Cladribine treatment of multiple sclerosis is associated with depletion of memory B cells

The study was approved by the Health and Social Care Research Ethics Committee B and the Health Research Authority, UK. People were recruited following informed consent. Purchased blood samples were collected with informed consent and did not require additional ethical review.

Participants (Supplementary Table S1) were diagnosed with MS based on the revised McDonald criteria. Subjects provided demographic details and drug history at the time of informed consent and completed an online Extended Disability Status Scale (EDSS) (https://​edss.​clinicspeak.​com). Inclusion criteria are diagnosis of MS; aged 18–65 years and either on alemtuzumab or cladribine therapy or drug-naive (MS controls) undergoing investigations to start treatment. Healthy control volunteers also provided blood samples with informed consent. People were excluded if they could not give valid consent or comply with the study requirements.

PwMS were treated with 12 mg/day alemtuzumab for 5 days at month 1 and 12 mg/day for 3 days at month 13 [5]. Subcutaneous cladribine was used off-label on compassionate grounds. This consisted of 3 × 10 mg adjusted for weight (4 × 10 mg if > 90 kg) at month 0 and repeated 4 weeks later and further dose-adjusted from 0 to 4, 10 mg doses depending on subject weight and level of lymphopenia with > 1.0 × 10⁹ cells/L receiving three doses, 0.8 to  < 1.0 × 10⁹ cells/L receiving two doses; 0.5 to  < 0.8 × 10⁹ cells/L receiving one dose and those with < 0.5 × 10⁹ cells/L receiving no additional doses [15].

A cross-sectional study was performed on PwMS attending the Royal London Hospital, for routine blood monitoring as part of their standard of care. Samples were collected from healthy controls; drug-naïve PwMS and PwMS at their end of year blood test, prior to their second cycle of treatment.

The level of lymphopenia was assessed via standard haematology laboratory measures. One 1-year data was extracted from the pivotal CLARITY dataset of oral cladribine [4–5 daily dosing of 10–20 mg tablets (4–8 mg/day generic cladribine equivalent) week 1 and 5] of the group that were immunophenotyped (n = 309) supplied by the European Medicines Agency (EMA) under a Freedom of Information request [19]. The data from the 3.5 and 5.25 mg/kg groups [9] were merged as they had both received identical dosing up to 9 weeks. Analysis was restricted to data, where baseline 5- and 9-week data were all available. This was compared and contrasted with anonymous audits of people treated with subcutaneous, generic cladribine (3–4, 10 mg subcutaneous cladribine) doses at week 1 and week 5 or 5 daily 12-mg doses of alemtuzumab. The nadir at week 4 or 9 was selected for cladribine and the depletion at week 5 was analysed for alemtuzumab. The grade of lymphopenia was based on standardised definitions of the Common Terminology Criteria for adverse events V4, National Cancer Institute, Bethesda. Differences in the frequency of severe lymphopenic events were estimated using Chi-squared analysis.

Anonymised heparinised venous blood samples were collected and cells were labelled for 1 h with mouse anti-human antibody–fluorophore conjugates: CD3-PE (T cells), CD10-PECy7, CD19-PECy5, CD25-BV421, CD27-BV205, CD38-APC, IgD-BB515 and IgG1k-BV421 isotype (Becton-Dickinson, Oxford, UK) and cells were analysed by flow cytometry. Memory B cells were identified as CD19⁺/CD27⁺ lymphocytes. The presence of class-switched (CD19⁺/CD27⁺/IgD⁻), unswitched (CD19⁺/CD27⁺/IgD⁺) and interleukin-2 receptor expressing (CD19⁺/CD25⁺/CD27⁺) memory B cell subsets was assessed. The number of immature (CD10⁺/CD19⁺/CD27⁻/CD38⁺); naïve/mature (CD10⁻/CD19⁺/CD27⁻/CD38⁺) and plasmablasts (CD19⁺/CD27⁺/CD38⁺) were estimated. The absolute number of cell subsets was calculated based on automated full blood counts (Sysmex XE-2100, Kobe).

Anonymised healthy donor blood samples were purchased from Cambridge Bioscience UK and mononuclear cells extracted. These were incubated with various concentrations of cladribine with or without 250 μM deoxycytidine to neutralise DCK activity [20]. Cells were stained with CD3-PE, CD19-PECy5 and CD27-BV205, Annexin V (Invitrogen, Loughborough, UK) and 4′,6-diamidino-2-phenylindole (DAPI) to detect live (Annexin V−, DAPI−), early (Annexin V+ , DAPI−) and late stage (Annexin V+, DAPI+) apoptosis. Cells were analysed using flow cytometry.

Gene expression of T cell and notably B cell subsets of ADA, DCK and NT5-related genes were identified through searches in public databases: at the Human Protein Atlas for immunohistology (http://​www.​proteinatlas.​org, [21]) and Microarray and RNAseq data at BioGPS (http://​www.​biogps.​org, [22]) and the Gene Expression Omnibus at the National Center for Biotechnology Information, Bethesda, USA (https://​www.​ncbi.​nlm.​nih.​gov, GEO profiles/DATA sets).

Sample size calculations were based on data within the CARE-MS I alemtuzumab trial data set [18], with 80% power to detect an 80% memory B cell depletion, comparable with the 12-month alemtuzumab depletion data [18], at the P = 0.05 (n = 4/arm) and P = 0.01 (n = 8/arm) levels. Relative proportions of B populations and live/apoptotic cells were determined using FlowJo 8 (FlowJo, Ashland, USA) software. Analysis of variance with Bonferroni post hoc test or Kruskal–Wallis analysis of variance with a Dunn’s post hoc test was performed using Sigmaplot software.

Analysis of the in vitro effect of cladribine demonstrated that healthy human lymphocytes were dose-dependently killed by cladribine (Fig. 1a). This occurred by apoptosis and was blocked by excess deoxycytidine (Fig. 1b), supporting DCK as a major mediator of the cytotoxic effect. Memory B cells were sensitive to cladribine, similar to T cells (Fig. 1a), at concentrations achievable in plasma in humans [20]. However, as anticipated, without additional activating stimuli, B cells rapidly died in culture even in the absence of cladribine, limiting detection of any possible influence for different immune subsets. Therefore, we examined the in vivo response to cladribine in MS.

Our dose-adapted, subcutaneous cladribine induced a low (1.8%. n = 1/57) incidence of severe (grade 3/4) lymphopenia during the first cycle of treatment (Table 1). This was marginally lower than the 6.2% (n = 12/194) found in the immunophenotyping dataset of people taking oral cladribine (Table 1). Importantly the incidence of severe lymphopenia induced by subcutaneous cladribine was significantly (p < 0.0001) lower than the marked (84.1%, n = 106/126) lymphopenia induced by the first cycle of alemtuzumab treatment (Table 1). However, there was lymphocyte recovery towards the end of year 1 (Fig. 2).

Although serial analysis of B memory cell numbers following cladribine is needed and will be undertaken as a consequence of this study, initial analysis of cladribine-treated PwMS (n = 11) indicated that switched and unswitched memory B cell were consistently depleted to levels below reference levels [23] across a wide time-range (Supplementary Figure S1A). This suggests that once depleted, levels of B memory cells remain persistently low for over 12 months, as found in alemtuzumab- and rituximab-treated PwMS [18, 24]. Therefore, the memory B response was examined at a single point at the end of the treatment cycle in additional individuals (Figure S1B, Fig. 2a–d).

Healthy controls did not significantly differ from MS controls for any outcome measure assessed. However, there were significant treatment effects of both cladribine and alemtuzumab (Fig. 2a–e). There was depletion in the absolute numbers of CD3⁺ T cells in people treated with cladribine (45.0% depletion) and alemtuzumab (67.1%) at year 1 (Fig. 2c), which are similar with results of serial blood samples from the CLARITY [16] and CARE-MS I [18] studies. These pivotal phase III trials indicated that there was a marked depletion of CD19⁺ B cells by both alemtuzumab and cladribine [16, 18]. Consistent with the repopulation kinetics of alemtuzumab and rituximab [18, 24], there was an increase in the absolute numbers of immature and mature cells in PwMS treated with cladribine (Fig. 2c). These were comparable with the absolute B cell numbers in alemtuzumab-treated individuals (Fig. 2c). Most importantly, there was a marked (80.3%) and significant (P = 0.001) depletion of memory B cells by cladribine, compared to healthy controls, to levels at least as low to that seen in year 1 (75.6%) and year 2 (65.8%) samples post-alemtuzumab treatment (Fig. 2c).

Similarly there was a marked reduction (76.5 and 86.5% depletion compared to healthy controls) in the absolute numbers of both class-switched (P < 0.001) and unswitched (P = 0.003) memory B cells, respectively (Fig. 2d). This was comparable (67.8 and 84.9% depletion compared to healthy controls) to the class-switched and unswitched memory B cell depletion seen 1 year after the initial alemtuzumab treatment, respectively. Interestingly, there was also selective depletion of the CD25⁺ subset of memory B cells (Fig. 2d). These cells represented 63.3 ± 4.3% of the CD19⁺, CD27⁺ pool in healthy controls, but only 34.1 ± 2.0% of the memory B cell pool in PwMS treated with cladribine and 27.0 ± 3.6% (year 1) and 32.4 ± 4.4% (year 2) in PwMS treated with alemtuzumab, indicating a marked depletion in this subset by cladribine (88.9% depletion compared to healthy controls) and alemtuzumab (88.6% depletion at year 1, Fig. 2d).

To aid comparison with other published studies [4, 19], the percentage of total B memory cells within the CD19⁺ population (Figure S1B and 2E) was also analysed. Both class-switched and unswitched memory B cells were significantly (P < 0.05) reduced following cladribine treatment to a level comparable with that found following alemtuzumab treatment at year 1 and year 2 (Fig. 2d).

Gene and protein expression data within public repositories were searched to determine whether differences in purine salvage pathway genes could help explain selectivity of the action of cladribine for B cells. It was evident that ADA and DCK are relatively specific for leucocytes in contrast to nucleoside phosphorylases that have broad expression [13]. This was readily observed following examination of microarray data (Fig. 3a, b). It was found that the tissue distribution of ADA message correlated well with the previously reported [13] protein activity (Fig. 3a). Furthermore, although there was variation in lymphocyte expression levels between different microarray studies, it was evident that B cells often express lower levels of ADA than T cells (Fig. 3a, b, E-GEOD-22886, GSE62584 from blood during first demyelinating event) and importantly B cells may, but not always (E-GEOD-22886, GSE62584), express higher levels of DCK than T cells (Figs. 3a, b, 4). This is consistent with observations measuring protein or functional activity of the enzymes within normal cells and malignant cells, where B lineage cells tend to exhibit higher activity than T lineage cells [25]. However, it was evident that B cell subsets are very heterogeneous with regard to expression (Fig. 3b). Whilst there was variation between different microarray studies (GPS_00013; E-GEOD-22886; GSE68878; GSE68245; GSE68878) on balance it was found that immature, mature and memory populations, which populate the blood compartment, had similar levels of DCK (Fig. 3b). These expressed low levels of ADA (Fig. 3b). However, it was consistently found (GPR_00013; GSE68878; E-GEOD-22886) that plasma cells in blood, tonsil and bone marrow (Fig. 3b) exhibited significantly lower levels of DCK compared to memory and germinal centre cells. Interestingly, it was evident that germinal centre cells and notably lymphoblasts, which localise to the dark zone of the germinal centre exhibit high levels of DCK (Fig. 3b, E-GEOD-38697; E-GEOD-15271). This profile was consistent with protein expression within human lymphoid tissue (Fig. 4). Indeed B cells within the follicles express more staining than cells within the paracortical areas, which contain T cells (Fig. 4a–d). Importantly there was high expression of DCK within the dark zone of the secondary follicles (Fig. 4a–d). Within the light zone there were intensely stained, modestly stained and poorly stained cells, which is perhaps consistent with levels of DCK message in centrocytes, memory cells and plasma cells (Fig. 3b) that reside in these areas.

Although DCK is expressed at high levels within B cells, some 5′NT subtypes, notably NT5C3 isoforms are expressed at high levels within B cells (Fig. 3c). This could counter the action of DCK. Likewise, NT5E was expressed at higher levels in the circulatory: immature, mature and memory subtypes than T cells (Fig. 3c), which is consistent with reports on protein expression [25]. Whilst NT5E will dephosphorylate deoxyadenosine monophosphates, as it is an ecto-enzyme, it may not be particularly relevant to control the cytosolic action of DCK [26]. Likewise, as cytosolic NT5C2 preferentially targets inosines and NT5C3 isoforms are involved in pyrimidine catabolism [26], their relevance to dephosphorylation of 2-deoxycladribine monophosphate remains to be properly identified. However, based on the ratio of DCK activity to total 5′NT activity, the germinal centre cells would appear to be a potentially highly sensitive to cladribine (Fig. 3d). This aspect is particularly notable, when the balance between DCK and the cytosolic adenosine monophosphate degradative NT5C1 isoforms [26] was examined (Fig. 3d). These are expressed at low levels (Figs. 3c, 4e). There was a significantly higher DCK:NT5C1 ratio in CD4 T cells than CD8 T cells and higher DCK:NT5C1 ratio between mature and memory B cells and memory CD4⁺ and CD8⁺ T cells (Fig. 3d). In contrast neutrophils expressed little DCK message and more NT5C1B (Fig. 3d). This may help explain the pattern of neutrophil, T cell and B cell depletion profiles observed following use of cladribine.