Approach to the patient after relapse of hairy cell leukemia

Abstract

The current hairy cell leukemia (HCL) treatment is excellent, but evidence of cure with purine analogs cladribine and pentostatin, is lacking. Significant long-term immune suppression, particularly to CD4 + T-cells, and declining complete remission rates with each course, make the identification of new therapies an important goal. The anti-CD20 monoclonal antibody (Mab) rituximab displays significant activity, and, while causing prolonged normal B-cell depletion, spares T-cells. Recombinant immunotoxins, containing an Fv fragment of a Mab fused to truncated Pseudomonas exotoxin, have shown efficacy in HCL resistant to both purine analogs and rituximab. LMB-2 targets CD25 and can induce remission providing the HCL cells are CD25+. All HCL cells display CD22. Recombinant anti-CD22 immunotoxin BL22, targeting CD22, has shown significant efficacy in phase I and II testing, and avoids prolonged suppression of both normal B- and T-cells. An improved high-affinity version of BL22, termed HA22, is currently undergoing phase I testing.

Need for additional therapies

The results of therapy with purine analogs cladribine and pentostatin for hairy cell leukemia (HCL) are excellent, with 85–95% of patients achieving complete remission (CR), only about 40% of patients relapsing by 10 years [1–3], and 75% achieving second CR [1]. However, disease-free survival curves fail to show a plateau even after 10 years, and hence there is no evidence of cure. Moreover, while third and fourth CRs are common with repeated courses of purine analogs, the CR rates decline significantly with each successive course, irrespective of whether the same purine analog is used or not [4,5]. Because a single course of cladribine or pentostatin is reported to suppress CD4+ lymphocytes below the lower limit of normal for a median of 40 or 54 months, respectively [6,7], it may be unsafe to use repeated courses of purine analogs to maintain HCL patients, particularly at short intervals. The use of rituximab, while not approved for HCL, is an important advance because this anti-CD20 monoclonal antibody (MAb) spares T-lymphocytes. CR rates among the six reported rituximab studies in HCL (10–25 patients each, total 97) vary from 10 to 54% [8–13]. However, in the 51 patients from five studies who demonstrated a need for treatment based on cytopenias and who had at least one prior purine analog, there were 10 (20%) CRs and 10 (20%) partial responses (PRs) [8–12]. In the largest single trial enrolling 24 such patients, there were 3 (13%) CRs and 3 (13%) PRs [10]. Thus, new treatments are needed for relapsed HCL which have both high efficacy and lack of cumulative toxicity, particularly to T-cells. Recombinant immunotoxins are currently being developed to meet this need.

Recombinant immunotoxins

Protein toxins are among the most potent natural substances known, in that they act catalytically and can therefore kill a cell with a single molecule in its cytoplasm [14]. Plant toxins inactivate ribosomes by preventing their association with elongation factor-1 and −2 (EF-1 and EF-2). Bacterial toxins such as Pseudomonas exotoxin (PE) and diphtheria toxin (DT) induce ADP-ribosylation of EF-2. This leads to protein synthesis inhibition in either case, and cell death by apoptosis [15,16]. Bacterial toxins, which are more often used to fuse to ligands, are naturally made by bacteria in single-chain form, composed of domains for binding and ADP-ribosylation at opposite ends, and a translocation domain in between [16–19]. The orientation of the domains are opposite in PE and DT, with the binding domain at the amino terminus of PE and at the carboxyl terminus of DT. They intoxicate cells by binding to the cell surface, undergoing internalization, unfolding within an acidic vesicle, undergoing proteolytic cleavage within the translocating domain, and translocating to the cytosol where EF-2 is inactivated by ADP-ribosylation [19]. Recombinant toxins are produced by replacing the binding domain with a cancer cell-binding ligand. In denileukin diftitox, approved for relapsed and refractory cutaneous T-cell lymphoma, human interleukin-2 replaces the binding domain of DT at the carboxyl terminus [20–22]. In recombinant immunotoxins, the cell-binding ligand is an Fv fragment, a recombinant antibody containing the variable domains of a Mab [23]. In the recombinant immunotoxin LMB-2, the variable domains of the anti-CD25 Mab are fused together by a peptide linker and then fused to the amino terminus of PE38, a fragment of PE which is missing its binding domain [19,24]. In the recombinant immunotoxin BL22, the variable domains of the MAb RFB4 are fused to PE38, but the Fv is stabilized with an engineered disulfide bond rather than a peptide linker [25–27].

LMB-2 targeting CD25

LMB-2 was the first recombinant immunotoxin reported to be of benefit for HCL, with four (100%) responses out of four patients treated as part of a phase I trial, one of whom had a durable CR [28,29]. The three patients with PR had inadequate treatment due to pneumonia, dose-limiting toxicity and immunogenicity, suggesting that significant clinical activity might be observed in a phase II setting. Unlike purine analogs, myelosuppression and severe lymphopenia were not observed with LMB-2 [28,29], which was expected since normal T-and B-cells have insufficient CD25 expression to be sensitive [30,31]. The most common toxicities of LMB-2 included non-dose limiting hypoalbuminemia, transaminase and alkaline phosphatase elevations, and fever. A disadvantage of targeting CD25 in HCL is that 10% of patients have a variant (HCLv) in which CD25 is absent [32,33], and this variant is over-represented by patients with relapsed/refractory disease after purine analogs. Both HCLv and classic HCL uniformly express high levels of CD22, and have more recently been targeted by the recombinant immunotoxins BL22.

BL22 targeting CD22

In phase I testing, 31 of the 46 patients had HCL relapsed or refractory after purine analogs, and of these 31 patients there were 19 (61%) CRs and 6 (19%) PRs [34,35]. A completely reversible form of hemolytic uremic syndrome (HUS) was observed in four (13%) of the patients, but only during retreatment with a second or third cycle. However, most (58%) of the CRs were achieved with one cycle, suggesting that retreatment was not always necessary or advisable. Therefore, in the phase II trial, patients were observed without retreatment if they achieved CR or at least the complete blood count (CBC) values required for CR, and if not, they were retreated at a lower dose level. One cycle of BL22 induced CR in 25% of 36 patients, and by retreating the 56% of patients who were eligible for retreatment, the CR rate rose to 47%, with an overall response rate of 72% [36]. Compared to the phase I trial, the dose-limiting HUS rate was less than half (6% vs. 13%) and the immunogenicity rate less than one-third (11% vs. 35%). The median CR duration was 5–46+ (median 22+) months, with 76% of 17 CRs ongoing [36]. Response was not significantly related to length of prior purine analog response, but was significantly higher in patients with spleens <200 mm in maximal dimension than in patients with either larger spleens or prior splenectomy. Pharmacokinetic studies showed significant inverse relationships between area under the curve and tumor burden measured by either the log of the HCL count in the blood, or the spleen size. It was concluded that BL22 is highly active in HCL despite prior purine analog treatment and resistance, and that experimental therapy should be considered in patients prior to removing the spleen or allowing it to become massive in size.

HA22, a high affinity mutant of BL22

Patients with chronic lymphocytic leukemia (CLL) were not as sensitive to BL22, attributed to much lower expression of CD22 on CLL compared to HCL cells [35]. To allow a higher percentage of bound immunotoxin to be internalized by CD22 rather than disassociating from this antigen, the off-rate of BL22 was lowered by mutagenesis within the third complementarity determining region of the heavy chain. The resulting anti-CD22 recombinant immunotoxin HA22 contained three amino acid mutations and bound with 15-fold higher affinity and lower off-rate to CD22, and had improved cytotoxicity [37–39]. HA22 is undergoing phase I testing in HCL, CLL and non-Hodgkin lymphoma. The HCL trial has neared completion with excellent results, and a pivotal trial is being planned in this disease.

Approach to patients with relapsed hairy cell leukemia

Many different treatment approaches to the relapsed patient exist, and lack of randomized data prevents discounting most of them as inappropriate. A suggested algorithm for the treatment of early and relapsed HCL is shown in Table I. Because of the long-term immunosuppression of purine analogs and their decreased efficacy with retreatment, a general strategy has been to avoid multiple courses of these agents with short time intervals in between. After first relapse with first purine analog response duration less than 1–2 year, or after multiple relapses from purine analog, salvage treatment with rituximab or experimental treatment with LMB-2, BL22 or HA22 (CAT-8015) has shown positive outcomes [28,29, 35,36,40,41,59,60]. Combined treatment with both purine analog and rituximab has been successfully employed in both the 1st line and relapsed setting [5,61–64], although this approach is not standard and long-term suppression of both normal B- and T-cells is expected. Physicians with patients who are eligible for recombinant immunotoxin clinical trials may choose to consider HA22 (CAT-8012) after failure of at least 2 courses of purine analog. Patients with <12 month response to initial purine analog are considered primarily refractory and not appropriate for repeat courses of purine analog, thus their physicians may consider rituximab as 2nd line treatment and HA22 (CAT-8015) thereafter. Interferon is clearly inferior to purine analog for achievement of response [45], and has significant toxicities, but may offer palliation [55]. Splenectomy should be avoided in patients being considered for recombinant immunotoxin experimental treatment [36], but should always be considered in patients with even mild or moderate splenomegaly who have exhausted other options and risk fatal complications of the disease [55,56]. Finally, purine analogs alone or with other agents may be employed in the palliative setting, but their declining potential benefit with repeated use must be considered along with their cumulative and often non-reversible toxicities.

Table I.

HCL-specific open clinical trials.

Agents	Phase	Line	Center	Cancer.gov NCT#	Reference
Cladribine + rituximab	Phase II non-randomized	1st/2nd	MDA	NCT00412594	[44]
Cladribine + rituximab	Phase II randomized	1st/2nd	NCI	NCT00781235
LMB-2	Phase II	> 2nd	NCI	NCT00337311	[28,29]
BL22	Phase II retreatment	>3rd	NCI	NCT00850525	[34–36]
HA22 (CAT-8015)	Phase I	> 2nd	NCI, NW Stanford Lodz	NCT00462189	ASH, 2008, #3160

HCL, hairy cell leukemia.

For the two cladribine + rituximab trials, patients begin treatment at MD Anderson (Houston, TX) or at the National Cancer Institute, NIH (Bethesda, MD), respectively, and subsequent treatment may be given by local physicians.

HA22, also called CAT-8015, is administered at NCI, Northwestern (Chicago, IL), Stanford University (Stanford, CA) or at the University of Lodz (Lodz, Poland).

Prevention of relapse by combining rituximab with cladribine

Early treatment with rituximab and cladribine. One hypothesis is that relapse in HCL may be prevented or delayed by eliminating the minimal residual disease (MRD) left over after CR to cladribine. A preliminary report shows excellent activity of rituximab in eliminating MRD when used 4 weeks after cladribine in 13 patients [61]. This trial was not designed to determine whether rituximab should be used during 1st or 2nd line cladribine because comparison with cladribine alone was not done. To determine the benefit of adding rituximab to first or 2nd line cladribine, a randomized study is underway at the National Cancer Institute which also allows patients to be treated by their local physicians. Patients with 0 or 1 prior course of cladribine are randomized separately. Within each of these stratifications, half of the patients receive rituximab beginning with the 1st dose of cladribine, half receive cladribine without rituximab, and all patients receive delayed rituximab at least 6 months later if MRD can be demonstrated in the blood. Thus the 1st goal is to determine if rituximab can decrease the rate of MRD observed in the blood at 6 months, the time point at which patients generally achieve their minimum disease burden after cladribine [65,66]. The 2nd goal is to determine if MRD-free survival is better if rituximab is used up front or delayed. It is quite possible that rituximab may work better if delayed at least 6 months after cladribine when disease is minimal and clumps of cells are more permeable to rituximab. Alternatively, rituximab may work synergistically with cladribine and have optimal activity up front. Regardless of when the rituximab is added in combination with cladribine, we believe that patients with early HCL may benefit from combined treatment and should be referred for the randomized and non-randomized trials of cladribine and rituximab if possible. Because of the cumulative and long term toxicity or purine analogs and the lack of long-term toxicity with recombinant immunotoxins, we believe that patients with multiply relapsed HCL should avoid repeated courses of purine analog and if possible enroll in a clinical trial testing recombinant immunotoxin HA22 (CAT-8015).

Figure 1.

Flow-chart for the suggested treatment of early and relapsed HCL.