SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Objective

Adenosine deaminase (ADA; EC 3.5.4.4) activity is elevated in the synovial fluid (SF) of patients with rheumatoid arthritis (RA). Since the antiinflammatory effect of methotrexate is reportedly associated with increased levels of extracellular adenosine, the present study was undertaken to clarify the role of 2 ADA isozymes, ADA1 and ADA2, in the pathogenesis of RA.

Methods

The activities of ADA1 and ADA2 were measured in SF from RA and osteoarthritis (OA) patients, in sera from RA patients, and in lysates prepared from mononuclear and polymorphonuclear cells from SF from RA patients, peripheral blood from RA patients, and fibroblast-like synoviocytes (FLS) from RA and OA patients. Also measured were the effects of proinflammatory cytokines on ADA1 activity and ADA messenger RNA (mRNA) expression in RA FLS, as determined using real-time polymerase chain reaction. The adenosine concentration in RA SF was measured by radioimmunoassay.

Results

The adenosine concentration in RA SF ranged from 0.027 μM to 0.508 μM (mean ± SD 0.156 ± 0.132 μM). At those concentrations, ADA1 would be expected to be functionally dominant due to its higher affinity for adenosine. ADA1 activity in RA SF (mean ± SD 14.4 ± 8.5 IU/liter) was significantly higher than that in OA SF (3.0 ± 1.1 IU/liter) or RA sera (3.0 ± 0.6 IU/liter); moreover, ADA1 activity in RA FLS lysate was the highest among the cell lysates tested. Proinflammatory cytokines did not affect ADA1 activity or ADA mRNA expression in RA FLS.

Conclusion

Elevated ADA1 activity is an intrinsic characteristic of RA FLS, which likely contributes to the pathogenesis of RA by neutralizing the antirheumatic properties of endogenous adenosine.

Adenosine deaminase (ADA; EC 3.5.4.4) is a key enzyme in purine metabolism that catalyzes irreversible deamination of adenosine and 2′-deoxyadenosine to inosine and 2′-deoxyinosine, respectively (1). In humans, 3 ADA isozymes with differing molecular weights, kinetic properties, and tissue distributions have been identified (2): 1) a 35-kd enzyme (ADA1); 2) a 280-kd enzyme comprising two 35-kd ADA1 enzymes complexed with a nonenzymatic 200-kd combining protein that was recently shown to be identical to CD26 (ADA1 + combining protein) (3); and 3) a 100-kd enzyme (ADA2) (4). The first 2, which share the same catalytic subunit, do not differ significantly in their kinetic properties (4). In contrast, ADA2 has a lower affinity for adenosine and lower catalytic activity with deoxyadenosine than ADA1 (5).

Elevated serum ADA activities have been reported in patients with diseases in which cellular immunity is stimulated. For example, ADA1 activity is elevated in patients with acute lymphoblastic leukemia (6) or acute hepatitis (7), and ADA2 activity is elevated in patients with human immunodeficiency virus infection (8). In human tissues and cells the majority of ADA activity is derived from ADA1, but the prevalent form in serum is ADA2 (9, 10). Most human cells contain only small amounts of ADA2, and its tissue source is not yet completely clear, although the monocyte/macrophage cell system is likely a major source (10, 11). It is also known that total ADA activity is significantly higher in the synovial fluid (SF) of rheumatoid arthritis (RA) patients than in the SF of osteoarthritis (OA) patients (12, 13).

Methotrexate (MTX), one of the most effective antirheumatic drugs, reportedly increases extracellular adenosine concentrations at sites of inflammation and represses the infiltration of inflammatory cells (14). In addition, the nonselective adenosine receptor antagonists theophylline and caffeine are capable of reversing the antiinflammatory effects of MTX in rat adjuvant arthritis (15). Taken together, these findings suggest that adenosine is an effector molecule mediating the antirheumatic effects of MTX via adenosine receptor signaling, and that reduction of the local concentration of adenosine by ADA may contribute to the joint inflammation of RA. In the present study, therefore, we examined the activities and cellular sources of 2 ADA isozymes, ADA1 and ADA2, in RA, with the aim of gaining better understanding of the pathophysiologic role of ADA in rheumatoid inflammation.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Reagents.

All reagents were purchased from Sigma (St. Louis, MO) unless indicated otherwise.

Synovial cells, synovial fluid, and sera.

After informed consent was obtained, human fibroblast-like synoviocytes (FLS) were isolated from synovial tissues of patients with RA (RA FLS) or osteoarthritis (OA FLS) who were undergoing joint replacement surgery. RA and OA were diagnosed according to the criteria of the American College of Rheumatology (formerly, the American Rheumatism Association) (16, 17). The collected tissues were minced and incubated first for 3–4 hours at 37°C with 4 mg/ml collagenase and then for 30 minutes with 0.05% trypsin (Difco, Detroit, MI). The isolated cells were cultured in RPMI 1640 medium (Nissui, Tokyo, Japan) supplemented with 10% fetal calf serum (Life Technologies, Grand Island, NY), penicillin, streptomycin, and L-glutamine (complete medium). Adherent FLS were used from the second through fifth passages, a period during which they were highly proliferative.

RA and OA SF were transferred to sterile tubes containing 1,000 units of heparin and centrifuged at 1,500g for 15 minutes. The supernatants were then treated for 10 minutes at 37°C with 10 units/ml hyaluronidase (Nakalai Tesque, Kyoto, Japan) and centrifuged again at 1,500g for 15 minutes. The resultant cell-free supernatants were stored at −20°C until use. Sera from RA patients were obtained on the same day as SF collection and stored at −20°C until use.

Cell fractionation in synovial fluid and peripheral blood.

Human mononuclear cells (MNCs) and polymorphonuclear cells (PMNs) were isolated from SF and peripheral blood (PB) of RA patients by density-gradient centrifugation using Mono-Poly Resolving Medium according to the instructions of the manufacturer (Dainippon Pharmaceutical, Osaka, Japan), with slight modifications. Briefly, SF was transferred to sterile tubes containing 1,000 units of heparin and 40 mg of EDTA and centrifuged at 400g for 5 minutes. The pelleted cells were then washed 3 times with phosphate buffered saline and filtered through a Cell Strainer (pore size 40 μm; Becton Dickinson, Franklin Lakes, NJ) to remove any fibrous debris. MNCs and PMNs were then isolated using Mono-Poly Resolving Medium. PB was collected in sterile plastic tubes containing EDTA and then placed directly on Mono-Poly Resolving Medium, and separation was similarly performed. For hemolysis, MNCs and PMNs were resuspended in 7 ml of 0.2% (weight/volume) NaCl solution for ∼40 seconds. Osmotic pressure was then restored by addition of 7 ml of 1.6% (w/v) NaCl, after which the samples were centrifuged at 400g for 5 minutes, and MNCs and PMNs were resuspended in phosphate buffered saline. The purity of the preparations was estimated by microscopic examination of the cells stained using the May-Grünwald-Giemsa method; only preparations that were >90% pure were used for the subsequent assays.

Lysate preparation.

Cells were centrifuged at 400g for 5 minutes, resuspended in TNE buffer (10 mM Tris HCl, pH 7.8, 1% [volume/volume] Nonidet P40 [Nakalai Tesque], and 0.15M NaCl) supplemented with a protease inhibitor cocktail (Complete, Mini; Boehringer, Mannheim, Germany), and incubated at 4°C for 30 minutes with gentle mixing every 5 minutes. Samples were then stored at −80°C until use. After samples were thawed and centrifuged at 20,000g for 15 minutes, the supernatants were assayed for total ADA and ADA isozyme activities.

Assay of total ADA and ADA isozyme activity.

The activities of the total ADA and the ADA isozymes were assayed using an automatic analyzer (TBA-80FR; Toshiba, Tokyo, Japan) with a commercial kit according to the instructions of the manufacturer (Toyobo, Osaka, Japan). The enzyme activity was determined by quantifying inosine liberated from the substrate, adenosine (10 mM). Inosine was converted by purine nucleoside phosphorylase into hypoxanthine, which was further converted by xanthine oxidase into uric acid and hydrogen peroxide. The hydrogen peroxide was then converted into the quinone dyes by peroxidase, 4-aminoantipyrine, and N-ethyl-N-(2-hydroxy-3-sulfopropyl)-m-toluidine, and the absorbance at 548/700 nm was measured. ADA2 activity was measured in the presence 0.35 mM erythro-9-(2-hydroxy-3-nonyl) adenine hydrochloride (EHNA; Toyobo) since ADA1 activity is blocked by EHNA at that concentration but ADA2 activity is unaffected (18). Total ADA activity was measured in the absence of EHNA, and ADA1 activity was then calculated by subtracting the ADA2 activity from the total activity.

Determination of the adenosine concentration in SF.

The adenosine concentration in SF was determined by radioimmunoassay according to the instructions of the manufacturer (Yamasa, Chiba, Japan). Briefly, coformycin (an ADA inhibitor), dipyridamole (an adenosine uptake inhibitor), and EDTA (10 mM) were added to the SF immediately after it was withdrawn from each patient's joint. The radioimmunoassay was performed in cooperation with SRL Laboratory (Tokyo, Japan).

Determination of C-reactive protein levels.

The concentration of serum C-reactive protein (CRP) was determined using TBA-80FR, with a commercial kit (Iatron, Tokyo, Japan).

Quantitative analysis of ADA messenger RNA (mRNA) expression using reverse transcription–polymerase chain reaction (RT-PCR).

Total RNA from RA FLS and OA FLS was prepared using an RNeasy Mini Kit (Qiagen, Tokyo, Japan). After DNase I treatment, the first-strand complementary DNA (cDNA) was synthesized using the SuperScript First-Strand Synthesis System according to the instructions of the manufacturer (Invitrogen, Carlsbad, CA).

To evaluate the level of expression of ADA mRNA in the samples, quantitative real-time RT-PCR was performed using an ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). The human ADA sequence was amplified using real-time RT-PCR under standard conditions with primers corresponding to nucleotides 618–638 and 684–704, and the amount of PCR product was quantitated using a TaqMan probe corresponding to nucleotides 657—681 of the ADA open reading frame (GenBank accession no. NM_000022). Similarly, GAPDH mRNA was quantitated using primers corresponding to nucleotides 66–84 and 268–291 and a TaqMan probe corresponding to nucleotides 243–262 (GenBank accession no. M33197). The TaqMan probes consisted of oligonucleotides with a reporter dye (6-carboxyfluorescein) and a quencher dye (6-carboxytetramethylrhodamine) on their 5′ and 3′ ends, respectively. To control for differences in the quantity of cDNA in the various samples, ADA mRNA levels were normalized to those of GAPDH mRNA.

Incubation with proinflammatory cytokines.

RA FLS and OA FLS were incubated (24 hours, 37°C, 5% CO2) with 10 ng/ml interleukin-1β (IL-1β; PeproTech, London, UK), 100 ng/ml IL-6 (PeproTech), 100 ng/ml interferon-γ (IFNγ; PeproTech), or 100 ng/ml tumor necrosis factor α (TNFα; Chemicon, Temecula, CA). After incubation, total RNA and cellular lysates were prepared, and levels of ADA mRNA expression and ADA1 activity were determined as described above. The results are shown as percentages of control values determined in the absence of cytokines.

Statistical analysis.

Results are expressed as the mean ± SD. Statistical comparisons between groups were carried out using Student's unpaired t-test. We also applied Welch's unpaired t-test to confirm the evaluations of the significance of differences between 2 groups with different variances. Student's paired t-test was used to compare ADA activities in SF and sera from RA patients. Pearson's correlation coefficient was used to evaluate the correlation between ADA activity in SF and CRP concentration in serum of RA patients.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

ADA activity in SF and sera.

When ADA activity was measured in vitro, the concentration of substrate adenosine (10 mM) was sufficient to ensure that both ADA1 and ADA2 were functioning at maximal velocity (Km = 52 μM and 2,000 μM, respectively). Under these conditions, total ADA, ADA1, and ADA2 activities were all significantly higher in SF from patients with RA than in SF from patients with OA or in sera from patients with RA (Figures 1A and B). The elevated ADA activity seen in RA SF thus did not reflect a systemic change, but was a local event inside RA joints. In addition, although ADA2 activity was higher than that of ADA1 in vitro, we found that the adenosine concentration in RA SF ranged only from 0.027 to 0.508 μM (mean ± SD 0.156 ± 0.132 μM) (Figure 1C), which suggests that the in vitro findings might not be relevant to in vivo conditions, and ADA1 might actually be the dominant isoform within the RA joint.

thumbnail image

Figure 1. A, Comparison of total adenosine deaminase (ADA), ADA1, and ADA2 activities in synovial fluid (SF) from patients with rheumatoid arthritis (RA) and osteoarthritis (OA). In each case, the enzyme activity in RA SF (n = 55) was significantly higher than in OA SF (n = 20). Larger circles with bars show the mean ± SD. B, Comparison of total ADA, ADA1, and ADA2 activities in paired RA SF and RA serum samples. SF and sera were collected from the same patients (n = 49) on the same day. In each case, the enzyme activity in the SF was significantly higher than in the corresponding serum sample. C, Dependence of ADA activity on adenosine concentration. Based on the observed Km values, ADA1 would be expected to be functionally dominant within the range of adenosine (substrate) concentrations found in RA SF. Solid circles represent the percent activity of ADA1; open circles represent the percent activity of ADA2.

Download figure to PowerPoint

Activity of ADA isozymes in cell lysates.

To investigate the source of the ADA activity, cells obtained from RA joints were purified, and the enzyme activities of the lysates were assayed. MNCs and PMNs in SF and PB from RA patients and FLS from RA and OA joints were separated as described in Materials and Methods. The cells were then lysed, and the total ADA, ADA1, and ADA2 activities in the lysates were determined. We found that the ADA1 activity was significantly higher than the ADA2 activity in all of the cell populations tested (Figure 2). The highest level of ADA1 activity was found in RA FLS; this was >10-fold higher than in the other cell types from RA patients. Moreover, the ADA1 activity in RA FLS was 5-fold higher than that in OA FLS, although no significant difference was found in the levels of ADA2 activity. High levels of ADA1 activity appear to be characteristic for RA FLS, suggesting that FLS are the major source of ADA1 in RA joints.

thumbnail image

Figure 2. ADA1 and ADA2 activities in polymorphonuclear cells (PMNs), mononuclear cells (MNCs), and fibroblast-like synoviocytes (FLS). In each cell type, ADA1 activity (open bars) was significantly higher than ADA2 activity (solid bars). By far, the highest level of ADA1 activity was seen in RA FLS (>10-fold higher than in RA PMNs and RA MNCs and 5-fold higher than in OA FLS). Inset, ADA mRNA expression in RA FLS and OA FLS. Relative levels of ADA mRNA expression were determined using real-time reverse transcription–polymerase chain reaction. As an internal control, ADA mRNA levels were normalized to those of the housekeeping gene GAPDH. ADA mRNA was significantly more abundant in RA FLS than in OA FLS. Values are the mean and SD. PB = peripheral blood; NS = not significant (see Figure 1 for other definitions).

Download figure to PowerPoint

ADA mRNA expression.

The finding that the ADA activity was significantly higher in RA FLS than in OA FLS suggests that ADA gene expression was up-regulated in RA FLS. We tested this hypothesis using real-time RT-PCR and found that significantly higher levels of ADA mRNA were expressed in RA FLS than in OA FLS (Figure 2, inset). Up-regulation of ADA gene transcription thus appears to be at least partly responsible for the higher ADA activity in RA FLS.

ADA activity and CRP.

CRP is a cytokine-induced protein that is typically produced at the time of an acute-phase reaction such as infection, inflammation, or tissue injury (19), and its serum concentration is widely used as a surrogate marker for inflammatory disease activity. In RA patients, ADA1 activity in SF was positively correlated with the serum CRP concentration (data not shown). In contrast, no significant correlation was found between ADA2 activity and serum CRP concentration.

Failure of proinflammatory cytokines to induce ADA mRNA or increase ADA activity.

It is known that concentrations of proinflammatory cytokines such as IL-1β, IL-6, and TNFα are elevated in RA SF (20, 21) and that these proinflammatory cytokines induce production of CRP (22–24). We therefore tested the possibility that the observed increase in ADA1 activity in RA FLS was induced by proinflammatory cytokines. As shown in Figure 3, however, IL-1β, IL-6, IFNγ, and TNFα all failed to affect either expression of ADA mRNA or ADA1 activity.

thumbnail image

Figure 3. Failure of proinflammatory cytokines to induce ADA mRNA expression or ADA1 activity in fibroblast-like synoviocytes (FLS). Incubation of RA FLS (open bars) and OA FLS (solid bars) with the indicated cytokines for 24 hours had no effect on either ADA1 activity (A) or ADA mRNA expression (B) in either cell type. Values are the mean and SD. IL-1β = interleukin-1β; IFNγ = interferon-γ; TNFα = tumor necrosis factor α (see Figure 1 for other definitions).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Intracellular adenosine is involved in cellular energy production and purine metabolism, whereas extracellular adenosine acts as an autocrine and/or paracrine factor via specific cell surface receptors, exerting multiple physiologic actions in a variety of systems, including the immune system (25–28). Consider, for example, the case of inherited ADA deficiency. The resultant accumulation of extracellular and intracellular adenosine leads to severe combined immunodeficiency. The mechanism by which this occurs is currently thought to involve direct lymphotoxicity of intracellular adenosine as well as A2A adenosine receptor–mediated signaling (29). In that regard, investigators at our laboratory previously showed that adenosine signaling via A2A receptors represses T cell function by competing with T cell receptor–mediated signaling (30).

The antiinflammatory effects of MTX are reportedly due in large part to its capacity to enhance extracellular adenosine at sites of inflammation (14). Since several lines of evidence suggest that T cells, especially Th1-type cells, are responsible for the pathogenesis of RA (31), extracellular adenosine likely mediates the antiarthritic effects of MTX by negatively regulating T cell function. Supporting this notion is the finding that the nonselective adenosine receptor antagonists theophylline and caffeine are capable of reversing the antiinflammatory effects of MTX in rat adjuvant arthritis (15). In that context, an ADA-catalyzed reduction in the local concentration of adenosine would be expected to contribute to the joint inflammation of RA.

In addition to the cytosol, ADA is found on the surface of many cells (ecto-ADA) (32), and 2 cell surface receptors for ADA have been identified: CD26 (dipeptidyl peptidase IV) and A1 adenosine receptors (32). Ecto-ADA interacting with CD26 is reportedly a costimulatory molecule that facilitates a variety of signaling events in different cell types (32), including T cells (33). ADA may thus play a regulatory role either by catalyzing the irreversible deamination of adenosine to inosine and/or by transmitting signals when interacting with CD26 or A1 receptors.

When enzyme activities were assayed in vitro, the adenosine concentration used (10 mM) was sufficient for both ADA1 and ADA2 to function at maximum velocity. Under these conditions, we found that ADA2 activity was greater than ADA1 activity in both RA SF and OA SF (Figures 1A and B), which is consistent with an earlier report that the level of ADA2 activity in tuberculous pleural effusions is much higher than that of ADA1 activity (34). However, all of these measurements were obtained in vitro and may not be relevant to in vivo conditions. Indeed, the fact that ADA2 has a Km for adenosine of 2,000 μM, while that of ADA1 is 52 μM (4, 35), means that ADA1 should be the functionally predominant isozyme in RA SF (and OA SF), given the range of adenosine concentrations found there (Figure 1C). In contrast, considering that the half-life of adenosine in whole blood is ∼1 minute (36) due to cellular uptake via nucleoside transporters and degradation to inosine by ecto-ADA, the adenosine concentration measured in SF might underestimate the true concentration inside the joints, even with addition of both dipyridamole (a nucleoside transporter inhibitor) and coformycin (an ADA inhibitor) to the SF immediately after its withdrawal from the joint. Still, ADA1 should be the predominant ADA isozyme in RA joints at adenosine concentrations in the range of ≤1 mM.

Comparison of ADA isozyme activities in FLS, MNCs, and PMNs revealed that RA FLS possessed the highest ADA1 activity, >10-fold greater than that in the other types of cells obtained from RA patients. This could reflect in part the higher protein content of FLS since the size of the individual cells is significantly larger than MNCs or PMNs. This is not a complete explanation, however, because ADA1 activity in RA FLS was also 5-fold greater than in OA FLS, even though the respective sizes of the 2 cell types are similar. In addition, no significant difference in ADA2 activities in RA FLS and OA FLS was found.

Conversely, it was previously reported that there is a significant positive correlation between matrix metalloproteinase 9 (MMP-9) and ADA1 levels in RA SF (37), and a positive correlation between PMN count and MMP-9 (38). In that context, the fact that PMNs are usually the most abundant cell type in RA SF raises the possibility that they are the major source of ADA1, even though each cell contains little ADA1. Unfortunately, RA FLS, which constitutively express MMP-9 (39), were not investigated as a possible source of MMP-9 in those studies. In the present study, moreover, ADA1 activity was not significantly correlated with PMN count or with the total cell number in RA SF. Finally, proliferation of FLS in joint tissue, not SF, is characteristic of RA. Thus, the highly elevated ADA1 activity seen in RA SF would appear to be a characteristic feature of RA FLS, which suggests that these cells are the major source of ADA1 in RA joints.

It also seems possible that the apparently higher deaminase activity in RA FLS reflects an abundance of ecto-ADA bound to RA FLS via CD26 and/or A1 receptors. In fact, RA FLS express all 4 adenosine receptors, including A1 receptor (40), and the ecto-ADA from HEp-2 cells and from lymphocytes obtained from patients with chronic lymphocytic leukemia was identified as ADA1 (41). To further address the question of whether the increased ADA1 activity is an intrinsic characteristic of RA FLS, we compared levels of ADA gene expression in RA FLS and OA FLS. The fact that human ADA genes (GenBank accession nos. K02567, BC007678, NM_000022, X02994, or XM_029810) have essentially the same coding sequences suggests that ADA isozymes are derived from a single gene through mechanisms such as alternative splicing, posttranscriptional modification, and/or posttranslational modification (42). Whatever the case, the level of ADA gene expression appears higher in RA FLS than in OA FLS; certainly the amount of transcript was significantly higher in RA FLS (Figure 2, inset). This is in accordance with our conclusion that the increase in cellular ADA is an intrinsic characteristic of RA FLS. That the difference in the level of ADA mRNA expression in RA FLS and OA FLS is less marked than the difference in the ADA activities is consistent with the notion that ADA isozymes are derived from posttranscriptional modification of a single gene product rather than from different individual gene products (43).

Proinflammatory cytokines such as IL-1, TNFα, and IL-6 are generally believed to contribute to the pathogenesis of RA (22). This prompted us to test whether these cytokines might be responsible for the induction of ADA in RA FLS. Under the in vitro conditions in our study, however, none of the cytokines tested significantly affected ADA1 activity or ADA mRNA expression in either RA FLS or OA FLS (Figure 3).

We found a significant correlation between ADA1 activity in RA SF and CRP levels in RA sera, which raises the possibility that the elevated ADA1 activity in RA SF may be secondary to the systemic inflammation. Our other results provide evidence against this idea, however. For instance, in RA patients, serum CRP levels correlated significantly with ADA1 activity but not with ADA2 activity, even though ADA2 is the predominant ADA isozyme in serum. Moreover, the proinflammatory cytokine IL-6, which is responsible for increases in CRP production in the liver, had no effect on ADA activity or expression in RA FLS. We therefore conclude that the observed increases in ADA1 activity and ADA gene expression in RA FLS reflect an intrinsic abnormality of RA FLS and are not an inflammation-induced secondary effect.

Finally, our findings raise the possibility that novel approaches to the treatment of RA might be developed based on a strategy of increasing extracellular adenosine and normalizing the intrinsic abnormality of RA FLS. Such an approach has the potential to be highly efficacious and would seem to warrant further investigation.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors are grateful to Ms Kyoko Tanaka (Kobe University) for technical assistance and to Dr. William F. Goldman (MST Editing, Baltimore, MD) for editorial assistance and preparation of the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES