Serum hepatocyte growth factor is associated with viral hepatitis ...

Nephrol Dial Transplant (2002) 17: 637–644

Original Article

Serum hepatocyte growth factor is associated with viral hepatitis, cardiovascular disease, erythropoietin treatment, and type of heparin in haemodialysis patients Jacek Borawski and Michak Mys´liwiec Department of Nephrology and Internal Medicine, Medical Academy, Biakystok, Poland

Abstract Background. Increased serum hepatocyte growth factor (HGF) level is a part of the counter-system against tissue damage and predicts mortality in maintenance haemodialysis (HD) patients. We studied which of the common co-morbid and clinical conditions, and surrogates of metabolic disorders or specific organ damage determine HGF levels in these subjects. Methods. In 86 patients, pre-dialysis serum HGF, soluble endothelial markers—such as thrombomodulin (TM), von Willebrand factor and plasminogen activator inhibitor-1—and hepatitis B and C markers were measured by ELISAs. Inflammatory reactants such as C-reactive protein (CRP), a1-antitrypsin, a1 acid-glycoprotein, and immunoglobulin M and G were assayed by nephelometry, and lipoprotein(a) was determined by ELISA. Cardiovascular disease (CVD) was identified on a clinical basis. Results. Serum HGF was directly associated with the presence of viral hepatitis, alanine aminotransferase and TM levels, time on HD, the presence of CVD, CRP and a1-antitrypsin levels, use of unfractionated heparin (UFH) (vs enoxaparin) during HD, dose of UFH, use of recombinant erythropoietin (rHuEpo) treatment, and KtuV. In 36 patients not treated with rHuEpo, HGF directly correlated with haemoglobin, but not with endogenous Epo levels. There was no association between HGF and the other endothelial and inflammatory markers, gender, age, smoking, cause of renal failure, body mass index, normalized protein catabolic rate, dialysate buffers, dialysers, blood pressure, antihypertensive treatment, leukocyte and platelet counts, albumin, fibrinogen, lipoprotein(a), markers of iron and calcium–phosphorus metabolism, or metabolic acidosis. Positive viral hepatitis markers, prevalent CVD and rHuEpo treatment (in descending

Correspondence and offprint requests to: Dr Jacek Borawski, Department of Nephrology and Internal Medicine, Medical Academy, 14 Zurawia Street, PL-15-540 Biakystok, Poland. Email: [email protected] #

order of significance) were independent predictors of high HGF level. In another 20 HD patients, a 4-week course of rHuEpo treatment resulted in a significant 17% increase in circulating HGF levels. Conclusion. Serum HGF levels in HD patients are determined by inflammatory conditions such as viral hepatitis and CVD, increase in response to rHuEpo treatment, and may be influenced by type and dose of heparin used during HD procedures. Keywords: cardiovascular disease; erythropoietin; haemodialysis; heparin; hepatocyte growth factor; viral hepatitis

Introduction Hepatocyte growth factor (HGF) is a pleiotropic cytokine involved in embryonic development, and the repair, regeneration, and protection of various organs from injuries w1,2x. It exhibits mitogenic and antiapoptotic activities, and enhances motility of different cell types, including hepatocytes, renal epithelial and proximal tubular cells, and vascular endothelial cells (ECs). Following tissue damage, HGF is expressed in mesenchymal cells (e.g. fibroblasts, mononuclear cells, megakaryocytes), while its high-affinity receptor c-Met is expressed by almost all epithelial cells, ECs, and erythroid progenitors w1,2x. HGF administration and gene therapy are of value in various models of liver disease, renal failure, diabetes mellitus, and cardiovascular disease (CVD) w2x. However, their use in humans is restricted due to the possible role of HGF in propagation of some tumours w2x. Regarding cardiovascular aspects, increased blood levels of HGF have been recently proposed to be a part of the counter-system against endothelial damage in patients with arterial hypertension w3,4x, and in those with systemic arteriosclerosis w5x. In vitro studies have shown that HGF stimulates

2002 European Renal Association–European Dialysis and Transplant Association

J. Borawski and M. Mys´liwiec

638

proliferation of ECs w6,7x and retards their apoptosis w7x. Another clinically interesting function of HGF is stimulation of erythropoiesis due to enhanced proliferation and differentiation of bone marrow progenitor cells w8,9x. HGF may be of particular interest in maintenance haemodialysis (HD) patients, in whom viral hepatitis, chronic inflammation, CVD, hypertension, EC dysfunction, and anaemia are particularly common w10–12x. Serum levels of HGF are increased in these patients w13–17x, partially due to its enhanced production by leukocytes, and possibly tissue fibroblasts w15x. Pro-inflammatory cytokines are supposed to mediate HGF synthesis during the extracorporeal circulation w15x, and the role of heparin in this phenomenon cannot be ruled out w13,15x. Interestingly, the repeated and striking increase in HGF levels due to HD procedures has been proposed to protect against liver damage caused by hepatitis C virus w16x. Recently, a comprehensive Italian study w17x has revealed some important clinical meanings of the increased serum HGF levels in dialysis patients. Namely, the cytokine was found to be a strong independent predictor of mortality, being directly associated with a state of chronic inflammation and severity of carotid arteriosclerosis. On the other hand, a positive association between HGF and haemoglobin (Hb) levels found in this study w17x seems to be in favour of the anti-anaemic role of HGF in HD patients. We aimed to determine which of the common co-morbid conditions and surrogates of metabolic disorders or specific organ damage may be predictive of increased serum HGF levels when related to clinical and HD treatment-specific variables in maintenance HD patients. In addition, we investigated the mechanisms underlying the erythropoietic effects of HGF in these patients.

Patients and methods Patients Eighty-six non-diabetic, clinically stable patients (32 (37.2%) were females) of a median age of 57 (range 16–77) years, maintained on HD therapy for 14 (1–110) months were enrolled in a cross-sectional study. Their renal diagnoses were chronic glomerulonephritis (ns40, 46.5%), interstitial nephritis (ns22, 25.6%), polycystic kidney disease (ns12, 14.0%), hypertensive nephropathy (ns3, 3.5%), secondary amyloidosis (ns2, 2.3%), and unknown (ns7, 8.1%). Dialyses were performed for 3.5–5 h three times per week using the double-needle technique, native arteriovenous fistulas, low-flux cuprophane (ns38, 44.2%) or polysulfone dialysers (ns48, 55.8%), acetate (ns24, 27.9%) or bicarbonate (ns62, 72.1%) buffers, and unfractionated heparin (UFH) (Heparin, Biochemie, Kundl, Austria; ns47, 54.7%; mean total dose of 4298"1443 IUuHD session) or low molecular weight heparin (LMWH) enoxaparin (Clexane, Bellon Rho´ne-Poulenc Rorer, Montrouge, France; ns39, 45.3%; median single dose of 20 (40–60) mguHD session). UFH was administered as a loading dose

(one-third of the total amount; minimal bolus of 500 IU) followed by its continuous infusion that was discontinued 30 min prior to the end of HD. The average dialysis dose (KtuV) was 1.13"0.20, and the average normalized protein catabolic rate (nPCR) was 0.95"0.21 gukguday as calculated from pre-dialysis and immediate post-dialysis blood urea nitrogen levels using single pool kinetics. The patients’ body mass index was 23.0 (17.6–36.8) kgum2. Thirty-nine (45.3%) patients were seropositive for either hepatitis B virus surface antigen (ns5, 5.8%), hepatitis C virus antibodies (ns30, 34.9%), or for both the markers (ns4, 4.7%). Thirty-two (37.2%) patients had CVD defined on a clinical basis as described previously w11,12x. All the patients with CVD had ischaemic heart disease, eight (9.3%) had peripheral vascular disease, and one (1.2%) had a history of ischaemic stroke. Nineteen (22.1%) patients were current smokers. Systolic, diastolic arterial blood pressure (BP), and pulse pressure calculated as a mean of 10 consecutive predialysis readings preceding the study onset were 147"20.4, 87.5 (62.5–125), and 60.5"14.3 mmHg, respectively. Fifty (58.1%) patients were treated with recombinant erythropoietin (rHuEpo) (Eprex, Cilag AG Int., Zug, Switzerland) at a median subcutaneous dose of 109 (28.4 –190) IUukguweek for a median period of time of 34 (4 –96) weeks. Seventy (81.4%) patients were treated with antihypertensive drugs such as ACE inhibitors (ns34, 39.5%), calcium channel (ns55, 64.0%) and b-receptor (ns13, 15.1%) blockers, alone or in combination, as well as with mononitrites (ns24, 27.9%). None of the subjects received interferon or lamivudine therapy, steroids, oral anticoagulants, acetylsalicylic acid, or lipid-lowering drugs. The patients were treated with calcium carbonate (ns53, 61.6%), oral alphacalcidol (ns69, 80.2%), i.v. iron saccharate (ns72, 83.7%), and folic acid (ns86, 100%). During the month preceding the study onset, no patient suffered from any inflammatory or infectious disease. Approval by our institutional ethical board was obtained and all patients gave informed consent.

Methods The patients were investigated in the morning of midweek HD days under fasting conditions. Blood was withdrawn from the arterial outlet of the fistula before heparinization and immediately transferred to an accredited laboratory. Some of the serum and plasma samples were also aliquoted and stored at 408C until further assay. Complete blood counts, blood chemistries, serum calcium, and whole blood total carbon dioxide levels were determined using routine automated methods. Serum intact parathormone was assayed by immunoradiometry. Serum albumin was measured by the bromocresol green method. Plasma fibrinogen was determined with the clotting method of Clauss. Serum levels of C-reactive protein (CRP), a1 acidglycoprotein (also known as orosomucoid), a1-antitrypsin (also known as a1-proteinase inhibitor), and immunoglobulin (Ig) M and G were analysed by nephelometry (Orion Diagnostica, Espoo, Finland). The CRP assay was performed using the latex-enhanced method. Serum hepatitis C virus antibodies and hepatitis B virus surface antigen were detected with third generation enzyme-linked immunosorbent assays (ELISA) using an AxSYM analyser and kits purchased from Abbott Laboratories (Abbott Park, IL, USA). Serum HGF levels were determined using the ELISA kit manufactured by R & D Systems, Inc. (Minneapolis, MN,

HGF in HD patients

USA). Commercially available ELISAs were also used to measure plasma levels of von Willebrand factor antigen (Boehringer, Mannheim, Germany), plasminogen activator inhibitor-1 antigen (Biopool, Umea, Sweden) and lipoprotein(a) (Biopool), and serum levels of soluble thrombomodulin (sTM) (American Diagnostica, Greenwich, CT, USA), ferritin (Dialab, Wien, Austria), and endogenous erythropoietin (Epo) (Boehringer). The assays were performed in duplicates on a 400 SFC photometer (SLT-Labinstruments, Gro¨diguSalzburg, Austria), and were calibrated using provided reference samples and standards. For calculations of the results, a computer and a curve-fitting program were used. Their calculated intra- and inter-assay coefficients of variations were -10%.

Statistical analyses The normally distributed data provided by Shapiro–Wilk’s W test were expressed as mean"1 SD. The non-Gaussian data were presented as median (range), and were transformed to natural logarithm in order to normalize their distribution prior to statistical analysis. The problem of zero values in the CRP, endogenous Epo and lipoprotein(a) assays, which are lost in the log transformation, was addressed as follows: all CRP values -6 mgul were treated as 5 mgul, a value equal to one-third of the detection limit of 8.5 IUul was added to the results of the Epo assay, and a small non-zero value of 0.1 was added to the results of the lipoprotein(a) assay for all patients. As such modified data remained skewed, other mathematical transformations were also attempted. The square root transformation normalized the distribution of lipoprotein(a), but none of the methods normalized the CRP and endogenous Epo data. Bivariate associations between variables of interest were determined by Pearson’s linear or quasi-Newton’s logistic regression analysis. The CRP and endogenous Epo data were analysed using non-parametric Spearman’s correlation test. For inter-group comparisons, Student’s t-test for independent samples, x2 test, and one-way analysis of variance with post hoc Scheffe’s procedure were used when appropriate. Stepwise multiple linear regression analysis with both backward elimination and forward selection was employed to evaluate any associations between HGF level as the outcome variable and multiple independent variables. All P were two-tailed, and values -0.05 were considered statistically significant. Computations were performed using Statistica 5.1 (StatSoft Inc., Tulsa, OK, USA).

Results Bivariate correlates of serum HGF Table 1 shows detailed laboratory data. Serum HGF levels were directly associated with HD duration (rs0.489, P-0.0001), use of UFH (vs enoxaparin) ( x2s20.2, P-0.0001) (Figure 1A), positive hepatitis markers ( x2s14.2, Ps0.0002) (Figure 1B), alanine aminotransferase (ALT) (rs0.354, Ps0.0008), sTM (rs0.299, Ps0.005), use of rHuEpo ( x2s7.15, Ps0.007) (Figure 1C), alkaline phosphatase (ALP) (rs0.254, Ps0.018), a1-antitrypsin (rs0.267, Ps0.013), KtuV (rs0.247, Ps0.022), and prevalence

639 Table 1. Laboratory data HGF (nguml) White blood cells (3 103uml) Hb (gudl) Platelets (3 103uml) Iron (mgudl) Total iron-binding capacity (mgudl) Transferrin saturation (%) Ferritin (nguml) Total carbon dioxide (mEqul) Calcium (mmolul) Intact parathyroidhormone (pguml) Phosphorus (mgudl) ALT (IUul) ALP (IUul) Albumin (gudl) Total cholesterol (mgudl) Triglycerides (mgudl) Fibrinogen (mgudl) IgM (gul) IgG (gul) CRP (mgul) a1 Acid-glycoprotein (gul) a1-Antitrypsin (gul) Lipoprotein(a) (mgudl) Thrombomodulin (nguml) Von Willebrand factor antigen (%) Plasminogen activator inhibitor-1 antigen (nguml) Endogenous Epo (IUul)**

1.89 (0.87–8.88) 5.69"1.57 9.34"1.80 169 (92– 448) 87.0 (17.0–231) 240 (147– 429) 34.2 (9.29–91.7) 540 (15.1–1687) 21.6"3.21 2.10"0.29 112 (8.96–1150)* 5.70 (2.43–13.6) 21 (4 –256) 82 (35–261) 3.73"0.47 158 (95–292) 93 (38–310) 312 (171–548) 1.12 (0.31–3.54) 12.9 (6.60–25.4) 7.50 (0–280) 1.04 (0.54 –3.90) 1.45"0.38 14.5 (0–93.3) 11.3 (5.88–30.7) 129"27.2 31.1 (11.5–65.9) 9.88 (0–88.4)

*Mean"1 SD, 210"255 pguml; **measured in 36 patients not treated with rHuEpo. Data are shown as mean"1 SD or median (range) depending on their normal or skewed distribution.

of CVD ( x2s4.73, Ps0.029) (Figure 1D). Median HGF levels in patients stratified by the above dichotomous variables are shown in Figure 1. There was a non-linear positive association between HGF and CRP levels (rhos0.289, Ps0.002), and a negative correlation between HGF and total cholesterol levels (rs 0.276, Ps0.010). No statistically significant associations were found between HGF levels and gender, age, renal failure cause, type of dialysate buffer or dialyser membrane, nPCR, body mass index, smoking, or the other laboratory parameters listed in Table 1. Correlations between HGF and pre-dialysis systolic BP (rs0.128, Ps0.240), diastolic BP (rs 0.094, Ps0.391) or pulse pressures (rs0.119, Ps0.274) were non-significant even when adjusted for the number of antihypertensive drugs used. There were no associations between HGF levels and use of any class of these medications, either (data not shown). Pre-dialysis HGF levels positively correlated with the dose of UFH (rs0.333, Ps0.022) (Figure 2), but not with that of enoxaparin (rs0.199, Ps0.225). The positive association between the use of rHuEpo treatment and HGF levels ( x2s7.15, Ps0.007) was independent of the dose of the hormone and duration of the therapy (data not shown). In 36 patients not treated with rHuEpo, serum HGF positively correlated with Hb levels (rs0.365, Ps0.029) (Figure 3), but not with those of endogenous Epo (rhos0.110, Ps0.539).

640


Fig. 1. Bivariate logistic regressions between serum HGF and type of heparin (A), viral hepatitis marker seropositivity (B), rHuEpo treatment (C), and prevalence of CVD (D).

Independent predictors of serum HGF

Fig. 2. Relation between serum HGF levels and UFH dose.

Fig. 3. Relation between serum HGF and Hb levels in HD patients not treated with rHuEpo.

For the multivariable analysis, the parameters showing significant correlations with serum HGF levels either by linear or logistic bivariate analysis were considered as potential independent variables. Additional analyses were, however, performed in order to exclude intervening variables. Viral hepatitis seropositivity was positively associated with ALT ( x2s12.8, Ps0.0004) and sTM levels ( x2s14.9, Ps0.0001). There was a direct correlation between ALT and ALP (rs0.395, Ps0.0002), and an inverse one between ALT and total cholesterol levels (rs 0.314, Ps0.002). Based on these associations and our previous findings w11x, the above parameters were not considered for the multivariable analysis as being intervening variables, causally linked with the presence of viral liver disease. The duration of time on HD directly correlated with HGF levels (rs0.489, P-0.0001) as well as with positive hepatitis markers ( x2s32.1, P-0.0001) and use of UFH (vs enoxaparin; x2s10.4, Ps0.001). The patients with hepatitis markers vs those without were more frequently anticoagulated with UFH (ns29 (74.4%) vs ns18 (38.3%), Ps0.001). These reciprocal associations could reflect our bias to the preferred use of UFH in the long-term HD patients with prevalent viral hepatitis and, as a result, with higher HGF levels. Therefore, both HD duration and type of

HGF in HD patients

641

heparin were included in the multivariable model to establish whether these confounders could be predictive of serum HGF when adjusted for viral hepatitis status. Of the significant bivariate correlates of HGF, presence of CVD was directly associated with increased CRP ( x2s9.06, Ps0.003) and a1-antitrypsin ( x2s5.89, Ps0.012) levels. These inflammatory markers also correlated with one another (rhos0.296, Ps0.001). Therefore, they were not entered into the multivariable analysis as being intervening variables, reflecting a state of chronic inflammation underlying the development of CVD in chronic HD patients w10x. Moreover, the use of CRP as an independent variable was not allowed due to its non-normal distribution. Following the above analyses, the independent variables for the multivariable analysis of the HGF level were duration of time on HD, viral hepatitis status, type of heparin, use of rHuEpo, CVD status, and KtuV. The backward stepwise multiple regression analysis (Table 2) showed that long time on HD and presence of CVD independently predicted high HGF levels, with time on HD being the stronger predictor. When HD duration was excluded from the analysis, hepatitis marker seropositivity became the strongest independent predictor of high HGF levels, followed by presence of CVD and use of rHuEpo treatment. In the latter analysis, use of UFH (vs enoxaparin) was a near-significant (Ps0.051) positive predictor of high HGF levels. The stepwise multiple linear regression analyses with forward selection yielded the same results (data not shown).

BuC markers, nine (45.0%) had evidence of CVD, and six (30.0%) were anticoagulated with UFH (vs enoxaparin). Following a 4-week course of the rHuEpo treatment, pre-dialysis HGF levels increased from 1.54"0.34 to 1.80"0.36 nguml (paired Student’s t-test Ps0.020) (Figure 4). The post-treatment HGF levels were not corrected for haemoconcentration because, despite a prominent increase in Hb levels (7.99"2.14 vs 9.95"1.64 gudl, P-0.0001), there was no rise in serum total protein or albumin concentrations (data not shown). Platelet counts increased from 175 (141–202) to 204"54 3 103uml (Wilcoxon’s matched pairs test Ps0.003), whereas white blood cell counts remained unchanged (5.69"1.50 vs 5.64" 1.59 3 103uml, paired Student’s t-test Ps0.741). There was no correlation between the increase in platelet count and that in the HGF level (Spearman’s rhos0.147, Ps0.535). No changes in hepatitis markers, liver enzymes, inflammatory markers, or KtuV were noted (data not shown).

Effect of rHuEpo treatment on serum HGF To verify the unexpected independent association between the use of rHuEpo and increased serum HGF levels, we performed an additional, prospective, case-controlled study. Twenty maintenance HD patients were started on rHuEpo (Eprex) treatment at a fixed dose of 50 IUukg injected subcutaneously three times per week after each HD session. Six (30.0%) of them were seropositive for viral hepatitis

Fig. 4. Effects of rHuEpo treatment on serum HGF levels.

Table 2. Variables predicting serum HGF level in the backward stepwise multiple linear regression analysis All variables entered

Duration of dialysis removed*

Independent variable

Regression coefficient

Standard error

P value

Independent variable

Regression coefficient

Standard error

P value

Duration of dialysis CVD rHuEpo treatment Viral hepatitis marker UFH KtuV

0.277 0.213 0.176 0.182 0.141 0.048

0.037 0.077 0.074 0.086 0.081 0.191

0.016 0.029 0.064 0.100 0.176 0.615

Viral hepatitis marker CVD rHuEpo treatment UFH Kt uV

0.291 0.213 0.191 0.203 0.105

0.081 0.079 0.076 0.081 0.192

0.006 0.024 0.043 0.051 0.274

Multiple r for variables in the models0.612 (*0.571), multiple r2s0.374 (*0.326), adjusted r2s0.327 (*0.284), Fs7.88 (*7.74), P-0.0001 (*-0.0001), standard error of estimate: 0.321 (*0.331).

642

Discussion This study shows that serum HGF levels in maintenance HD patients are, in part, determined by inflammatory conditions such as viral liver disease and CVD. The novel finding of our study is the direct dependency of pre-dialysis HGF levels on rHuEpo treatment and, probably, on the use of UFH instead of LMWH enoxaparin for blood anticoagulation during HD procedures. Long duration of HD treatment was an independent positive predictor of increased HGF levels in the study by Malatino et al. w17x, as well as in our patients in the initial multivariable analysis. However, it is known that increasing time on HD directly relates to the development of viral liver disease w28x, whereas viral hepatitis is a model disease resulting in increased circulating HGF w2x. Both these facts were exemplified in our HD patients. In order to reveal whether the relationship between the HGF levels and the duration of HD treatment represents any specific effects of prolonged therapy or just reflects the increased prevalence of viral hepatitis, we performed another multivariable analysis with HD duration being excluded from the model. Then, the previously nonsignificant viral hepatitis seropositivity became the strongest independent predictor of increased HGF levels. These analyses indicate that a long time on HD therapy and increasing prevalence of viral hepatitis are interchangeable markers for chronic liver disease underlying the increase in HGF levels. Moreover, serum HGF levels in HD patients depend on the degree of liver damage anduor dysfunction. The clinically consistent associations between HGF levels and those of liver enzymes, liver-synthesized cholesterol and TM, the latter being an endotheliumderived marker of both viral infection and dysfunction of the liver in chronic HD patients w11x, support this dependency. Thus, our results obtained in the HD population are consistent with the theory that the serum HGF level raises in response to liver disease, and is further determined by the severity of liver dysfunction anduor reduced hepatic clearance of the cytokine w2x. Presence of CVD was the second strongest independent predictor of increased HGF levels. Because both CVD and HGF were directly associated with inflammatory markers such as CRP and a1-antitrypsin, our data confirm that subclinical chronic inflammation underlying the development of CVD w10x is an important stimulus for HGF synthesis in HD patients w17x. The increase in HGF in this situation is likely to represent a part of the counter-system against cardiovascular damage. We did not find any associations between pre-dialysis blood HGF levels and those of EC dysfunction markers such as von Willebrand factor and plasminogen activator inhibitor-1. This suggests that plasma levels of these reactants are not useful estimates of HGF protection against vascular damage. At variance with previous studies w17,19x, HGF levels were not related to pre-dialysis BP in


our cohort even after adjustment for the intensity of antihypertensive treatment. These w17,19x and our findings are, however, in contrast to the positive relationship between HGF and arterial BP levels in hypertensive subjects without renal disease w4,5x. It implies existence of different mechanisms governing HGF production in response to hypertensive vascular injury in dialysis patients compared with the general population. Regarding inflammatory markers studied previously by Malatino et al. w17x, we confirm that CRP is a good correlate of serum HGF levels, and that serum albumin is not. We did not, however, find serum IgG to be related to HGF w17x, which could be due to the higher number of patients with viral hepatitis in our population studied. Moreover, we found a1-antitrypsin to be another significant correlate of HGF. No such relations were evidenced for fibrinogen, IgM, a1 acid-glycoprotein and lipoprotein(a). To our knowledge, this is the first study showing that rHuEpo treatment increases serum HGF levels in HD patients. Evidence for this phenomenon has been derived from the multivariable analysis where the use of rHuEpo therapy was an independent positive predictor of HGF levels. It has been further confirmed by the short-term prospective study in which rHuEpo treatment produced a 17% increase in serum HGF. It is a matter of speculation whether the effect is a result of tissue damage or specific stimulation of cells that can both express receptors for rHuEpo and synthesize HGF. Although megakaryocytes w2,20x and leukocytes w2,15,21x are cells of this type, there was no change in the white blood cell count or association between the increase in the number of circulating platelets and that of the serum HGF level following the rHuEpo treatment. This provides indirect evidence against the latter hypothesis. On the other hand, based on well-established erythropoietic actions of HGF w8,9x, the increase in this reparative cytokine during rHuEpo therapy may be viewed as synergistic and desirable. Malatino et al. w17x have been the first to report that Hb levels are directly associated with serum HGF in dialysis patients. However, they did not investigate associations between rHuEpo treatment itself and HGF levels, and from their data one could not ascertain whether improved anaemia was a result of this therapy, increased HGF, or both. In the present study, we found the positive correlation between Hb and HGF levels in the subgroup of HD patients not receiving rHuEpo. On the other hand, there was no association between serum HGF and endogenous Epo levels in these subjects. Therefore, our data imply that HGF stimulates erythropoiesis in maintenance HD patients, and that this effect is not linked with enhanced release of Epo. It is plausible that the anti-anaemic actions of HGF in HD patients are due to enhanced proliferation and differentiation of bone marrow erythroid progenitors w8,9x. Heparin is an important modulator of HGF blood levels and actions due to its ability to induce HGF

HGF in HD patients

production in several cell types w22x, and displace HGF from its tissue complexes with heparan sulfate proteoglycans w23x. Heparin can also bind and stabilize circulating HGF w24x, and facilitate its c-Met receptor activation w25x. In maintenance HD patients, heparin seems to be partially responsible for the repeated, marked, and prolonged increase in HGF levels w13,15x. Sugimura et al. w13x showed that HD sessions employing UFH resulted in increased posttreatment HGF levels, while the procedures using nafamostat mesilate were devoid of this effect. Rampino et al. w15x showed that both HD with and without heparin produced a marked rise in HGF, the peak serum HGF level being higher in the standard HD than in the non-heparin one. The HD-induced increase in circulating HGF is prolonged and evident during at least 24 h following the end of dialysis w16x. In our analysis, the use of UFH instead of LMWH enoxaparin for blood anticoagulation during HD procedures was a near-significant (Ps0.051) independent positive predictor of pre-dialysis serum HGF levels. Moreover, the HGF level directly correlated with the dose of UFH. These effects of different heparins have not been studied in HD patients so far. However, UFH has been found to be more potent than LMWH in inducing the increase in serum HGF when injected intravenously to healthy volunteers w26,27x. In another study w28x, the increase in HGF levels in patients with coronary artery disease was directly dependent on the dose of i.v. UFH. The different abilities of UFH and LMWH to increase HGF may be due to heparin molecule size, which affects its affinity to bind with HGF w24x, as well as to different hepatic clearance of their plasma complexes with the cytokine w29x. Prospective studies are needed to verify our hypothesis that UFH in the doses used for HD induces higher serum HGF levels as compared with LMWH. Regarding other variables that are linked with increased co-morbidity anduor specific organ damage in HD patients, serum HGF levels were not affected by either gender, age, smoking, underlying renal disease, body mass index, nPCR, type of dialysate buffer and dialyzer, or markers of iron metabolism, parathyroid function, and metabolic acidosis. In conclusion, increased serum HGF levels in chronic HD patients represent a part of the reparative reaction against liver and cardiovascular damage. The anti-anaemic effect of HGF itself, as well as the increase in serum HGF due to rHuEpo treatment, may synergistically augment erythropoiesis. Prospective studies are needed to establish whether the use of UFH instead of its low molecular weight derivative during HD procedures is more beneficial in these high cardiovascular risk and anaemic patients. Acknowledgements. We thank Dr Krystyna Pawlak for her technical contribution to this study. This work was supported by grants No. 4 PO5B 014 15 from the National Research Committee (Komitet Badan´ Naukowych), Warsaw, Poland, and No. 4-54669

643 from the Medical Academy (Akademia Medyczna), Biakystok, Poland.

References 1. Zarnegar R, Michalopoulos GK. The many faces of hepatocyte growth factor: from hepatopoiesis to hematopoiesis. J Cell Biol 1995; 129: 1177–1180 2. Vargas GA, Hoeflich A, Jehle PM. Hepatocyte growth factor in renal failure: promise and reality. Kidney Int 2000; 57: 1426–1436 3. Nakamura Y, Morishita R, Nakamura S et al. A vascular modulator, hepatocyte growth factor, is associated with systolic pressure. Hypertension 1996; 28: 409– 413 4. Nakamura S, Morishita R, Moriguchi A et al. Hepatocyte growth factor as a potential index of complication in diabetes mellitus. J Hypertens 1998; 16: 2019–2026 5. Nishimura M, Ushiyama M, Nanbu A, Ohtsuka K, Takahashi H, Yoshimura M. Serum hepatocyte growth factor as a possible indicator of arteriosclerosis. J Hypertens 1997; 15: 1137–1142 6. Van Belle E, Witzenbichler B, Chen D et al. Potentiated angiogenic effect of scatter factoruhepatocyte growth factor via induction of vascular endothelial growth factor: the case for paracrine amplification of angiogenesis. Circulation 1998; 97: 381–390 7. Nakagami H, Morishita R, Yamamoto K et al. Mitogenic and antiapoptotic actions of hepatocyte growth factor through ERK, Stat3, and Akt in endothelial cells. Hypertension 2001; 37: 581–586 8. Galimi F, Bagnara GP, Bonsi L et al. Hepatocyte growth factor induces proliferation and differentiation of multipotent and erythroid hemopoietic progenitors. J Cell Biol 1994; 127: 1743–1754 9. Takai K, Hara J, Matsumoto K et al. Hepatocyte growth factor is constitutively produced by human bone marrow stromal cells and indirectly promotes hematopoiesis. Blood 1997; 89: 1460 –1465 10. Zimmermann J, Herrlinger S, Pruy A, Metzger T, Wanner C. Inflammation enhances cardiovascular risk and mortality in hemodialysis patients. Kidney Int 1999; 55: 648–658 11. Borawski J, Naumnik B, Pawlak K, Mys´liwiec M. Soluble thrombomodulin is associated with viral hepatitis, blood pressure, and medications in haemodialysis patients. Nephrol Dial Transplant 2001; 16: 787–792 12. Borawski J, Naumnik B, Pawlak K, Mys´liwiec M. Endothelial dysfunction marker von Willebrand factor antigen in haemodialysis patients: associations with pre-dialysis blood pressure and the acute phase response. Nephrol Dial Transplant, 2001; 16; 1442–1447 13. Sugimura K, Kim T, Goto T et al. Serum hepatocyte growth factor levels in patients with chronic renal failure. Nephron 1995; 70: 324 –328 14. Sugimura K, Lee C-CR, Kim T et al. Production of hepatocyte growth factor is increased in chronic renal failure. Nephron 1997; 75: 7–12 15. Rampino T, Libetta C, de Simone W et al. Hemodialysis stimulates hepatocyte growth factor release. Kidney Int 1998; 53: 1382–1388 16. Rampino T, Arbustini E, Gregorini M et al. Hemodialysis prevents liver disease caused by hepatitis C virus: role of hepatocyte growth factor. Kidney Int 1999; 56: 2286–2291 17. Malatino LS, Mallamaci M, Benedetto FA et al. Hepatocyte growth factor predicts survival and relates to inflammation and intima media thickness in end-stage renal disease. Am J Kidney Dis 2000; 36: 945–952 18. Niu MT, Coleman PJ, Alter MJ. Multicenter study of hepatitis C infection in chronic hemodialysis patients and hemodialysis center staff members. Am J Kidney Dis 1993; 22: 468– 473 19. Nishimura M, Ushiyama M, Maruyama Y, Mabuchi H, Takahashi H, Yoshimura M. Association of human hepatocyte growth factor with hemodialysis hypotension. Hypertens Res 2000; 23: 581–586 20. Dessypris EN, Graber SE, Krantz SB, Stone WJ. Effects of recombinant erythropoietin on the concentration and cycling


644

21. 22.

23. 24. 25.

status of human marrow hematopoietic progenitor cells in vivo. Blood 1988; 72: 2060–2062 Sela S, Shurtz-Swirski R, Sharon J et al. The polymorphonuclear leukocyte–a new target for erythropoietin. Nephron 2001; 88: 205–210 Matsumoto K, Nakamura T. Heparin functionates as a hepatotrophic factor by inducing production of hepatocyte growth factor. Biochem Biophys Res Commun 1996; 227: 455– 461 Ashikari S, Habuchi H, Kimata K. Characterization of heparan sulfate oligosaccharides that bind to hepatocyte growth factor. J Biol Chem 1995; 270: 29586–29593 Zioncheck TF, Richardson L, Liu J et al. Sulfated oligosaccharides promote hepatocyte growth factor association and govern its mitogenic activity. J Biol Chem 1995; 270: 16871–16878 Sakata H, Stahl SJ, Taylor WG et al. Heparin binding and oligomerization of hepatocyte growth factoruscatter factor

26.

27. 28.

29.

isoforms: heparan sulfate glycosaminoglycan requirement for Met binding and signaling. J Biol Chem 1997; 272: 9457–9463 Seidel C, Hjorth-Hansen H, Bendz B et al. Hepatocyte growth factor in serum after injection of unfractionated and low molecular weight heparin in healthy individuals. Br J Haematol 1999; 105: 641–647 Salbach PB, Bru¨ckmann M, Turovets O, Kreuzer J, Ku¨bler W, Walter-Sack I. Heparin-mediated selective release of hepatocyte growth factor in humans. Br J Clin Pharmacol 2000; 50: 221–226 Okada M, Matsumori A, Ono K, Miyamoto T, Takahashi M, Sasayama S. Hepatocyte growth factor is a major mediator in heparin-induced angiogenesis. Biochem Biophys Res Commun 1999; 255: 80–87 Kato Y, Liu KX, Nakamura T, Sugiyama Y. Heparinhepatocyte growth factor complex with low plasma clearance and retained hepatocyte proliferating activity. Hepatology 1994; 20: 417–424

Received for publication: 28.5.01 Accepted in revised form: 14.11.01