HLA-DR4 predicts haematological response to cyclosporine in T-large granular lymphocyte lymphoproliferative disorders
A. John Barrett, Stem Cell Allotransplantation, NHLBI, NIH, Bldg 10, Rm 7C103, 9000 Rockville Pike, Bethesda, MD 20892, USA. E-mail: email@example.com
Summary. T-cell large granular lymphocytic lymphoproliferative disease (T-LGL) is often associated with life-threatening cytopenias. Twenty-five subjects with anaemia and/or neutropenia caused by T-LGL were treated with cyclosporin A (CSA) 5–10 mg/kg/d for at least 3 months. Eighteen patients survived between 35 and 77 months after starting treatment. Fourteen patients [56%; 95% confidence interval (CI) 35–76%] responded to CSA with sustained improvement in the neutrophil count or transfusion independence. Seven had complete normalization of blood counts, and four achieved a durable response only after the addition of erythropoietin. Sustained response required continued low-dose CSA. In a multivariate analysis, HLA-DR4 was highly predictive of CSA responsiveness (odds ratio 18; 95% CI 1·8–184). T-LGL subtype, LGL counts after therapy, lymphocytic marrow infiltration and bone marrow cellularity did not significantly affect the probability of response. We conclude that CSA is effective in inducing haematological responses in HLA-DR4-positive patients and that T-LGL is likely to have an immune pathogenesis.
T-cell large granular lymphocyte (T-LGL) proliferations are diagnosed on the basis of typical morphology, clinical associations (autoimmune disease and cytopenias) and the immunophenotype of an activated cytotoxic T lymphocyte (CTL) (CD3+, CD8+, CD4–, TCR αβ+, CD 57+, CD16+/–). Both clonal (T-LGL leukaemia or T-cell granular lymphocytic leukaemia) and non-clonal [LGL-lymphoproliferative disease (LGL-LP)] proliferations may be complicated by life-threatening autoimmune cytopenias in about two-thirds of patients (Loughran, 1993; Dhodapkar et al, 1994). T-LGL is associated with rheumatoid arthritis (RA) and may overlap with Felty's syndrome. Immunogenetic studies to differentiate between Felty's syndrome and LGL disease have shown that LGL patients with associated RA are mostly HLA-DR4 positive, whereas those without RA have similar HLA-DR4 frequencies to the normal population (Bowman et al, 1994; Starkebaum et al, 1997).
The optimum therapy is undefined, but case series have shown limited responses to low-dose methotrexate (10 mg/m2/week), steroids, cyclophosphamide (100 mg daily), purine analogues and granulocyte colony-stimulating factor (Dhodapkar et al, 1994; Loughran et al, 1994; Weide et al, 1994; Witzig et al, 1994). Recently, reports have described improvements in the cytopenias of T-LGL patients treated with cyclosporine A (CSA) (Bible & Tefferi, 1996; Sood et al, 1998). Responses to CSA occur despite the persistence of T-LGL cells and may require low-dose maintenance therapy.
Here, we describe the response rate, long-term outcome and prognostic factors for treatment response in a prospective trial of CSA in 25 patients with T-LGL and profound neutropenia or anaemia. HLA-DR4 was found to be strongly predictive for haematological response.
Patients. Twenty-five subjects with T-LGL, referred for the investigation of their cytopenias, gave informed consent for a trial of CSA under a National Heart, Lung and Blood Institute institutional review board approved protocol. Eligibility criteria included age of 18 years or over and no previous CSA therapy.
Diagnostic criteria. Diagnosis of T-LGL required a peripheral blood absolute LGL count of ≥ 0·3 × 109/l or an immunophenotype consistent with T-LGL (CD3+, CD8+, either CD57+ or CD16+, CD56+/–). In addition, the diagnosis of T–LGL accompanying myelodysplastic syndrome (T-LGL/MDS) required a clonal cytogenetic abnormality in myeloid cells or megakaryocytic dysplasia or myeloid dysplasia or 5–20% myeloblasts in the bone marrow aspirate.
Criteria for cytopenia. In addition, all subjects either had (1) severe neutropenia (absolute neutrophil count ≤ 0·5 × 109/l) or (2) severe thrombocytopenia (platelets ≤ 20 × 109/l) or (3) moderate thrombocytopenia (platelets ≤ 50 × 109/l) with active bleeding or (4) anaemia (Hb ≤ 9 g/dl) or (5) red cell transfusion requirement of ≥ 2 units/month for at least 2 months before therapy.
Laboratory studies. Flow cytometry, T-cell receptor (TCR) γ-gene rearrangement by polymerase chain reaction and high-resolution molecular typing for human leucocyte antigen (HLA) classes I and II were performed as described previously (McCarthy et al, 1992; Bunce et al, 1995; Saunthararajah et al, 2001). Subjects with HLA-DR4 were retyped for confirmation and consistency with current nomenclature.
Cyclosporine. Oral CSA was started at 5–10 mg/kg/d, in two divided doses, with dose adjusted to maintain a therapeutic level between 200 and 400 ng/ml. After 3 months, the dose of CSA was slowly tapered to that required in order to sustain a response or discontinued if there was no response, or after relapse. Erythropoietin, 10 000–20 000 U subcutaneously (s.c.) three times per week, was added for patients with an inadequate response to CSA alone.
Response criteria. Complete haematological response (CR) to CSA was defined as a complete normalization of blood counts. Partial response (PR) required a > 50% improvement in deviation of blood counts from normal in the absence of normalization of blood counts. For anaemia, a twofold improvement in transfusion requirements, or an increment in haemoglobin ≥ 2 g/dl, satisfied the criteria of PR. All responses had to be maintained for at least 2 months.
Statistical analysis. Statistical analyses were performed using statview 5.0.1. The normal approximation to the binomial distribution was used to estimate the 95% confidence intervals (CI) for overall mortality and response to CSA. Categorical factors were analysed by Fisher's exact test, and continuous variables were analysed by the non-parametric Mann–Whitney U-test. A P-value < 0·05 was considered to be significant.
Patient characteristics are summarized in Table I. The full spectrum of T-LGL was represented, including nine subjects with pure T-LGL leukaemia (clonal), 10 with T-LGL/MDS (one non-clonal, one oligoclonal, eight clonal) and six with T-LGL LP (one oligoclonal and five non-clonal). Molecular typing at the HLA-DR B1 locus showed the allelic frequency of HLA-DR4 to be 0·22 for all subjects with LGL disease, not significantly different from the DR4 frequency of the normal Caucasian population (0·17) by Fisher's exact test (two-tailed P = 0·26) (Mori et al, 1997). Only three patients in our series had refractory anaemia (RA), and the presence of HLA-DR4 in CSA responders was not likely to be confounded by coincidental RA.
Table I. Baseline patient characteristics.
|Gender (F/M)||14/11|| |
|Median patient age (years)||60||(21–79)|
|Median LGL count (× 109/l)||0·735||(0·076–4·481)|
|Median ANC (× 109/l)||0·546||(0–3·450)|
|Median platelet count (× 109/l)||0·2||(0·013–0·397)|
| RA||6|| |
| RARS||1|| |
| RAEB||3|| |
|Clonality by TCR analysis|
| Clonal||19|| |
| Oligoclonal||3|| |
| Non-clonal||3|| |
| Neutropenia||16|| |
| Thrombocytopenia||9|| |
| RBC transfusion dependent||12|| |
| Normal||22|| |
| Abnormal [t(4;11); −20q; der 18]||3|| |
| Hypocellular||8|| |
| Normocellular||6|| |
| Hypercellular||10|| |
|Lymphoid aggregates in the marrow||13/24|| |
|History of an immune disorder||15|| |
|Abnormal immunoglobulin levels||12|| |
|Antinuclear antibody positive||8/22|| |
|Rh factor positive||8/22|| |
A response to CSA was seen in 14 patients (overall response rate of 56%, 95% CI 35–75%) at a median of 70 d; there were seven CR, seven PR and 11 primary non-responders. Clinical outcomes with respect to disease subtype are described in Table II. Responses were sustained in 10 patients. Three patients converted to CR and one to a PR only after the addition of erythropoietin. Two other patients developed reversible erythroid hypoplasia with CSA.
Table II. Clinical outcomes by disease subtype.
Several factors were assessed for their effect on response (Table III). The presence of HLA-DRB1 04 (odds ratio 18·0, 95% CI 1·75–184; P = 0·01) and high baseline LGL counts (P = 0·03) were the two factors most closely associated with response in univariate analysis. However, the influence of baseline LGL count was eliminated as a contributory factor to response in logistic regression analysis. Responses were not associated with other HLA types, T-LGL subtype, the nature and degree of pretreatment cytopenia, age, lymphocytic marrow infiltration or marrow cellularity (Table IV).
Table III. Characteristics of responders.
| 59/F||T-LGL leukaemia||St/Ch/ATG||0.470||N||46xx||CR|| ||1101, 1302|
| 27/M||T-LGL leukaemia|| ||0.506||E||46xy|| * CR (epo)|| ||0301, 0401|
| 38/F||T-LGL leukaemia||GF/St||0.853||E, N, P||46xx||PR|| ||11, 1501|
| 28/F||LGL/MDS-RA||ATG||4.481||E, N||46xx||CR|| ||0101, 0408|
| 58/F||T-LGL leukaemia||St/Ab/GF/Ig||1.904||E, N||46xx||PR|| ||0401, 1601|
| 76/F||T-LGL leukaemia||St||0.798||E, N||46xx||CR|| ||11, 13|
| 61/M||T-LGL LP||GF/Ab/Sp||2.735||E, N||46xy||PR||+||0301, 0401|
| 54/F||T-LGL leukaemia||GF||2.304||E, N||46xx||PR|| ||0101, 0101|
| 40/M||LGL/MDS-RA(oligo)||Sp/GF||1.005||N||der(18)||PR|| ||0401, 0401|
| 79/F||T-LGL leukaemia||St||0.542||E, P||46xx||CR|| ||0301, 0401|
| 69/M||LGL/MDS-RAEB|| ||0.839||E, P||20q–|| * PR (epo)|| ||0401, 1101|
| 65/M||LGL/MDS-RA||St||0.285||E, N, P||46xy|| * CR (epo)|| ||0404, 1501|
| 67/F||T-LGL leukaemia|| ||0.735||E, P||46xx|| * CR (epo)|| ||0402, 1502|
| 43/F||T-LGL LP||Sp/St||1.614||E, N||46xx||PR|| ||1001, 1501|
| 60/F||T-LGL leukaemia||St/Ch/Ab||1.421||E, N||46xx||NR|| ||08, 11|
| 36/F||LGL/MDS-RA||St||0.370||E||t(4;11)||NR|| ||0404, 1103|
| 37/M||T-LGL leukaemia||St||1.453||E||46xy||NR|| ||14, 14|
| 71/M||LGL/MDS-RAEB||GF||0.76||E, N, P||46xy||NR|| ||0301, 0701|
| 67/F||LGL/MDS-RARS|| ||0.322||E||46xx||NR|| ||1510, 1510|
| 46/F||T-LGL leukaemia||GF/St/Ch||0.480||E, N, P||46xx||NR|| ||0101, 01|
| 21/M||T-LGL LP (oligo)||Sp/Ig/GF/St||0.386||N||46xy||NR|| ||0301, 11|
| 60/M||LGL/MDS-RA|| ||0.346||E, N||46xy||NR|| ||0103, 1301|
| 66/M||LGL/MDS-RAEB (oligo)|| ||0.98||E, P||46xy||NR||+||1302, 1502|
| 79/F||LGL/MDS-RA|| ||0.109||E, N, P||46xx||NR|| ||03, 08|
| 75/M||T-LGL LP|| ||4.455||E||46xy||NR||+||0801, 13|
Table IV. Analysis of response.
|HLA-DR4||9||1 (of 10)||0·01|
|Median LGL count (× 109/l)||0·846||0·370||< 0·03|
|Median ANC (× 109/l)||0·441||0·580||NS|
|BM lymphoid aggregates||7 (of 13)||6 (of 11)||NS|
|BM hypocellularity||4 (of 13)||4 (of 11)||NS|
Seven patients died with an overall mortality of 28% (95% CI 12·0–49·4%) over 4 years, a rate comparable to published accounts of mortality from T-LGL (including those without cytopenias) of between 10% and 20% at 4 years (Pandolfi et al, 1990; Dhodapkar et al, 1994). Overall, five out of 11 (45·5%) non-responders compared with two out of 14 (14·3%) responders died. Two partial responders died after stem cell allotransplant of progressive disease and sepsis respectively. Five deaths occurred in primary non-responders. One non-responder was lost to follow-up after 3 months. Median survival has not yet been reached at median follow-up of 46 months.
Responses to CSA are commonly attributed to the reversible inhibition of CD4+ and CD8+ T lymphocytes by blocking signalling downstream of the TCR (Kahan, 1989). CSA may also influence T-LGL by modulating p-glycoprotein (Foxwell et al, 1989; Yamamoto et al, 1993). Our ability to rescue three CSA-unresponsive patients by adding a brief course of erythropoietin is supportive of growth factors acting synergistically with CSA.
Opinion is divergent on the necessity of an immune stimulus and the identity of the antigenic trigger for the clonal CD8 expansion seen in T-LGL. It has been proposed that LGL represents a clonal expansion to a haematopoietic antigen (Barrett et al, 2000) or to human T-cell lymphotropic virus type 1 (HTLV-1; Loughran, et al, 1998) followed by clonal evolution to leukaemia (Langerak et al, 2001). Like activated CTLs, T-LGL cells express CD57, fas ligand (Perzova & Loughran, 1997) and perforin (Oshimi et al, 1990). The occurrence of the clone in both CD57– (memory) and CD57+ (effector) compartments is consistent with chronic antigenic stimulation by an autoantigen (Melenhorst et al, 2001). Our finding that HLA-DR4 predicts response to immunosuppression in T-LGL is further support for an antigen-specific immunological mechanism underlying the pancytopenia. HLA-DR15, a different class II allele, has been found to be over-represented and predictive of response to immunosuppression in myelodysplasia, paroxysmal nocturnal haemoglobinuria and aplastic anaemia, which are bone marrow failure states considered to have a T-cell-mediated autoimmune basis (Maciejewski et al, 2001; Saunthararajah et al, 2002). Significantly, HLA-DR15 was neither over-represented nor associated with CSA response in any subgroup in this cohort.
In conclusion, we have shown that CSA is a safe and effective treatment for the cytopenias of T-LGL. The addition of erythropoietin enhances the response. CSA compared favourably with antimetabolite therapy in response rate and has a good safety profile for long-term administration. The presence of HLA-DR4 predicts response to CSA and supports an underlying antigen-driven immunological mechanism in T-LGL.
The authors gratefully acknowledge contributions by Laura Wisch for clinical support, and Maria Bettinotti for helpful discussion regarding HLA association. This study was conducted and supported by the National Heart, Lung and Blood Institute, National Institutes of Health.