Serum levels of p55 and p75 soluble TNF receptors in adult acute leukaemia at diagnosis: correlation with clinical and biological features and outcome

Authors


Dr GiovanniPizzolo Cattedra di Ematologia, Ospedale Policlinico, 37134 Verona, Italy.

Abstract

The tumour necrosis factor (TNF)/TNF-receptor (TNFR) complex plays a role in the growth of leukaemic cells. We retrospectively investigated the relationship between pre-treatment serum concentration of soluble TNFR (p55- and p75-sTNFRs) and outcome in adult acute myeloid (AML 82 cases) and lymphoid (ALL 44 cases) leukaemia. Both sTNFRs were significantly higher in AML (p55-sTNFR 4.53 ± 3.7, median 3.75; p75-sTNFR 6.51 ± 5.25 ng/ml, median 4.72) and ALL sera (3.31 ± 1.5, median 2.95; 5.30 ± 2.3 ng/ml, median 4.56, respectively) than in controls (1.89 ± 0.5, median 1.98; 2.22 ± 0.8 ng/ml, median 2.37) (P < 0.01 for both sTNFRs). Fresh leukaemic cells expressed p55- and p75-sTNFRs, which were modulated and released into the supernatant (SN) following short-term in vitro culture, suggesting that in vivo sTNFRs were also leukaemia-derived. Whereas no correlation was observed between sTNFRs and outcome in ALL, in AML higher p55-sTNFR levels (> 3.75 ng/ml) were associated with shorter disease-free survival (DFS) (P = 0.006) and overall survival (OS) (P = 0.0004). At multivariate analysis p55-sTNFR was the most significant predictor of DFS (P = 0.006) and OS (P < 0.001). Our data suggest that the prognostic significance of p55-sTNFR in AML could be related to relevant biological features of AML blasts.

Experimental evidence suggests that tumour necrosis factor (TNF) is involved in the proliferation of acute leukaemia cells ( Hoang et al, 1989 ; Oster et al, 1989 ; Elbaz et al, 1991a , b; Delwelet al, 1992 ; Carter et al, 1994 , 1996). TNF acts as a pleiotropic cytokine enhancing leukaemic growth ( Hoang et al, 1989 ; Oster et al, 1989 ; Carter et al, 1994 , 1996), inducing the transcription of other proliferation-relevant cytokine or cytokine receptor genes ( Elbaz et al, 1991a , b; Delwelet al, 1992 ; Vinante et al, 1993 , 1996; Khoury et al, 1994 ; Carter et al, 1996 ) and triggering apoptosis ( Elbaz et al, 1991b ; Delwel et al, 1992 ; Khoury et al, 1994 ; Carter et al, 1996 ).

Both TNF receptors (p55- and p75-TNFR), which bind to TNF with roughly equal affinity ( Bazzoni & Beutler, 1995; Gruss & Dower, 1995), have been reported to be expressed on acute leukaemia cells and to mediate distinct either inhibitory or stimulatory effects ( Delwel et al, 1992 ; Carter et al, 1996 ). In acute myeloid leukaemia (AML) blasts the engagement in vitro of p55-TNFR can be followed by enhanced DNA synthesis ( Hoang et al, 1989 ; Oster et al, 1989 ; Delwel et al, 1992 ; Carter et al, 1994 , 1996), induction of granulocyte-macrophage colony stimulating factor (GM-CSF) ( Carter et al, 1996 ) and regulation of the receptors for stem cell factor (SCF), granulocyte colony stimulating factor (G-CSF) and GM-CSF ( Elbaz et al, 1991a , b; Delwelet al, 1992 ; Khoury et al, 1994 ; Carter et al, 1996 ) and by cytotoxic effects ( Kobayashi et al, 1997 ). p75-TNFR seems to be mainly involved in the induction of secondary cytokines ( Delwel et al, 1992 ; Carter et al, 1996 ). In acute lymphoblastic leukaemia (ALL) TNF has been reported to exert either proliferative or cytotoxic effects on primary blast cells ex vivo ( Carter et al, 1994 ; Kobayashi et al, 1997 ), though its functional role is supported by less clear-cut data than in AML. Similar effects have been reported in leukaemic cell lines ( Kobayashi et al, 1997 ). Moreover, endogenous TNF has been demonstrated to be correlated with resistance to TNF or doxorubicin treatment in either cell lines or primary blasts ( Kobayashi et al, 1997 ).

Following TNF binding, TNFRs are partly internalized leading to mobilization of cytoplasmic factors including NFkB, and partly cleaved by proteases into soluble forms ( Schall et al, 1990 ; Loetscher et al, 1990 ; Heller et al, 1990 ; Brockhaus et al, 1990 ; Dembic et al, 1990 ; Smith et al, 1990 ; Kohno et al, 1990 ; Hwang et al, 1993 ; Higuchi & Aggarwal, 1994; Bazzoni & Beutler, 1995; Gruss & Dower, 1995; Heaney & Golde, 1996). The soluble forms compete with surface receptors for TNF, resulting in a potent inhibition of TNF activities ( Kohno et al, 1990 ; Bazzoni & Beutler, 1995; Gruss & Dower, 1995; Heaney & Golde, 1996). In some instances sTNFRs may function as TNF carriers or be shed independently of TNF ligation ( Bazzoni & Beutler, 1995; Heaney & Golde, 1996).

Although data generated in vitro suggest that TNF/TNFRs complex plays a role in the pathogenesis of acute leukaemia proliferation, no clinical evidence has been so far provided on the possible involvement of TNFRs in vivo. In the present study we partially addressed this issue by investigating the circulating levels of sTNFRs in acute leukaemia at diagnosis looking for a possible correlation with clinical and biological features and outcome.

MATERIALS AND METHODS

Patients

Pre-treatment sera used for sTNFRs determination were obtained from 126 adult patients with de novo acute leukaemia referred to our Institution from January 1985 to December 1994. Sera from patients observed thereafter were not included in the study since their median follow-up period was considered too short for a reliable assessment of their outcome and due to the substantial modification of the induction treatment strategies introduced in January 1995 for our AML patients. Patients whose serum samples were investigated represented 60% of all cases observed from 1985 to 1994. Sera of the remaining cases were not investigated for the following reasons: (a) sample unavailability (29%); (b) samples from patients with concomitant infection, a condition associated with increased serum concentration of sTNFRs (8%) (12); (c) samples from patients excluded from standard treatment because of advanced age (> 70 years) and/or poor performance status (ALL one, AML six cases) and/or clear-cut evidence of secondary leukaemia.

Eighty-two patients had AML and 44 ALL. No patient with FAB-L3 Burkitt's type ALL was included. Patients' characteristics are detailed in Table I. The diagnosis of AML or ALL and their subtypes was based on clinical findings and on established morphological, cytochemical and cytofluorimetric features of peripheral blood and/or bone marrow (BM) cells. Cytogenetic findings were available in only 74 cases: 41/82 AML and 33/44 ALL. In AML, 14/41 had a normal karyotype, 12/41 were unfavourable (deletions of various types and degree and/or complex rearrangements), and 15/41 were favourable [14 t(15;17) (q22; q12) and one inv(16)]. A normal karyotype was observed in 19/33 ALL, Ph1 chromosome in 14/33.

Table 1. Table I. Patients' characteristics and pre-treatment serum concentration of p55- and p75-sTNFR.Thumbnail image of

Treatment

AML patients received three induction courses of the following combinations: i.v. doxorubicin (35 mg/m2, days 1 and 2), i.v. cytosine arabinoside (100 mg/m2, days 1–7), oral 6-thioguanine (100 mg/m2, days 1–7) (courses 1 and 3) and i.v. doxorubicin (50 mg/m2, day 1), i.v. vincristine (1.3 mg/m2, day 2), i.v. cytosine arabinoside (500 mg/m2, days 3–8) (course 2). Induction treatment was followed by 1-year maintenance therapy, alternating each month course A [i.v. doxorubicin (25 mg/m2, day 1), i.m. cytosine arabinoside and oral 6-thioguanine (both 100 mg/m2, days 1–5)] with course B [i.v. vincristine (1 mg/m2, day 1), i.v. VP-16 (100 mg/m2, days 1–3), i.m. cytosine arabinoside (100 mg/m2, days 1–3)]. Bone marrow transplantation (BMT) was performed in first complete remission in 15 cases (allogeneic four cases, autologous 11 cases). In 3/14 FAB M3 patients treatment included all-trans retinoic acid.

ALL patients were treated with induction therapy (ALLVR74: eight cases; ALLVR89: 36 cases) including i.v. daunorubicin, vincristine, prednisone and asparaginase, followed by consolidation therapy, central nervous system chemo- and radio-prophylaxis, and periodic re-induction courses over a 3-year maintenance period, as previously reported ( Todeschini et al, 1994 , 1998). ALLVR89 regimen was characterized mainly by high dosages of daunorubicin delivered during the early phase of induction period ( Todeschini et al, 1998 ). One high-risk ALL patient underwent allogeneic BMT.

Detection of soluble TNFR and TNF

Serum and/or supernatant (SN, see below) was kept frozen at −70°C until use and then tested in duplicate for the concentration of sTNFRs and TNF. sTNFRs were measured in all 126 patients using an enzyme-linked immunological binding assay [ELIBA (Roche, Basel, Switzerland)], as previously described ( Trentin et al, 1995 ). TNFα was measured in 24 AML (four M0, four M1, five M2, three M3, five M4, three M5) using a commercially available EASIA (Medgenix, BioSource Europe, Fleurus, Belgium). Preliminary experiments showed that sample deep freezing did not affect sTNFR or TNF determination. p55-, p75-sTNFR and TNF mean concentrations in 88 healthy blood donors were 1.89 ± 0.5 (range 0.6–3.2), 2.22 ± 0.8 (range 0–3.5) ng/ml and 0.97 ± 5.11 (range 0–28) pg/ml, respectively.

Isolation of leukaemic cells

Viable leukaemic cells purified by conventional methods from freshly heparinized peripheral blood obtained from 22 cases with high circulating blast count [geqslant R: gt-or-equal, slanted 30 × 109/l; AML 12 cases (two M1, two M2, one M3, four M4, three M5), ALL 10 cases (four B- and six T-lineage derived)] were frozen in liquid nitrogen. In all cases frozen cell samples contained > 95% blasts. Cell viability after thawing was always > 90%, as assessed by Trypan-blue staining. Leukaemic cells were used in short-term cultures.

Cultures of leukaemic cells

Thawed cells were added to a 24-well plate (Falcon, Lincoln Park, N.J.) and incubated for 72 h at 37°C in 5% CO2 in RPMI-1640 (Gibco Life Technologies, Paisley) supplemented with 10% heat-inactivated fetal calf serum (FCS) (Gibco Life Technologies), and streptomycin (50 μg/ml), each well containing 2 × 106/ml blasts. After blast harvesting, SN were collected, concentrated 15 times using Centricon 10 (Amicon, Donyers, Mass.), and frozen at −70°C until use.

Flow cytometric detection of membrane TNFRs

Flow cytometry analysis of p55 and p75-TNFRs membrane expression by uncultured and cultured blasts was performed using HTR-9 and UTR-1 mAbs, respectively (gift of Dr M. Brockhaus, Basel, Switzerland) ( Brockhaus et al, 1990 ). Primary mAb incubation was followed by a biotinylated second antibody [goat anti-mouse Ig-G1 (Amersham Corp., U.K.)] and then by PE-conjugated streptavidin (Becton Dickinson, Sunnyvale, Calif.). An irrelevant IgG1 mAb was used as control. The analysis was carried out with a FACScan flow cytometer using the Lysis II software (Becton Dickinson). The results were expressed as mean fluorescence intensity (MFI, 0–1023 channels) minus the control.

RT-PCR detection of TNFRs mRNA

Total cellular RNAs of purified blasts were isolated as previously reported ( Vinante et al, 1993 ). 4 μg of RNA were reverse transcribed using universal primers and 1.25 U of AMV reverse transcriptase (Gibco Life Technologies) (20 μl final volume). cDNA was PCR amplified using the following primers (Genenco, m-medical, Florence, Italy): p55-TNFR (441 bp) sense 5-GAAGTCCAAGCTCTACTCCAT-3 and antisense 5-ATCGATCTCGTGGTCGCTCAG-3; p75-TNFR (418 bp) sense 5-AGAGAG-

AAGCCAAGGTGCCTC-3 and antisense 5-CAGCTGTGACCGAAAGGCACAT-3; Vimentin (266 bp) sense 5-GCTCAGATTCAGGAACAGCAT-3 and antisense 5-TAAGGGCATCCACT-TCACAGG-3. Controls for any contaminating genomic DNA were also performed. The cDNA was denatured for 5 min at 94°C followed by 30 cycles in a thermal cycler (GeneAmp PCR System 2400, Perkin Elmer, Norwalk, Ct.) using 1.25 U of Taq polymerase (Promega, Madison, Wis.) in 50 μl (94°C 40 s, 60 or 57°C 40 s, depending on the primer base composition, 72°C 50 s) followed by 5 min at 72°C. PCR products were separated by electrophoresis on 1.5% agarose gel.

Statistics

Student's t-test (after data normalization) for paired or unpaired samples, ANOVA or Pearson's correlation test were used to compare the sTNFR values with other variables. The design of the study was a retrospective evaluation with multivariate analysis, assuming (a) the following end-points: achievement of complete remission (CR) (failures included blast persistence or death from any cause during induction), disease-free survival (DFS) (patients who did not achieve CR were not considered and failures included relapse or death from any cause) and overall survival (OS) (failures included death at any time from any cause); (b) the following prognostic factors at diagnosis, chosen on the basis of their significance in univariate analysis or of their recognized relevance to outcome in AML or ALL: pre-treatment serum concentrations of p55- and p75-sTNFR, age and WBC count as continuous variables, and FAB subtypes (M4–M5 versus other AML FAB types), spleen size (geqslant R: gt-or-equal, slanted 2 v < 2 cm), karyotype (normal, favourable, unfavourable, unknown), sex (F v M) and BMT (only for DFS) as discrete variables. Response to treatment was analysed using a logistic regression (LR) model. The Cox's proportional hazard regression model ( Cox, 1972) was applied to the DFS and OS analyses. The Odds Ratios/Relative Risks were calculated for 1 SD increase, except for leucocytes, for which an arbitrarily chosen increase of 20 × 109/l was used. DFS and OS actuarial curves were estimated ( Kaplan & Meier, 1958) after recoding p55-sTNFR serum concentrations as a two-level variable, with its median value as cut-off, using the Log-Rank test to evaluate the difference between the two levels. Statistics for comparisons were regarded as significant for a P value < 0.05.

RESULTS

Soluble TNFRs levels in pre-treatment sera

As shown in 1Table I and Fig 1, the values of p55- and p75-sTNFRs were significantly higher in AML (4.53 ± 3.7, median 3.75, and 6.51 ± 5.25 ng/ml, median 4.72, respectively) and ALL (3.31 ± 1.5, median 2.95, and 5.30 ± 2.3 ng/ml, median 4.56, respectively) as compared to controls (1.89 ± 0.5, median 1.98, 2.22 ± 0.8 ng/ml, median 2.37) (AML and ALL v controls: P < 0.01 for both sTNFRs; AML v ALL: P= 0.035 and P= 0.61 for p55- and p75-sTNFR, respectively). Circulating sTNFRs higher than the median value observed in healthy controls were detected in 79.5% of ALL and 89% of AML (p55-sTNFR), and in 97% of ALL and 89% of AML (p75-sTNFR). p55- and p75-sTNFR serum levels correlated poorly with each other (ALL, r= 0.46; AML, r= 0.42) as well as with age (AML, r= 0.14 and 0.16; ALL, r=0.12 and 0.14, respectively) and WBC count (AML, r= 0.49 and 0.28; ALL, r= 0.23 and 0.10, respectively). In AML, females had slightly lower values of p55-sTNFR (P= 0.023) than males. No sex-related difference was observed for either sTNFR in ALL. Patients with M4–M5 AML showed higher levels than those with M0–M3 subtypes (p55- and p75-sTNFR, 6.71 ± 5.1 and 9.78 ± 5.6 v 3.27 ± 1.6 and 4.82 ± 4.3 ng/ml, P < 0.001 for both sTNFRs). sTNFR values did not correlate with karyotype, though a trend towards lower levels of p55- and p75-sTNFR among favourable karyotypes was observed ( Table I). No significant difference was detected in ALL between B- and T-lineage for p55-sTNFR (P= 0.15). p75-sTNFR levels were slightly lower in T- than in B-ALL (P= 0.043).

Figure 1.

.

TNFRs expression and release by leukaemic cells

We investigated whether sTNFRs actually derived from leukaemic cells. To this purpose, a subset of 22 cases with a high circulating blast count (geqslant R: gt-or-equal, slanted 30 × 109/l) were studied ex vivo. RT-PCR for TNFRs mRNA on purified uncultured blasts demonstrated in all 22 (10 ALL and 12 AML) cases the presence of the expected bands of 441 and 418 bp for p55- and p75-TNFR, respectively. This was confirmed by cytofluorimetric analysis of the same cases, all of which expressed both p55- and p75-TNFRs at basal conditions. The typical membrane pattern of p55- and p75-TNFRs compared with RT-PCR mRNA expression is depicted in Fig 2 in one representative ALL (panel A) and AML (panel B) patient. The overall intensity of p55- and p75-TNFRs expression was rather low with some variability in individual patients. The MFI of p55- and p75-TNFRs was 130.1 ± 85.8 and 94.0 ± 81.0 in AML and 149.1 ± 83.4 and 119.8 ± 97.7 in ALL, respectively. After culture, the MFI increased significantly for both p55- and p75-TNFR in ALL (from 149.1 ± 83.4 and 119.8 ± 97.7 to 219.9 ± 81.3 and 202.3 ± 95.0, P= 0.046 and 0.036, respectively). In AML, this pattern was observed only for p75-TNFR MFI (which increased from 94.0 ± 81.0 to 154.7 ± 83.1, P= 0.016); p55-TNFR did not show a unique trend, a third of cases being characterized by a clear-cut decrease of MFI after culture. A spontaneous release of p55- and p75-sTNFRs into culture SN was documented in the majority of investigated cases (14/16 and 12/16 for p55- and p75-sTNFR, respectively), with values ranging from 0 to 160 and from 0 to 850 pg/ml, respectively. p55-sTNFR median concentration was 55 pg/ml in AML (range 20–160) and 16.5 pg/ml in ALL (range 0–328). p75-sTNFR median value was 19.5 pg/ml in AML (range 0–850) and 9 pg/ml in ALL (range 0–69). Overall, patients with high values of p55-sTNFRs in vivo tended to show high SN concentrations. No release could be documented in SN of unstimulated PBMC.

Figure 2.

66 bp) and p55- (441 bp) or p75-TNFR (418 bp).

sTNFRs serum concentration and outcome

To assess the prognostic power of sTNFRs we performed either univariate or multivariate analysis correlating the achievement of CR, DFS and OS with the pre-treatment concentration of sTNFRs and other parameters usually recognized as bearing prognostic significance. No association was found between p55-sTNFR or p75-sTNFR serum concentration and the achievement of CR, DFS or OS in ALL. On the contrary, the higher the serum levels of p55-sTNFR in AML the poorer was the outcome of patients in terms of CR, DFS and OS. The probability of no response, relapse or death associated with the variables included in the multivariate analysis (LR model for CR and Cox's model for DFS and OS) are listed in Table II and the significant variables are represented in Fig 3.

Table 2. Table II. Adjusted Odds Ratios (OR) for no response to treatment (Logistic Regression model) and Relative Risk (RR) for relapse and death (Cox's proportional hazard regression model) assigned to each covariate included in the multivariate analysis on evaluable AML patients. OR and RR are calculated for an increase of 20 × 109/l for WBC, and an increase of 1 SD for the remaining variables.* Continuous variables. Thumbnail image of
Figure 3.

Fig 3. Odds ratio (OR) for no response and relative risk (RR) for relapse and death as observed in multivariate analysis including 82 AML patients. Only variables significant for CR achievement (panel A), DFS (panel B) or OS (panel C) analyses were represented (see Table II). p55-sTNFR and age: continuous variables; FAB subtype: discrete variable (M4–M5 v other subtypes). Data are the OR and RR with 95% CIs on a logarithmic scale.

CR achievement

The LR model analysis applied to our AML patients showed that age (P= 0.014, OR = 2.1, CI = 1.16–3.88) and p55-sTNFR serum levels (P= 0.066, OR = 3.4, CI = 0.92–12.62) were associated with a low probability of achieving CR. The high significance shown by p55-sTNFR at univariate analysis (P= 0.008, OR = 4.39, CI = 1.47–13.11) was only partially retained (P= 0.066) after controlling for the covariates ( Table II and Fig 3).

DFS and OS

Univariate analysis showed that p55-sTNFR level was a significant risk factor for both DFS (P < 0.001, RR = 6.94, CI = 2.34–20.63) and OS (P < 0.001, RR = 2.56, CI = 1.85–3.55) in AML. In multivariate analysis, p55-sTNFR was found to be an independent prognostic factor for both DFS (P= 0.006, RR = 8.48, CI = 1.85–38.93) and OS (P < 0.001, RR = 2.51, CI = 1.63–3.86) together with age (OS, P < 0.001) and FAB subtype (OS, P= 0.048) ( Table II and Fig 3). Monocytic differentiation, which was associated with the highest levels of sTNFRs, had a lower prognostic impact in terms of OS than p55-sTNFR ( Table II). p75-sTNFR level, though a significant risk factor for DFS (P= 0.013, RR = 1.54, CI = 1.09–2.16) and OS (P= 0.004, RR = 1.34, CI = 1.10–1.63) in univariate analysis, lost its significance in multivariate analysis ( Table II). Fig 4 shows the DFS (panel A) and OS (panel B) curves in AML by using the p55-sTNFR median (3.75 ng/ml) as an arbitrary cut-off value ( Table III). The difference between the curves was statistically significant for both DFS (P= 0.006) and OS (P= 0.0004). The 5-year DFS and OS rates were only 11.8% and 9.0% respectively for AML patients with a p55-sTNFR value > 3.75 ng/ml, as opposed to 47.4% and 34.0% respectively for AML patients with a p55-sTNFR value leqslant R: less-than-or-eq, slant 3.75 ng/ml.

Figure 4.

.

Table 3. Table III. Covariate figures according to the median pre-treatment concentration of p55-sTNFR in 82 AML patients. * Continuous variables.Thumbnail image of

Relationship between serum levels of p55-sTNFR and TNF

We also examined if p55-sTNFR and TNF levels were correlated. To this purpose, we assessed the serum levels of TNF in 24 AML patients. Table IV summarizes the results grouped by the median value of p55-sTNFR, which defined two subsets of patients with different prognosis (Fig 4). Two more variables, CR rate and OS, were also correlated. Though the levels of TNF, when detectable, and p55-sTNFR did not correlate (r= 0.11), high TNF levels clustered in the group of patients presenting high p55-sTNFR and vice versa.

Table 4. Table IV. Serum TNF levels, CR rates and OS according to the median value of p55-sTNFR. *P < 0.01.Thumbnail image of

DISCUSSION

The present retrospective study shows that the serum levels of p55- and p75-sTNFR before treatment were elevated in our AML and ALL (P < 0.01 v controls), that blasts substantially contributed to the accumulation of circulating sTNFRs, and that the serum concentration of p55-sTNFR, but not of p75-sTNFR, was an important prognostic factor in 82 newly diagnosed AML patients.

Our main finding is represented by the demonstration that p55-sTNFR concentration, as a continuous variable, was an independent prognostic factor in AML. Higher p55-sTNFR serum concentration at diagnosis was associated with shorter DFS (P= 0.006) and 0S (P < 0.001). The OS in our AML series was 23% at 5 years, yet it was as low as 9.0% in patients with p55-sTNFR value > 3.75 ng/ml and as high as 34.0% in those with values leqslant R: less-than-or-eq, slant 3.75 ng/ml. The RR for relapse or death was 8.48 and 2.51, respectively, for 1 SD increase of p55-sTNFR levels.

These findings should be evaluated in the context of some peculiar aspects of our study, which, being retrospective, suffers a number of limitations. First, our series includes only a proportion (60%) of cases observed from 1985 to 1994. However, patients were selected on the sole basis of sample availability among those without infections at diagnosis, or secondary leukaemia, or excluded from standard treatment because of advanced age (> 70 years) and/or poor performance status. Second, our AML patients, who were observed during the 10-year period before 1995, received less effective treatments than those in use at present. However, our overall results appear in line with those reported by others for the same period ( van den Heuvel-Eibrink et al, 1997 ), taking into account the proportion of patients > 50 years of age (39/82, 47.6%). On the other hand, the long follow-up (at least 33 months in patients still alive, median 80 months) allows the results to be considered as consolidated. Furthermore, since supportive care changed constantly during the decade in which patients were observed, one could ask whether the observed prognostic effect of p55-sTNFR levels applies throughout the period. Indeed, this was the case (data not shown). Third, cytogenetic analysis was not available in all patients (41/82 AML cases), thus limiting the reliability of its impact on prognosis in relation to p55-sTNFR. With this limitation in mind, in our multivariate model, karyotype appeared devoid of any statistically significant prognostic relevance, at variance with literature data ( van den Heuvel-Eibrink et al, 1997 ). Nevertheless, our patients with favourable karyotype actually had a better outcome (DFS rate at 5 years was 7/15, 47%; OS rate 8/15, 53%) than that of other AML, suggesting that the lack of statistical significance might be due to the small size of our karyotype sample.

An explanation for the prognostic significance of p55-sTNFR serum level is that it might merely reflect the total bulk of leukaemic cells. At least two points do not fit with this interpretation. First, although p55- and p75-sTNFRs, which are probably released via similar mechanisms, were co-expressed by leukaemic cells, their serum levels were poorly correlated (r= 0.42 and 0.46 for AML and ALL respectively). Second, the prognostic impact of p55-sTNFR was independent of WBC count, which can be regarded as a parameter associated with the neoplastic mass.

In our AML patients, low p55-sTNFR levels were associated with low serum concentration of TNF, thus suggesting that the prognostic significance of p55-sTNFR might be related to TNF activity. In its soluble form, p55-TNFR is a potent regulator of TNF bioactivity ( Bazzoni & Beutler, 1995; Heaney & Golde, 1996). As a membrane-bound molecule, p55-TNFR is characterized by the presence of a death domain in its cytoplasmic tail and by NFkB mobilization following its triggering. In vitro p55-TNFR engagement on myeloid blasts enhances DNA synthesis in an autocrine fashion ( Hoang et al, 1989 ; Oster et al, 1989 ; Delwel et al, 1992 ; Carter et al, 1994 , 1996), and up-regulates IL-3/GM-CSF receptors ( Elbaz et al, 1991a ; Delwel et al, 1992 ; Carter et al, 1996 ) and GM-CSF production ( Carter et al, 1996 ). Conversely, p55-TNFR can transduce growth inhibition signals mainly down-regulating the receptors for SCF ( Khoury et al, 1994 ) and G-CSF ( Elbaz et al, 1991a ; Delwel et al, 1992 ; Carter et al, 1996 ) and can induce apoptosis of leukaemic cells ( Kobayashi et al, 1997 ). Instead, data available to date support the view that in AML p75-TNFR is involved mainly in the induction of secondary cytokines ( Delwel et al, 1992 ; Carter et al, 1996 ). If the two TNFRs play a different role in the cytokine loops which allow the growth of myeloid blasts, this might support a different prognostic value of p55- and p75-TNFR in AML. Finally, as far as the release of sTNFRs is concerned, a possible mutational mechanism leading to constitutive activation and shedding of TNFRs has been suggested ( Bazzoni & Beutler, 1995).

The adverse prognostic impact of high p55-sTNFR levels observed in AML was not found in ALL, despite similar patterns of expression and release. The reason for such a difference is obscure. It could depend on the different treatment given to ALL and AML patients. TNF gene regulation has been shown to involve the quite restricted ATF2/Jun and NAFTp and the ubiquitous NFkB ( McHugh & Rowland, 1997), also mobilized by TNFR-driven signalling ( Hsu et al, 1995 ). NFkB is down-regulated by steroids ( McHugh & Rowland, 1997), which were widely used in our ALL treatment ( Todeschini et al, 1994 , 1998), but not in AML.

In conclusion, our retrospective analysis demonstrated that the level of p55-sTNFR, but not of p75-sTNFR, correlates with outcome in AML. The definition of the biological basis of our findings was beyond the aim of this work. However, we have demonstrated that low levels of leukaemia-derived p55-sTNFRs were associated with low or undetectable levels of TNF and we have underlined that some pathways relevant to leukaemic cell survival seem to be under the control of transduction signals mediated through p55-TNFR rather than p75-TNFR. Hence, though prognosis in AML is eventually the result of complex interactions involving a great number of variables, this study has determined a new prognostic factor which is probably related to an underlying biological diversity of leukaemic blasts.

Acknowledgements

This work was supported by grants from Associazione Italiana per la Ricerca sul Cancro (AIRC, Milano), and from Progetto Sanità 96/97 Fondazione Cariverona.

Ancillary