited size of BM-IF, these samples were diluted, usually 1:100 or 1:50, before analyses of the protein contents. TNFa, adiponectin, VEGF, and b-FGF in serum and in BM-IF were measured with commercial ELISA kits (Quantikine, R&D Systems). For VEGF, this assay is specific for the VEGF 165 isoform. All plates were precoated with the relevant cytokine antibodies, and controls, standards, and experimental samples were added. The washing and adding of antibodies, substrate solutions, and stop solutions were done according to the specifications of the manufacturer. For each well, the absorbance was read using a microplate reader (SpectraMax Plus, Molecular Devices, Sunnyvale, CA). The detection limits were 0.30 pg/mL for TNFa, 0.25 ng/mL for adiponectin, 5 pg/mL for VEGF, and 3 pg/mL for b-FGF. Due to limited sample volumes of BM-IF, analyses of VEGF and b-FGF were done in six patients.
We used an RNase protection assay to detect gene expression of TNFa and adiponectin, as previously described (24). mRNA for the two cytokines was assayed in extracts from bone marrow -derived blast cells (>95% purity) sampled from the AML patients and from primary human osteoblasts (Cambrex, Walkersville, MD), as well as from very immature CD34+CD38À bone marrow cells from healthy subjects.
Statistics. The various experiments were done in triplicates for each study subject and the corresponding medians were used to calculate the mean and SE for the group of study subjects. Differences were evaluated with a two-tailed Wilcoxon signed-rank test for paired samples and were considered significant for P < 0.05.

Results
Acute myeloid leukemia bone marrow interstitial fluid represses hematopoietic progenitor cell growth. We used a colony assay to determine the effect on growth of normal hematopoietic progenitor cells on addition of BM-IF samples from AML patients. Colonies were cultured at diagnosis (i.e., before treatment) as well as 2 to 4 weeks after start of induction chemotherapy (i.e., at the time of remission evaluation). Figure 1A and B shows that addition of BM-IF from all 10 AML patients markedly attenuated both short-and long-term colony growth before the induction treatment commenced. This indicates that both the more mature and partly lineagerestricted colony-forming cells (short-term colony formation), as well as the very immature, uncommitted, long-term colonyinitiating cells, were negatively affected on addition of BM-IF from the AML patients. The fractions of morphologic colony subtypes remained unaltered on induction chemotherapy ( Table 2). Addition of AML-derived plasma to cultures of normal CD34 cells had no effect on colony numbers (data not shown). Of the 10 included AML patients, 7 entered complete hematologic remission. BM-IF sampled from these seven patients at the time of remission evaluation had apparently lost most of the inhibitory effect on colony growth. In contrast, BM-IF from the three remaining patients not entering complete hematologic remission retained its growth inhibitory effect ( Fig. 1A and B). These three AML patients, who all had normal karyotypes, were then offered a second course of induction chemotherapy which brought two patients into complete hematologic remission. Scoring of short-and long-term colony numbers after this second induction course revealed that the growth inhibitory effect of BM-IF was markedly reduced in these two patients, but remained virtually unchanged in the single patient not responding to therapy (Table 3). Notably, the percentages of bone marrow blast cells at the time of diagnosis in the seven AML patients entering complete hematologic remission after the first induction course was lower (range, 33-57%) than the corresponding values obtained in the three AML patients not entering complete hematologic remission (range, 61-90%).
Analyses of single colonies revealed no change in either the normal karyotypes or the surface expression of a panel of CD antigens (data not shown), indicating that addition of BM-IF had not driven the cultured normal CD34 cells into a malignant phenotype or genotype.
Tumor necrosis factor a and adiponectin inhibit growth of normal hematopoietic progenitor cells. To examine whether the growth inhibitory effect of AML-derived BM-IF could be  explained by elevated levels of TNFa and adiponectin, normal CD34 cells were plated with BM-IF from untreated AML patients, and in the absence or presence of neutralizing antibodies raised against these two cytokines. It is evident from Fig. 2 that inhibition of either cytokine markedly stimulated growth of both short-and long-term colonies, and a significant, synergistic effect was noted when both antibodies were administered simultaneously. Moreover, this combined treatment with both anti-TNFa and antiadiponectin antibodies increased the colony numbers to values comparable to those scored after addition of posttreatment BM-IF (Fig. 1). Furthermore, there were no apparent differences in the increase in colony numbers on antibody blocking between the seven AML patients that entered complete hematologic remission after the first induction course compared with the three patients that did not achieve complete hematologic remission. Importantly, addition of an irrelevant immunoglobulin G antibody did not affect colony formation. We next measured the concentrations of TNFa and adiponectin in plasma and in BM-IF obtained at diagnosis and at the time of remission evaluation. Whereas no apparent changes were detectable for the plasma concentrations, the BM-IF concentrations of both cytokines clearly diminished on induction therapy in the seven AML patients entering complete hematologic remission, but not in the remaining three patients (Fig. 3A and B; Table 3). This decrease in BM-IF concentrations of the two cytokines, to about 1/4 of the pretreatment values, correlated with the 3-to 4-fold increase from pretreatment to posttreatment in colony numbers (Fig. 1). Strikingly, the concentrations of both TNFa and adiponectin in the BM-IF were markedly decreased in the two patients entering complete hematologic remission following the second course of induction therapy, but not in the therapy-resistant patient (Table 3).
In search of the cellular source of TNFa and adiponectin, an RNase protection assay was used to determine the gene expression of these two cytokines. We could readily detect TNFa mRNA in AML blast cells (Fig. 3C) and TNFa protein was found in culture medium from these cells (data not shown). Adiponectin transcripts were not found in the blast cells, but the human osteoblasts showed a marked gene expression of adiponectin, as expected ( Fig. 3C; ref. 6). In contrast, neither TNFa nor adiponectin transcripts could be detected in the very immature CD34+CD38À bone marrow cells from healthy subjects (Fig. 3D).
In Fig. 4, we have shown the colony numbers scored after single additions of TNFa, adiponectin, their respective antibodies, and combinations thereof to normal bone marrow CD34 cells. As expected, TNFa exerted a marked inhibitory effect on colony growth and this could be inhibited by addition of the anti-TNFa antibody. Interestingly, adiponectin also induced a strong inhibitory effect on colony growth, an effect that was reversed on addition of the antiadiponectin antibody. Furthermore, a substantial depression of colony numbers was obtained when a combination of TNFa and adiponectin was given to the culturing cells, which could be inhibited with simultaneous addition of both the anti-TNFa and the antiadiponectin antibodies. No change in the relative fractions of morphologic subtypes of colonies could be detected on antibody blocking (data not shown).   No change in acute myeloid leukemia bone marrow interstitial fluid concentrations of proangiogenic factors. The concentrations of the proangiogenic factors VEGF and b-FGF in plasma and in BM-IF apparently did not change in the AML patients who reached complete hematologic remission at the time of remission evaluation compared with measurements done at diagnosis. Whereas VEGF was undetectable in BM-IF samples obtained at diagnosis, it was 1.7 F 1.1 pg/mL (n = 6) after the induction course. The concentrations of VEGF in plasma before and after the induction course were 1.92 F 1.2 and 1.5 F 0.6 pg/ mL (P > 0.05), respectively. Whereas we could not detect b-FGF in BM-IF sampled either at diagnosis or after the induction course, the plasma concentrations of b-FGF obtained before and after induction chemotherapy were 0.008 F 0.005 and 0.007 F 0.004 pg/mL (P > 0.05), respectively. We could not detect VEGF or b-FGF in plasma or in BM-IF of normal subjects.

Discussion
The main and novel finding of the present study was the inhibitory effect of AML-derived BM-IF on colony formation of normal bone marrow -derived CD34 progenitor cells. This decrease in bone marrow hematopoiesis was, at least, partly the result of the cytokines TNFa and adiponectin because BM-IF of AML patients contained elevated levels of both cytokines, and addition of neutralizing antibodies blocking specifically the effects of either cytokine efficiently abrogated their growth inhibitory effects. Furthermore, the elevated BM-IF concentrations of these two cytokines probably reflected local production and/or secretion from the bone marrow itself because (a) bone marrow -derived blast cells from AML patients, but not very immature progenitor cells from healthy subjects, expressed mRNA for TNFa and primary human osteoblasts expressed mRNA for adiponectin; (b) the plasma concentration of either cytokine was not appreciably changed compared with healthy subjects; and (c) the concentrations of both TNFa and adiponectin in BM-IF decreased in all seven AML patients reaching complete hematologic remission. Hence, both cytokines seemed to be involved in the repression of normal bone marrow hematopoiesis among our AML patients. Moreover, because the BM-IF concentrations of TNFa and adiponectin, but not the plasma concentrations, decreased in responsive, but not in therapy-resistant, AML patients, BM-IF concentrations of these two cytokines apparently provide more reliable prognostic information than concentrations obtained from the systemic circulation. Notably, we specifically studied hematopoiesis of a small number of AML patients having elevated TNFa and adiponectin concentrations in their BM-IF, and hence cannot generalize to the entire population of AML patients, particularly because the incidence of elevated BM-IF concentrations of these two cytokines among AML patients is unknown. Furthermore, we chose to study two cytokines that reportedly could impair the formation of blood cells. Whether the BM-IF harbors other inhibitors of hematopoiesis among the possible hundreds of different compounds is not known.
An important question is whether the isolated fluid is representative for the fluid bathing the bone marrow cells (i.e., true BM-IF). Importantly, the BM-IF of neither the controls nor the AML patients contained any cells. We used a recently developed centrifugation method to isolate BM-IF, a method that was validated in normal rats and humans (16). Hence, most likely, the BM-IF from AML patients yielded representative samples of the interstitial fluid from the bone marrow microenvironment.  A major problem associated with AML patients and the intensive chemotherapy regimens offered to them is the invariable depression of normal bone marrow hematopoiesis. Hence, these patients gradually lose their ability to sustain a normal release of mature blood cells and platelets. The microenvironmental changes that occur within the bone marrow compartment during initiation and progression of AML, as well as those appearing during induction of cytostatic treatment, have not been satisfactorily clarified. We have previously shown that blood flow to the hematopoietic bone marrow in rats varied according to the bone marrow metabolism (25). In a rat model of human AML, we further showed that the bone marrow perfusion gradually decreased during leukemic development (26). Therefore, the concomitant inhibition of bone marrow hematopoiesis could at least partly be attributed to a declining blood supply. Moreover, in the leukemic rats, a combined hypoxia and acidity clearly hampered the normal bone marrow metabolic activity, leading to a profound decrease in the proliferative capacity among normal hematopoietic cells (27). Although the DNA-synthesizing fraction of leukemic cells also decreased as the bone marrow microenvironment changed, a subpopulation of leukemic cells retained their proliferative capacity, suggesting that they could renew and expand the leukemic cell load even during marked anaerobic conditions within the marrow microenvironment (28). Intriguingly, a spatial localization of normal hematopoietic stem cells has been proposed, leading to the notion that certain endosteal niches selectively harbor these very immature cells (29). Whether this also applies to leukemic stem cells remains to be investigated, and the possibility to now sample interstitial fluid from the bone marrow might serve as a useful approach in this context.
Several studies, both in rodent models of AML and human AML, have implicated angiogenesis as a pathogenic factor in leukemogenic process (11,12,30,31). In contrast to others, we could not detect any specific pattern on neither the plasma nor the BM-IF concentrations of the angiogenic surrogate markers VEGF or b-FGF in our AML patients. Although our values for these surrogate markers were in the low range in relation to the sensitivity of the assay, these data still reflect the heterogeneous nature of AML because not all patients in previous studies have increased angiogenesis in their bone marrows (9,11). We found a variable expression of the transcripts for both VEGF and b-FGF in primary AML blasts retrieved from the bone marrow of our patients (data not shown). Hence, it was not possible to ascertain whether any of these two angiogenic factors primarily stem from these cells or some other cells within the bone marrow. Importantly, a variable expression of b-FGF as well as of VEGF has been observed by others (12,13). Furthermore, it is known that platelets are rich in VEGF and b-FGF (32,33). The fact that we found low platelet counts in some of our patients may thus partly explain the low VEGF and b-FGF levels. The mRNA expression of both VEGF and b-FGF in isolated leukemic blasts from the bone marrow was variable (data not shown). Whether other cell types might contribute a local production of these angiogenic factors within the bone marrow microenvironment is not known. Moreover, a more direct estimate of angiogenesis, like bone marrow microvessel density, was not determined for our AML patients.
TNFa negatively affected normal hematopoiesis in several leukemic disorders including adult AML and juvenile myelomonocytic leukemia (3 -5). Our present study confirms and extends these previous observations. In particular, we could now directly show biologically active TNFa and adiponectin proteins in AML-derived BM-IF (i.e., in the liquid phase immediately adjacent to the hematopoietic progenitor cells). The concentration of TNFa in BM-IF was markedly lower than that used to study the effect of exogenous TNFa on colony growth. Yet, the TNFa in BM-IF substantially inhibited colony numbers because this effect was markedly impaired on blocking with the anti-TNFa antibody. Possibly, TNFa synergizes with other, yet unidentified, compounds in the BM-IF to inhibit blood cell formation.
An important novel finding was the increase in AML-derived BM-IF concentration of adiponectin. We now provide evidence that a locally dysregulated adiponectin synthesis and/or secretion resulting in increased BM-IF concentrations might act in concert with TNFa to inhibit normal bone marrow hematopoiesis. Fat occupies a considerable fraction of the bone marrow, and the adipocyte might modulate hematopoiesis through energy supply and/or secretion of compounds with hematopoietic activity (e.g., leptin). Adiponectin is one of the most abundant cytokines produced and released from adipocytes (34 -36). In addition to its role in cardiovascular disease and the metabolic syndrome, emerging evidence suggests that adiponectin can impair growth and differentiation of immature cells of the lymphoid and myeloid lineages, hence corroborating the present findings (7,8,36). Although we cannot delineate the molecular mechanism of how TNFa and adiponectin mediate their growth inhibitory effects on hematopoeitic progenitor cells, these two cytokines share some similarities in structure, and intracellular signaling molecules, such as the transcription factor complex nuclear factor nB, are influenced by both, indicating a possible common operating mechanism (37,38). Notably, we previously showed that the impaired hematopoiesis observed in a murine model of heart failure was at least partly due to activation of the TNFa/Fas apoptosis system (22). However, whether this death machinery can explain the inhibition of normal hematopoiesis in AML has been challenged, partly due to variable Fas expression on AML cells and partly because elevated expression of Fas on AML cells predicted increased relapse-free survival (39,40). Finally, we only studied two cytokines and cannot exclude the possibility that other substances might be involved in the down-regulation of bone marrow hematopoietic activity in AML.
Only a minority of AML patients are cured. Improvements in treatment consist frequently of intensified therapeutic regimens restricted to younger patients, limiting most elderly (>70 years) AML patients to a more palliative approach (1). New treatment modalities for this group may consist of a more mechanismbased therapy including antibodies raised against hematopoietic growth factors, inhibitors of signal transduction, and antiangiogenic compounds. These approaches will require a better understanding of the contents and distribution of the molecular targets within the bone marrow microenvironment, and specific analyses of the BM-IF like those presented here are likely to yield important and relevant information about normal and malignant hematopoietic activity, thus possibly adding to the design of new drugs.