Tumor necrosis factor-alpha promotes survival in methotrexate-exposed macrophages by an NF-kappaB-dependent pathway

Introduction Methotrexate (MTX) induces macrophage apoptosis in vitro, but there is not much evidence for increased synovial macrophage apoptosis in MTX-treated patients. Macrophage apoptosis is reported, however, during clinical response to anti-tumor necrosis factor-alpha (TNF-α) treatments. This implies that TNF-α promotes macrophage survival and suggests that TNF-α may protect against MTX-induced apoptosis. We, therefore, investigated this proposal and the macrophage signaling pathways underlying it. Methods Caspase-3 activity, annexin-V binding/7-aminoactinomycin D (7-AAD) exclusion and cell-cycle analysis were used to measure steps in apoptosis of primary murine macrophages and cells of the RAW264.7 macrophage cell line that had been exposed to clinically-relevant concentrations of MTX and TNF-α. Results MTX induces apoptosis in primary murine macrophages at concentrations as low as 100 nM in vitro. TNF-α, which has a context-dependent ability to increase or to suppress apoptosis, efficiently suppresses MTX-induced macrophage apoptosis. This depends on NF-κB signaling, initiated through TNF Receptor Type 1 ligation. Macrophage colony stimulating factor, the primary macrophage survival and differentiation factor, does not activate NF-κB or protect macrophages from MTX-induced apoptosis. A weak NF-κB activator, Receptor Activator of NF-κB Ligand (RANKL) is likewise ineffective. Blocking NF-κB in TNF-α-exposed macrophages allowed pro-apoptotic actions of TNF-α to dominate, even in the absence of MTX. MTX itself does not promote apoptosis through interference with NF-κB signaling. Conclusions These findings provide another mechanism by which TNF-α sustains macrophage numbers in inflamed tissue and identify a further point of clinical complementarity between MTX and anti-TNF-α treatments for rheumatoid arthritis.


Introduction
Synovial inflammatory macrophages have a central role in maintaining disease activity in rheumatoid arthritis (RA). Macrophage numbers in tissue are regulated by recruitment, local proliferation, local cell death and emigration to draining lymph nodes [1,2]. Synovial macrophage apoptosis has also been observed in synovium in RA [3,4]. A suppressed rate of apoptosis would contribute to maintaining inflammatory macrophage numbers, and thus clinical activity, in macrophage-dependent conditions. An enhanced rate of synovial macrophage apoptosis is reported in RA patients responding to anti-TNF-α treatments [4] as a delayed, rather than an early phenomenon [5]. Macrophage apoptosis has also been reported in patients with Crohn's disease after anti-TNF-α treatment [6]. These observations suggest that TNF-α directly, or indirectly, sustains macrophage survival in these conditions. TNF-α activity supports recruitment of macrophages into RA synovium [7], but is not known to enhance the proliferation of macrophages or to prevent the emigration of macrophages through lymphatics [2]. TNF-α, however, may deliver either apoptotic or survival signals, depending on the cell context. TNF-α is a ligand for two related receptors, TNF-R1 (p55) and TNF-R2 (p75). Macrophages and their bone marrow and blood-borne progenitors express both receptors [8]. TNF-α ligation to TNF-R1 leads to assembly of a Death-Inducing Signaling Complex (DISC) and ultimately activation of downstream effector caspases-3/6/7. Where this is the dominant consequence of TNF-R1 ligation, apoptosis follows. In other circumstances, TNF-R1 initiates survival signaling [9] through activation of the NF-B pathway. TNF-α ligation to TNF-R1 can thus bring apoptosis or survival, depending on the relative activities generated in the DISC-initiated and NF-B pathways [10]. Macrophage survival is notably dependent on NF-B signaling. The transcriptionally active RelA/NF-B 1 (p65/p50) complex appears in macrophage nuclei constitutively during late differentiation and is important for continuing survival [10,11]. In vivo, there is enhanced nuclear expression of RelA/NF-B 1 in synovial macrophages in RA [12], consistent with a role for NF-B activity in maintaining macrophage survival.
The observations of enhanced apoptosis during anti-TNF-α treatment imply that TNF-α predominantly antagonises apoptosis in RA and Crohn's disease [4,6]. On the other hand, methotrexate (MTX), an antirheumatic drug that has demonstrable pro-apoptotic effects on monocyte/macrophage cells in vitro, does not appreciably alter the synovial membrane macrophage apoptosis rate in RA patients [3,13]. This led us to question whether the presence of TNF-α in synovium was providing a survival signal to macrophages in MTXtreated patients, thus antagonizing one potential therapeutic function of MTX. That would provide additional explanation for the clinical complementarity of MTX and anti-TNF-α treatments in RA [14]. In the study of Catrina et al., which demonstrated synovial macrophage apoptosis in response to anti-TNF-α therapy, the majority of patients were co-treated with MTX [4].
MTX is a folate analogue with numerous effects on inflammatory cell proliferation and survival, cytokine expression, angiogenesis, cell adhesion and reactive oxygen production [15]. The known direct targets of MTX are dihydrofolate reductase, thymidylate synthase and 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) transformylase [15]. Consequences of inhibiting these enzymes include reduced availability of purines and pyrimidines for DNA and RNA synthesis and accumulation of AICAR. AICAR, by inhibiting enzymatic deamination of adenosine and adenosine monophosphate, is proposed to increase availability of adenosine extracellularly, for anti-inflammatory effect through adenosine cell surface receptors [16]. These effects are generally believed to be mediated by long-lived polyglutamate forms of MTX, which are generated in many cell types, including normal myeloid precursor cells and myelocytic cancer cells [17].
Cell culture RAW 264.7 cells were obtained from the American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium (Life Technologies, Carlsbad, California, USA) with 10% low endotoxin fetal calf serum (CSL, Melbourne, Australia), penicillin and gentamicin (DMEM-FCS). Bone marrow-derived monocytes/ macrophages (BMDM) were collected from the femora and tibiae of eight-week old female C57BL or BALB/c mice, as previously described [19], with the approval of the Animal Ethics Committee of The University of Western Australia. Cells were expanded in M-CSF (30 ng/ mL)-supplemented DMEM-FCS for seven days at 37°C/ 5% CO 2 /95% air. The experiments reported here were conducted with BMDM from C57BL mice, but results were comparable using BALB/c mice.
Caspase-3 protease activity RAW 264.7 cells and BMDM were cultured in six-well plates at a density of 0.3 × 10 6 cells/well for 48 hr. After treatments, trypsin-detached RAW 264.7 cells and scraped BMDM were lysed and assayed for caspase-3 activity as described previously [20]. Cell cycle analysis RAW 264.7 cells and BMDM (0.25 × 10 6 /well of a 12-well plate) were collected and washed in cold Dulbecco's Phosphate Buffered Saline (DPBS) before being fixed with 70% ethanol at -20°C. After 24 hr, cells were centrifuged at 500 × g for 10 minutes at 4°C and resuspended in 500 μL of staining buffer (50 μg/mL propidium iodide and 25 μg/mL RNase in 0.1% Triton X-100). After incubating at 37°C for 30 minutes, cells were analysed on a FACSCalibur flow cytometer (Becton Dickinson) using a doublet discrimination protocol.

Electrophoretic Mobility Shift Assay (EMSA)
EMSA was performed on nuclear extracts from RAW 264.7 cells (8 × 10 6 per treatment) that were treated with 10 ng/ mL TNF-α for up to one hour, according to previously described protocols [21]. A double-stranded NF-B consensus oligonucleotide probe 5'-GGGCATGGGAA TTTCCAACTC-3' (0.25 pmol) with 5'-G overhangs was filled in with labeled (α-32 P)dCTP (Amersham Pharmacia Biotech, Buckinghamshire, England) using the Klenow fragment of E. coli DNA polymerase I (Promega, Madison, Wisconsin, USA). This DNA probe was incubated with 3 μg of nuclear proteins for 10 minutes at room temperature. Where indicated, antibodies (1 μg) to specific NF-B factors or unlabelled oligonucleotide probes at 100-fold molar excess were also included for supershift EMSA and competition experiments, respectively. Samples were loaded onto a 4% polyacrylamide gel, containing 0.25X Tris-Borate-EDTA buffer, which had been pre-run for two hours in the same buffer. After separation, gels were exposed to Cronex X-ray film, using a single intensifying screen.

Transient transfection of RAW 264.7 cells and BMDM with an NF-B reporter plasmid
RAW 264.7 cells (1 × 10 7 ) were transiently transfected with 0.5 μg of pNFB-TA-Luc (Mercury™ Pathway profiling system, BD Biosciences) by the DEAE-dextran procedure, as previously described [21]. Transfected cells were resuspended in DMEM-FCS and distributed into a 24-well culture plate at a density of 3.3 × 10 5 cells/well. Cells were rested for 48 hr before experimentation. Firefly luciferase expression in transfected RAW 264.7 cells was measured using the Promega Luciferase Assay System, according to the manufacturer's instructions. Promoter activity was quantified by measuring light production in a multifunctional microplate reader (POLARstar OPTIMA, BMG, Germany), and presented as relative light units (RLU). BMDM were transfected with Lipofectamine™ and Plus reagents (Invitrogen) according to manufacturer's instructions. Briefly, for each 24-well, 1 μg pNF-B-TA-Luc was combined with 0.5 μL Plus reagent and 2 μL Lipofectamine™ in antibiotic-free medium. A total of 1.5 × 10 5 cells/well was added to the transfection mixture and plated in M-CSF-containing DMEM-FCS. After four to six hours of incubation at 37°C, wells were replaced with fresh growth medium. Luciferase expression was measured at least 48 hr after transfection.

Statistical analysis
Statistical significance (P < 0.05) between groups of experimental data was assessed using paired Student's t-tests.

TNF-a protects primary macrophages and a macrophage cell line from MTX-induced apoptosis
M-CSF is the primary growth, differentiation and survival factor for macrophages under physiological conditions. Serum concentrations of M-CSF are elevated to approximately 0.6 ng/ml in patients with RA [22], but concentrations in synovium are unknown. In vitro, a concentration of 30 ng/ml was optimal for sustaining the growth and survival of bone marrow-derived macrophages (BMDM) of C57BL/6 mice (results not shown).
In the presence of 30 ng/mL M-CSF, exposure to 10 μM MTX for 24 hr caused an approximately 25% loss of cell numbers in BMDM cultures ( Figure 1A). This was related to increased apoptosis, as estimated by caspase-3 activity ( Figure 1B), annexin-V binding (early apoptosis; Figure 1B) and increased sub-G 0 (apoptotic) fraction on flow cytometric cell cycle analysis ( Table 1). The proportion of cells in S-phase also increased with MTX, consistent with its known action on completing DNA synthesis [23] (Table 1).
Caspase-3 activation was evident at concentrations as low as 0.1 μM MTX in BMDM, with activity increasing dose-dependently up to 10 μM ( Figure 1C). MTX, therefore, induces macrophage apoptosis at levels usually attained in treated human RA patients [24]. This was replicated in the murine RAW 264.7 macrophage cell line ( Figure 1E). RAW 264.7 macrophages do not require exogenous M-CSF to proliferate, having gained growth signaling through introduction of the Abelson leukaemia virus [25]. They have been previously found to model primary macrophage apoptosis response [26]. A concentration of 10 μM MTX was used in subsequent experiments.
TNF-α could also maintain the viability of macrophages exposed to 10 μM MTX. TNF-α, when introduced three hours before MTX, completely prevented the loss of cell numbers in primary BMDM cultures exposed to MTX ( Figure 1A). TNF-α also almost completely inhibited MTX-induced caspase-3 activation ( Figure 1B) and annexin-V binding ( Figure 1B) and significantly suppressed the sub-G 0 (apoptotic) fraction on flow cytometric cell cycle analysis of MTX-exposed BMDM (Table 1). TNF-α did not reverse the MTXinduced increase in cells in S-phase. TNF-α, therefore, protected from MTX-induced cell loss by countering apoptosis, not through any action on cell proliferation. TNF-α also reduced the sub-G 0 population of M-CSFdeprived BMDM, without restoring S phase progression (Table 1). It is notable that TNF-α was protective even in the presence of M-CSF concentrations that are optimal for apoptosis prevention, suggesting distinct pathways for apoptosis protection. Further increases in M-CSF concentration in culture, up to 120 ng/mL, failed to substitute for the anti-apoptotic action of TNFα (results not shown). This suggested an action that was independent of M-CSF intracellular signaling and removed the possibility that TNF-α was acting through autocrine induction of M-CSF secretion. At least three hours of pre-exposure to TNF-α was required for optimal protection from MTX-induced apoptosis (results not shown).
TNF-α provided comparable protection from MTXinduced apoptosis in the RAW 264.7 macrophage cell line. TNF-α exposure significantly (P < 0.05) suppressed spontaneous apoptosis of RAW 264.7 cells as measured by caspase-3 activity and annexin-V binding ( Figure 1D). RAW 264.7 cells shared the susceptibility of primary macrophages to MTX, demonstrating enhanced caspase-3 activity and annexin-V binding after six hours of exposure ( Figure 1D). This increase was markedly suppressed when TNF-α was added three hours before MTX ( Figure 1D). Thus, the anti-apoptotic effect of TNF-α is replicated in RAW 264.7 cells, allowing use of this cell line for studying the phenomenon.

NF-B is required for the survival effect of TNF-a in macrophages
TNF-R1 ligation activates several pro-survival pathways via its association with TRAF2 (TNFR-Associated Factor 2), RIP (Receptor-Interacting Protein) c-Src and Jak2. These include the NF-B pathway, Phosphatidylinositol-3-Kinase/Protein Kinase B (PI3K/AKT) and Mitogen-Activated Protein Kinase pathways [28,29]. Constitutive NF-B activity is critical to macrophage survival [26]. Therefore, we next examined the kinetics of NF-B activation in TNF-α-stimulated macrophages and investigated whether NF-B activity was required for the anti-apoptotic effect of TNF-α.
Proteosomal degradation of the NF-B inhibitory protein, IBα, is an early indicator of activation in the canonical NF-B pathway [30]. TNF-α (10 ng/mL) addition in RAW 264.7 cells induced the degradation of IBα within 15 minutes ( Figure 3A, P < 0.05). This corresponded to the appearance of DNA-binding NF-B in the nucleus, with activity peaking at 30 minutes before returning to baseline levels by one hour ( Figure 3B).
Supershift analyses with specific antibodies to RelA, NF-B 1 , c-Rel and Rel-B identified RelA (p65) and NF-B 1 (p50) as the main components of the NF-B complex induced by TNF-α ( Figure 3B). A light c-Rel supershifted band was also detected ( Figure 3B). IBα This could be entirely prevented by introducing TNF-a three hours before MTX. Cell viability was assessed by the MTT reduction assay. Shown are mean ± SEM of normalised data from three independent experiments. B and D, TNF-a suppresses caspase-3 activation (left axis) and annexin-V binding (right axis) in MTXexposed primary BMDM (B) and RAW 264.7 cells (D). MTX was added three hours after TNF-a treatment, and caspase-3 activity or annexin-V binding was measured after 24 hr for BMDM and 6 hr for RAW 264.7 cells. Data are mean ± SEM of at least four independent experiments in each case. C and E, MTX dose-dependently increases apoptosis in BMDM (C) and RAW 264.7 cells (E). * = P < 0.05; ** = P < 0.01. levels recovered by one hour ( Figure 3A). The IKBA gene is also NF-B-responsive, so early recovery of IBα levels indicates functional NF-B signaling [30]. Lastly, TNF-α stimulation of RAW 264.7 cells transiently expressing a NF-B-responsive reporter construct showed that NF-B transcriptional activity was enhanced in a time-dependent manner ( Figure 3C). These results confirm that TNF-α is a rapid activator of NF-B function in macrophages.
To investigate whether NF-B activity is essential for TNF-α activity against MTX-induced apoptosis, the sesquiterpene lactone parthenolide (PAR) and BAY11-7085 (BAY) were used to specifically prevent NF-B signaling [19,31]. PAR and BAY pre-treatment for 30 minutes abolished both constitutive and TNF-α-stimulated NF-B transcriptional activity ( Figure 4A). They also completely prevented TNF-α from rescuing RAW 264.7 cells (Figure 4B,C) and primary macrophages ( Figure  4D-F) from MTX-induced apoptosis and caspase-3 activation. These findings indicate that NF-B activation is required for TNF-α protection against apoptosis induced by MTX. Notably, TNF-α became proapoptotic, rather than anti-apoptotic when NF-B signaling was blocked, when estimated by either annexin-V binding (P < 0.05) or caspase-3 activity (P < 0.05).
The involvement of PI3K/AKT, ERK, JNK and p38 MAP kinases was also explored using their specific inhibitors, but they were found to be dispensable for TNF-α-induced survival of macrophages, as assessed by annexin-V staining and caspase-3 activity (results not shown).

Autocrine TNF-a signaling is not required for basal survival of M-CSF-maintained macrophages
Macrophages are also sources of TNF-α, raising the possibility that constitutive NF-B activity (and thus survival) depended on autocrine TNF-α stimulation. We, therefore, treated BMDM cultures with etanercept, a chimeric protein comprising Fc domains of human IgG 1 and TNF-α-binding domains of the Type 2 TNF receptor, which neutralises both murine and human TNF-α [32]. An irrelevant human myeloma-derived IgG 1 served as a control. Etanercept, however, did not increase caspase-3 activation, indicating that autocrine TNF-α stimulation was not important for survival of macrophages cultured with optimal concentrations of M-CSF (results not shown).

RANKL, a weak activator of NF-B in macrophages, does not protect from MTX-induced apoptosis
The observation that TNF-α protected macrophages from MTX-induced apoptosis led us to question whether other cytokines present in rheumatoid synovium or erosions may also protect through NF-B induction. Receptor Activator of NF-B Ligand (RANKL) circulates at elevated concentration in RA and is demonstrable in rheumatoid synovium and erosions [33]. It is required for osteoclast formation from monocyte/macrophage precursors. Unlike TNF-α, however, RANKL did not suppress caspase-3 activation in MTX-exposed primary macrophage cultures at concentrations up to 400 ng/mL ( Figure 5A). A dose of 200 ng/mL of RANKL is sufficient to elicit classical  Figure 2 Human (h) and murine (m) TNF-a confer comparable protection against apoptosis in RAW 264.7 cells exposed to 10 mM MTX. Caspase-3 activity was measured after six hours with MTX, without TNF-a, or pre-treated with 10 ng/mL mTNF-a or hTNF-a for three hours. Shown are mean ± SEM of three independent experiments. * = P < 0.05. osteoclastogenesis in our hands [34], so was used for other experiments. Comparing effects on NF-B activation, we found that RANKL (six hours) stimulated an approximately three-fold increase in NF-B-luciferase reporter expression in transfected BMDM, well short of the approximately 22-fold increase with TNF-α ( Figure 5B). Western blot analyses also indicated weaker phosphorylation and degradation of IBα after RANKL treatment, compared to TNF-α ( Figure 5C). Therefore, RANKL does not protect differentiated primary macrophages from MTX, possibly due to insufficient activation of NF-B. Figure 4 TNF-a must activate NF-kB signaling to protect macrophages from MTX-induced apoptosis. A. PAR and BAY (10 μM) prevent TNF-a-induced NF-kB activity in RAW 264.7 cells, as measured by activity of the pNF-kB-TA-Luc reporter. Inhibitors were introduced 30 minutes before TNF-a and luciferase activity was measured nine hours later. Results are mean ± SEM of at least four independent experiments. B-F. PAR and BAY (10 mM or 7.5 mM) prevent TNF-a from rescuing MTX-treated RAW 264.7 cells (B and C) and BMDM (D-F) from apoptosis. Cells were exposed to PAR (B, D, and E) or BAY (C and F) for 30 minutes before TNF-a stimulation for three hours, after which MTX was introduced. Annexin-V binding and/or caspase-3 activity was measured 6 hours later in RAW 264.7 cells and 24 hours later in BMDM. Results are mean ± SEM of at least three independent experiments in each case. Normalised data are shown for BAY experiments. * = P < 0.05; ** = P < 0.01.

MTX itself does not induce apoptosis through NF-B suppression
The observation that NF-B induction countered the apoptotic action of MTX led us to test whether MTX itself acted through inhibiting NF-B. This has been previously observed in Jurkat T-cells and in the poorly differentiated myelo-monocytic U937 cell line [35]. However, MTX, at concentrations sufficient to induce apoptosis, failed to suppress either basal or TNF-αinduced NF-B activity in RAW 264.7 cells transiently transfected with a specific NF-B reporter construct (pNF-B-TA-Luc) ( Figure 6A). In BMDM, it also failed to prevent the TNF-α-stimulated phosphorylation of IBα, an essential step in IBα degradation and RelA/ NF-B 1 complex release ( Figure 6B). MTX, therefore, does not exert its apoptotic actions through NF-B suppression.
Anti-inflammatory actions of MTX on macrophages have been attributed to enhanced extracellular adenosine generation. However, neither adenosine itself (Figure 6C), or the pan-adenosine receptor agonist NECA ( Figure 6D), affected caspase-3 activity. Thus, adenosine accumulation alone is not a sufficient explanation for MTX-induced macrophage apoptosis.

Discussion
TNF-α protects macrophages from apoptosis induced by MTX. This offers an explanation for the clinical observation that anti-TNF-α treatments cause apoptosis of monocyte/macrophage lineage cells in peripheral blood and synovium of patients with RA [4], in a study where most participants were also taking MTX. The survival signal from TNF-α is transduced through TNF-R1 and the canonical NF-B pathway. M-CSF, which directs Figure 5 RANKL, a weak NF-B activator, does not protect primary BMDM from MTX-induced apoptosis. A. RANKL (100 to 400 ng/mL) was introduced to BMDM 6 hours before MTX (10 mM) treatment. Caspase-3 activity was measured 24 hours later. Results are mean ± SEM from at least six independent experiments. B. and C. Compared to RANKL (200 ng/mL), TNF-a (10 ng/mL) treatment causes greater activation of NF-kB in BMDM. Cells transiently expressing the pNF-kB-TA-Luc reporter were stimulated with RANKL or TNF-a for six hours (B). Shown are mean ± SEM of normalised data from three independent experiments. C. Cells were M-CSF-starved overnight before stimulation with RANKL or TNF-a for 10 minutes and 30 minutes. Whole-cell lysates were subjected to western blot analysis for phosphorylated IkBa, total IkBa and b-actin. Results are representative of three independent experiments. * = P < 0.05; ** = P < 0.01. macrophage differentiation, growth and survival, but does not activate NF-B signaling, cannot substitute for TNF-α. This is notable, because M-CSF is present in peripheral blood and synovial fluid in RA [22], and might have been expected to render anti-apoptotic effects of TNF-α redundant. M-CSF signals for growth and survival primarily through the PI3K/AKT pathway.
TNF-α can activate both pro-apoptotic and anti-apoptotic pathways, with outcomes that differ between cell types and conditions. The importance of NF-B signaling in survival is highlighted by the outcomes of PAR and BAY exposure, which converted TNF-α from a dominantly pro-survival signal to an apoptotic signal. NF-B-regulated proteins that have been linked to inhibition of apoptotic signaling include TRAF-1, TRAF-2, cIAP-1, cIAP-2, XIAP, FLIP, A20, GADD45β, the antioxidant Mn-SOD, and anti-apoptotic members of the Bcl-2 family, such as A1 and Bcl-xL [36].
The canonical NF-B pathway is activated in macrophages of rheumatoid synovium [12], consistent with a role in maintaining cell survival in untreated patients [10,11]. Other products of the inflamed synovium, including IL-1, RANKL, GM-CSF, VEGF, PDGF and leukotrienes [37] can promote NF-B signaling in responsive cells and may contribute, with TNF-α, to NF-B activation in macrophages of rheumatoid synovium. However, the experiments with RANKL indicated that not all NF-B activators protect from apoptosis. RANKL failed to protect primary murine macrophages from MTX-induced apoptosis, even at concentrations that promote osteoclastogenesis in our hands [34] and Figure 6 MTX does not inhibit TNF-a-induced NF-kB activity. A. Reporter-expressing RAW 264.7 cells were exposed to MTX for one hour or two hour before TNF-a stimulation for a further three hour. Shown are mean ± SEM of four independent experiments. B. BMDM were pretreated with MTX for two hour before TNF-a stimulation for 10 minutes. Whole-cell lysates were probed with antibodies for phosphorylated IkBa, total IkBa and b-actin. C. Adenosine and D. NECA do not increase basal macrophage apoptosis. BMDM were exposed to increasing concentrations of adenosine or NECA for 24 hour before caspase-3 measurements. Shown are mean ± SEM of results from three independent experiments in each case. ** = P < 0.01.
which exceed the concentrations reported in rheumatoid synovial fluid [38]. The difference may arise because RANKL activates NF-B less intensely, or because of differences in other pathways recruited by TNF-R1 and RANK. Finding that RANKL is not a survival factor for macrophages mirrors findings with osteoclast precursors, where RANKL promotes activation and differentiation, but not survival [39]. The mature osteoclast, however, is protected from apoptosis by RANKL with mixed reports of NF-B involvement [40]. The finding that RANKLinduced NF-B activation is insufficient to protect primary macrophages from apoptosis is also in accord with the conclusion from clinical studies, that RANKL is not an important driver of synovial inflammation in RA, even though it is critical for erosion formation [41].
MTX has been reported to suppress NF-B activity in a different cell context [35]. However, NF-B suppression could not be demonstrated in MTX-exposed macrophages, so cannot be the mechanism of MTXinduced macrophage apoptosis. MTX anti-inflammatory response has also been linked with the extracellular accumulation of adenosine. Adenosine is reported to promote apoptosis of colon cancer cells in vitro [42], but we could find no evidence that adenosine alone causes apoptosis in macrophages. The experiments could not disprove a permissive role for adenosine. The experiments do, however, prove that MTX and TNF-α do not just act conversely on NF-B signaling. MTX has many potentially apoptotic actions [3,13,15] and NF-B has many anti-apoptotic actions [36]. It, therefore, seems more likely that the fate of TNF-α and MTXexposed macrophages is decided by the late effectors of apoptosis signaling, particularly at mitochondrial steps, that integrate and balance multiple converging inputs [43]. This requires better definition.
Macrophage NF-B signaling is an attractive therapeutic target in RA because of roles in inflammation, differentiation and survival. However, NF-B pathways are employed in diverse tissues, with broad biological consequences. Anti-TNF-α treatments suppress NF-B activity in circumstances where it is promoted by TNF-α, such as in the synovial macrophage. Currently available NF-B-suppressing small molecules, however, lack such cellular specificity. Glucocorticoids suppress macrophage NF-B activity, but with poor specificity [21]. The proteosomal inhibitor, bortezomib, suppresses NF-B signaling by preventing IBα degradation and is reported to reduce synovial macrophage inflammatory activity in experimental murine collagen arthritis [44] but, again, is not specifically targeted to the macrophage. Thalidomide likewise suppresses clinical activity of systemic onset juvenile RA [45], and leads to monocyte apoptosis in vitro through actions on AKT-1 [46] and NF-B signaling [47], but with the attendant threat of teratogenesis and other serious off-target effects. Developing macrophage-specific NF-B suppressants may also have to contend with internal redundancy, with RelA and cRel able to serve similar functions for apoptosis suppression in haematopoietic cells [48].

Conclusions
We were able to confirm that the potent NF-B activator, TNF-α, counters MTX-induced apoptosis in primary murine macrophages. A less potent activator, RANKL, could not. Neither could the primary macrophage differentiation and survival factor, M-CSF, which we found did not activate NF-B signaling. However, MTX did not suppress either constitutive or induced NF-B signaling, ruling out a direct apoptotic action through NF-B inhibition. These findings provide an additional explanation for the clinical complementarity between MTX and anti-TNF-α treatments for rheumatoid arthritis.