Anti-CD22 and Anti-CD79B Antibody Drug Conjugates Are Active in Different Molecular Diffuse Large B-Cell Lymphoma Subtypes

Home , CD79, CD79B, Monomethyl auristatin E, Pinatuzumab vedotin, Polatuzumab vedotin

ORIGINAL ARTICLE Anti-CD22 and anti-CD79B antibody drug conjugates are active in different molecular diffuse large B-cell lymphoma subtypes

M Pfeifer1, B Zheng2, T Erdmann3,4, H Koeppen2, R McCord2, M Grau5, A Staiger6, A Chai2, T Sandmann2, H Madle3,4, B Dörken1, Y-W Chu2, AI Chen7, D Lebovic8, GA Salles9, MS Czuczman10, MC Palanca-Wessels11,12, OW Press11, R Advani13, F Morschhauser14, BD Cheson15, P Lenz5,GOtt6, AG Polson2, KE Mundt2 and G Lenz3,4

Antibody drug conjugates (ADCs), in which cytotoxic drugs are linked to antibodies targeting antigens on tumor cells, represent promising novel agents for the treatment of malignant lymphomas. Pinatuzumab vedotin is an anti-CD22 ADC and polatuzumab vedotin an anti-CD79B ADC that are both linked to the microtubule-disrupting agent monomethyl auristatin E (MMAE). In the present study, we analyzed the activity of these agents in different molecular subtypes of diffuse large B-cell lymphoma (DLBCL) both in vitro and in early clinical trials. Both anti-CD22-MMAE and anti-CD79B-MMAE were highly active and induced cell death in the vast majority of activated B-cell-like (ABC) and germinal center B-cell-like (GCB) DLBCL cell lines. Similarly, both agents induced cytotoxicity in models with and without mutations in the signaling molecule CD79B. In line with these observations, relapsed and refractory DLBCL patients of both subtypes responded to these agents. Importantly, a strong correlation between CD22 and CD79B expression in vitro and in vivo was not detectable, indicating that patients should not be excluded from anti-CD22-MMAE or anti- CD79B-MMAE treatment because of low target expression. In summary, these studies suggest that pinatuzumab vedotin and polatuzumab vedotin are active agents for the treatment of patients with different subtypes of DLBCL.

Leukemia (2015) 29, 1578–1586; doi:10.1038/leu.2015.48

INTRODUCTION monoclonal antibody (mAB) rituximab into the therapy of DLBCL – Diffuse large B-cell lymphoma (DLBCL) represents a hetero- has significantly improved prognosis.10 12 Nevertheless, a sub- geneous diagnostic category.1 This heterogeneity can partially stantial number of patients relapse after first-line therapy. These be explained by the existence of different molecular subtypes relapsed or refractory DLBCL patients are characterized by dismal identified by gene expression profiling.2–5 Applying the cell of outcome.13 origin classification, at least two major molecular subtypes can be A novel therapeutic approach is the use of antibody drug distinguished. The germinal center B-cell-like (GCB) DLBCLs are conjugates (ADCs) in which cytotoxic drugs are attached to derived from germinal center B cells, whereas activated B-cell-like antibodies that are directed against antigens expressed on tumor (ABC) DLBCLs originate from activated B cells.2,3 GCB and ABC cells.14 Recently, this strategy has been shown to be very effective DLBCLs depend on different oncogenic pathways that are in the treatment of malignant lymphoma patients. The anti-CD30 frequently activated by somatic mutations.1 Importantly, the ADC brentuximab vedotin achieved high response rates in existence of molecular subtypes as well as specific mutations patients with relapsed or refractory CD30-positive Hodgkin’s – appear predictive of response to pathway inhibitors such as lymphoma and anaplastic large-cell lymphoma (ALCL).15 17 ibrutinib, and immunomodulatory agents such as lenalidomide or CD22 and CD79B are physiologically expressed in the vast proteasome inhibitors.6–8 Thus, the assignment of DLBCL patients majority of B cells and therefore represent promising targets for into molecular subtypes is becoming increasingly important to ADCs. Preclinical data for an anti-CD79B and an anti-CD22 antibody ensure that patients are selected for the most efficacious conjugated to the microtubule-disrupting agent monomethyl treatment regimens. Finally, ABC and GCB DLBCLs also show auristatin E (MMAE) indicated the efficacy in B-cell lymphomas significant differences in outcome when treated with including Burkitt lymphoma, mantle cell lymphoma, follicular immunochemotherapy.9 The introduction of the anti-CD20 lymphoma and DLBCL.18–20 However, these studies did not

1Department of Hematology, Oncology and Tumor Immunology, Charité—Universitätsmedizin Berlin, Germany; 2Genentech Inc., 1 DNA Way, South San Francisco, CA, USA; 3Division of Translational Oncology, Department of Medicine A, University Hospital Münster, Münster, Germany; 4Cluster of Excellence EXC 1003, Cells in Motion Münster, Germany; 5Department of Physics, Philipps-University, Marburg, Germany; 6Department of Clinical Pathology, Robert-Bosch-Krankenhaus and Dr. Margarete Fischer-Bosch- Institute of Clinical Pharmacology, Stuttgart, Germany; 7Department of Hematology-Oncology, Knight Cancer Institute, Oregon Health and Science University, Portland, OR, USA; 8Department of Internal Medicine, Division of Hematology and Oncology, University of Michigan, Ann Arbor, MI, USA; 9Hematology Department, Hospices Civils de Lyon - Université de Lyon, Pierre-Bénite, France; 10Department of Medicine and Immunology, Roswell Park Cancer Institute, Buffalo, NY, USA; 11Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, WA, USA; 12Seattle Genetics Inc, Bothell, WA, USA; 13Stanford University Medical Center, Stanford University, Stanford, CA, USA; 14Department of Hematology, Centre Hospitalier Régional Universitaire de Lille, Lille, France and 15Lombardi Comprehensive Cancer Center, Georgetown University Hospital, Washington DC, USA. Correspondence: Dr KE Mundt, Genentech Inc., DNA Way 1, South San Francisco 94080, CA, USA or Professor G Lenz, Division of Translational Oncology, Department of Medicine A, University Hospital Münster, Albert-Schweitzer-Campus 1, Münster 48149, Germany. E-mail: [email protected] or [email protected] Received 25 August 2014; revised 8 January 2015; accepted 13 February 2015; accepted article preview online 24 February 2015; advance online publication, 8 May 2015 Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1579 investigate whether ABC or GCB DLBCL or lymphomas character- The anti-CD22-mAB SP104 (Ventana) was used according to the ized by specific mutations respond differentially to these agents. manufacturer’s instructions. The anti-CD79B-mAB AT-107-2 (0.25 μg/ml; Of particular interest in this respect are mutations affecting CD79B AbD Serotec, Oxford, UK) was incubated for 60 min at 37 oC. Human tonsils that have been reported to have an impact on B-cell receptor (BCR) were used as positive controls. The staining results are reported as H-score 21 as previously described that includes both numbers of positive tumor cells internalizationinmouseBcells. To this end, we analyzed the 28,29 efficacy of anti-CD22-MMAE and anti-CD79B-MMAE in a large panel as well as the staining intensities. The H-score was calculated for staining of tumor cells using the following formula: of cell lines derived from DLBCL patients of both molecular H-Score = (% at 0) × 0+(% at 1+) × 1+(% at 2+) × 2+(% at 3+) × 3. Thus, subtypes including cell lines characterized by CD79B mutations. To this score produces a continuous variable that ranges from 0 to 300.28,29 validate our in vitro findings, we investigated responses observed in A modified H-score was used in the cell lines that displayed a rather DLBCL patients treated in two recent phase-I studies. uniform intensity of protein expression in a given case: % cells positive × predominant expression intensity (0–3). MATERIALS AND METHODS Determination of molecular DLBCL subtypes Antibody drug conjugates Classification into molecular DLBCL subtypes was performed analogously The following ADCs were used in the present study. As previously to Wright et al.3 The detailed algorithm is described in Supplementary described, pinatuzumab vedotin is an anti-CD22-MMAE ADC (DCDT2980S), Materials and Methods and the data are summarized in Supplementary whereas polatuzumab vedotin (DCDS4501A) represents an anti-CD79B- Figure 1. MMAE ADC.18,22 Pinatuzumab vedotin and polatuzumab vedotin are the drug candidates used in two phase-I clinical trials in relapsed and refractory lymphoma patients.23,24 Determination of IC50 concentrations and correlation analyses The IC50 concentrations were calculated as described in detail in the Supplementary Materials and Methods. Cell lines Human DLBCL, ALCL and multiple myeloma (MM) cell lines were cultured as previously described in RPMI (Life Technologies, Carlsbad, CA, USA) with Mutation analysis of CD79B in primary DLBCL patient samples 10% fetal calf serum (Sigma-Aldrich, St Louis, MO, USA), except for OCI-Ly1, Detailed protocols are available in the Supplementary Materials and OCI-Ly2, OCI-Ly3, OCI-Ly4, OCI-Ly7, OCI-Ly10, OCI-Ly19 and TMD8 that Methods. The primer sequences are summarized in Supplementary Table 1. were cultured in Isocove’s modified Dulbecco medium (Life Technologies) 25,26 supplemented with 10% fetal calf serum. Western blotting Western blot analysis was performed as previously described.30,31 Viability assay, sub-G1 and cell cycle analysis Detailed protocols are available in the Supplementary Materials and Methods. RESULTS CD22 and CD79B expression in DLBCL Determination of CD22 and CD79B surface expression To assess the expression pattern of the targets of anti-CD22-MMAE using flow cytometry and anti-CD79B-MMAE, we measured CD22 and CD79B cell Detailed protocols are available in the Supplementary Materials and surface expression of DLBCL-derived cell lines using flow Methods. cytometry. Cell lines derived from MM and ALCL that do not express either CD22 or CD79B were used as negative controls CD79B antibody internalization assay (Supplementary Figures 2a and b). As expected, virtually all DLBCL cell lines displayed high CD22 and CD79B surface expression Cells were pre-incubated with 10 μg/ml of mouse anti-human CD79B − 7 − 6 antibody (clone SN8)27 on ice for 30 min and washed once to remove excess compared with MM (P = 2.2 × 10 for CD22 and P = 8.6 × 10 for CD79B) and ALCL cell lines (P = 2.2 × 10 − 7 for CD22 and unbound antibodies. Cells were subsequently incubated at 37 °C allowing for − 5 internalization. To terminate internalization, cells were washed twice and P = 1.3 × 10 for CD79B), albeit in a dynamic range stained with phycoerythrin-conjugated goat anti-mouse IgG antibody. (Supplementary Figures 2a and b and Supplementary Table 2). Surface CD79B expression was measured using flow cytometry. Percent of To analyze CD22 and CD79B cell surface expression in primary internalization was calculated as % internalization (Tn) = (gMFI(T0) − gMFI DLBCL, we performed immunohistochemical staining on 32 fl (Tn))/gMFI(T0) × 100 with gMFI = geometric mean uorescence intensity. patient samples for CD22 and on 28 patient samples for CD79B obtained from patients treated in two recently completed phase-I Patient population trials investigating the efficacy of anti-CD22-MMAE and anti- Both DCT4862g and DCS4968g were phase-I dose-finding studies for anti- CD79B-MMAE, respectively (NCT01209130 and NCT01290549). All CD22-MMAE or anti-CD79B-MMAE, respectively, in patients diagnosed with specimens expressed CD79B, consistent with its role as a signal- relapsed or refractory B-cell lymphoma. These studies are registered at transducing subunit of the essential BCR (median H-score 250; ClinicalTrials.gov under NCT01209130 and NCT01290549, respectively.23,24 Figures 1a, b and e and Supplementary Table 3). In contrast, the All patients gave informed written consent to participate in the clinical expression levels of CD22 determined with immunohistochemistry trials and to provide tissue for biologic studies. The respective institutional were more variable and unexpectedly low in several specimens ethics review boards approved both trials. To correlate preclinical data with (median H-score 90; Figures 1c–e and Supplementary Table 4). clinical results, we report in this manuscript CD22 and CD79B expression These results are surprising, especially as previous assessment of fi data as well as the correlation of expression and ef cacy in DLBCL patients CD22 surface expression in primary DLBCL samples using flow who were treated at doses of 1.8 mg/kg and above. The recommended cytometry indicated CD22 expression on virtually all cases.20 phase-II dose was determined as 2.4 mg/kg for both CD22 ADC and CD79B ADC as single agents. The complete patient characteristics and overall results of these studies will be published in a separate manuscript. Activity of anti-CD22-MMAE and anti-CD79B-MMAE in different subtypes of DLBCL in vitro and in vivo CD22 and CD79B immunohistochemistry in primary DLBCL patient Anti-CD22 and anti-CD79B ADCs have shown promising activity in samples and DLBCL cell lines preclinical models of various subtypes of malignant lymphoma.18 CD22 and CD79B immunohistochemical stainings were performed on the To specifically investigate the effects of anti-CD22-MMAE and anti- Ventana Discovery XT platform (Ventana, Tucson, AZ, USA) using CD79B-MMAE in different molecular DLBCL subtypes, we assessed the Ventana CC1 standard digestion pretreatment for antigen retrieval. cell viability following treatment with these agents in a large panel

© 2015 Macmillan Publishers Limited Leukemia (2015) 1578 – 1586 Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1580 of 27 DLBCL cell lines, previously assigned to either ABC or GCB DLBCL (Figures 2a and b and Supplementary Table 5). Using this DLBCL (7 ABC DLBCL and 20 GCB DLBCL cell lines). We analyzed cutoff, we observed a similar pattern of responsiveness of DLBCL dose-dependent sensitivity of every cell line in the range of cell lines to both ADCs. In all, 22 out of 27 (81%) cell lines were 0–50 μg/ml ADC and calculated the corresponding IC50.To sensitive to anti-CD22-MMAE and 22 out of 27 (81%) cell lines differentiate responding from non-responding cell lines, we showed toxicity following the anti-CD79B-MMAE treatment. The employed a cutoff value of IC50 ⩽ 5 μg/ml for responders, on the vast majority of DLBCL cell lines responded to both anti-CD22- basis of the clinical exposure at trough drug levels being above MMAE and anti-CD79B-MMAE, respectively (Figures 2a and b and 5 μg/ml for the recommended phase-II dose of 2.4 mg/kg for Supplementary Table 5). In contrast, a minority of cell lines was

CD22/CD79B expression in primary DLBCL patient samples

300

200

100

Staining intensity (H-score) 0

CD22 CD79B Figure 1. CD22 and CD79B expression in DLBCL. (a–d) Immunohistochemical staining of (a) a CD79B-positive DLBCL case with strong CD79B expression (H-score 300; case: DLBCL_63), (b) a CD79B-positive DLBCL case with weak CD79B expression (H-score = 120; case: DLBCL_69), (c)a CD22-positive DLBCL case with strong CD22 expression (H-score = 300, case: DLBCL_41) and (d) a CD22-positive DLBCL case with weak expression (H-score = 130; case: DLBCL_32). Magniﬁcation: × 20 (a) or × 40 (c–d). (e) CD22 and CD79B expression in primary DLBCL patient samples. CD22 and CD79B expression was determined by immunohistochemistry and reported as H-score. The black line indicates the mean expression of the respective surface marker.

Figure 2. Anti-CD22-MMAE and anti-CD79B-MMAE are active agents for the treatment of DLBCL. (a) Anti-CD22-MMAE displays potent in vitro activity across ABC and GCB DLBCL cell lines. The dashed red line indicates an IC50 value of 5 μg/ml that was used as the cutoff level to differentiate responding from non-responding cell lines. (b) Anti-CD79B-MMAE displays potent in vitro activity across ABC and GCB DLBCL cell lines. The dashed red line indicates an IC50 value of 5 μg/ml that was used as the cutoff level to differentiate responding from non-responding cell lines. (c) Determination of anti-CD79B antibody internalization rate in CD79B-mutated (TMD8) and wild-type (WSU-DLCL2 and SUDHL-4) DLBCL cell lines. The majority of the anti-CD79B antibody is rapidly internalized in all three cell lines within the first hour of incubation. The internalization rate in TMD8 and WSU-DLCL2 cells is virtually identical and higher compared with the rate in SUDHL-4 cells. (d) Treatment with anti-CD22-MMAE induces a significantly higher rate of cell death compared with the anti-CD22-mAB in the DLBCL cell lines OCI-Ly7 and RL as measured by an increase in the sub-G1 fraction. Error bars depict the s.d. (e) Treatment with anti-CD79B-MMAE induces a significantly higher rate of cell death compared with the anti-CD79B-mAB in the DLBCL cell lines OCI-Ly19 and RL as measured by an increase in the sub-G1 fraction. Error bars depict the s.d. *Po0.05, **Po0.01, ***Po0.001.

ABC DLBCL GCB DLBCL 10 9 8 7 Anti-CD22-MMAE 6

IC50 5 (µg/ml) 4 3 2 1 0 RL HT DLBCL cell lines DB RIVA K422 BJAB HBL1 TMD8 U2932 HS445 Farage Pfeiffer OCI-Ly1 OCI-Ly4 OCI-Ly7 OCI-Ly3 OCI-Ly2 NUDUL-1 OCI-Ly10 OCI-Ly19 SUDHL-4 SUDHL-2 SUDHL-6 WSU-NHL SUDHL-16 SUDHL-10 WSU-DLCL2

ABC DLBCL GCB DLBCL 10 9 8 7 Anti-C79B-MMAE 6

IC50 5 (µg/ml) 4 3 2 1 0 RL HT DLBCL cell lines DB RIVA K422 BJAB HBL1 TMD8 U2932 HS445 Farage Pfeiffer OCI-Ly3 OCI-Ly7 OCI-Ly1 OCI-Ly2 OCI-Ly4 NUDUL-1 OCI-Ly10 OCI-Ly19 SUDHL-2 SUDHL-6 SUDHL-4 WSU-NHL SUDHL-10 SUDHL-16 WSU-DLCL2

120 100 Cell DLBCL CD79B 80 line subtype mutation Internalization % status 60 TMD8 ABC Y196H 40 SUDHL-4 GCB WT 20 WSU-DLCL2 GCB WT 0 041 2 3 Hours

OCI-Ly7 RL 50 50 40 **40 *** Dead cells 30 30 (% of total) Dead cells (% of total) 20 20

10 10

0 0 Anti- Anti- Anti- Anti- CD22-mAB CD22-MMAE CD22-mAB CD22-MMAE

OCI-Ly19 RL 50 50 40 **40 30 30 Dead cells Dead cells (% of total) (% of total) 20 20

10 10

0 0 Anti- Anti- Anti- Anti- CD79B-mAB CD79B-MMAE CD79B-mAB CD79B-MMAE

© 2015 Macmillan Publishers Limited Leukemia (2015) 1578 – 1586 Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1582 sensitive to either anti-CD22-MMAE or anti-CD79B-MMAE treat- The in vivo patient data confirmed our in vitro observations that ment. Only two cell lines (SUDHL-2 and WSU-NHL) were resistant both anti-CD22-MMAE and anti-CD79B-MMAE are active agents to both drugs. For both agents, we could not detect a preferential in DLBCL. Treatment with anti-CD22-MMAE induced either a response in either ABC or GCB DLBCL cell lines (P = 0.10 for anti- complete (CR) or a partial remission in 12 out of 28 (43%) patients CD22-MMAE and P = 0.47 for anti-CD79B-MMAE; one-tailed two at dose levels ⩾ 1.8 mg/kg (Table 1). As suggested from our cell sample t-tests). These results demonstrate that anti-CD22-MMAE line data, we observed activity in both ABC and GCB DLBCL and anti-CD79B-MMAE are active in DLBCL models irrespective of patients. Three out of seven ABC (two CRs) and two out of seven their molecular subtype. GCB DLBCL patients (two CRs) responded to anti-CD22-MMAE To validate whether indeed both primary ABC and GCB DLBCL (dose level ⩾ 1.8 mg/kg; Table 2). Anti-CD79B-MMAE was similarly patients respond equally to anti-CD22-MMAE and anti-CD79- effective in DLBCL patients of both molecular subtypes. Responses MMAE, we determined the molecular DLBCL subtype in 52 out of were observed in 16 out of 29 (55%) patients. Four out of five ABC 87 (60%) patients treated in the recently completed phase-I trials DLBCL patients responded (one CR), whereas five out of nine GCB using these agents. Twenty-six patients (50%) were classified as DLBCL obtained a partial remission (Table 2). Despite the small GCB, twenty-two (42%) as ABC and four (8%) were unclassifiable numbers of evaluable patients, these data suggest that anti-CD22- DLBCL (Supplementary Figure 1). Overall, the response evaluation MMAE and anti-CD79B-MMAE are active in both ABC and GCB of 28 DLBCL patients treated with anti-CD22-MMAE and 29 DLBCL patients (Table 2). patients treated with anti-CD79B-MMAE at doses of ⩾ 1.8 mg/kg was available (Table 1). In these 57 patients, a molecular DLBCL fi subtype classi cation was available in 29 patients (14 patients Activity of anti-CD22-MMAE and anti-CD79B-MMAE treated with anti-CD22-MMAE and 15 patients treated with in CD79B-mutated DLBCL anti-CD79B-MMAE; Table 2). For the remaining patients we Roughly 20% of primary ABC DLBCL patient samples harbor could not obtain sufficient material to perform a molecular mutations affecting CD79B.32 Previous work suggested that these subtype classification. mutations might negatively affect BCR internalization in mouse models.21 Hence, we examined whether CD79B mutations would Table 1. Response to anti-CD22-MMAE and anti-CD79B-MMAE in interfere with anti-CD79B-MMAE activity, as ADC internalization is primary DLBCL patients crucial for its mechanism of action. To this end, we investigated Response to Response to the kinetics of internalization of the naked anti-CD79B antibody in anti-CD22-MMAE (n = 28) anti-CD79B-MMAE (n = 29) the CD79B (Y196H)-mutated cell line TMD8 compared with those in the CD79B wild-type cell lines SUDHL-4 and WSU-DLCL2.32 The ORR (CR+PR) 12 (43%) 16 (55%) majority of the anti-CD79B antibody was rapidly internalized in all CR 5 (18%) 3 (10%) fi PR 7 (25%) 13 (45%) three cell lines within the rst hour of incubation. The internaliza- SD 4 (14%) 5 (17%) tion rate in TMD8 and WSU-DLCL2 cells was virtually identical and PD 12 (43%) 8 (28%) higher compared with the rate in SUDHL-4 cells (Figure 2c), suggesting that the internalization rate in human CD79B-mutated Abbreviations: CR, complete remission; DLBCL, diffuse large B-cell lymphoma; ORR, overall response rate; MMAE, monomethyl auristatin E; and wild-type lymphoma cells is comparable. This was further PD, progressive disease; PR, partial remission; SD, stable disease. underscored as treatment of the two CD79B-mutated cell lines HBL1 and TMD8,32 with anti-CD79B-MMAE-induced toxicity in both cell lines. Both HBL1 and TMD8 were highly sensitive to anti- CD79B-MMAE treatment with IC s of 0.02 and 0.16 μg/ml, Table 2. Response to anti-CD22-MMAE and anti-CD79B-MMAE in 50 molecular subtypes of DLBCL respectively (Supplementary Table 5). This suggests that CD79B mutations do not seem to impair the efficacy of anti-CD79B-MMAE Response to anti-CD22-MMAE Response to anti-CD79B-MMAE in vitro. Next, we analyzed whether mutations in CD79B might have an ABC GCB Unclassified ABC GCB Unclassified impact on response to anti-CD79B-MMAE in primary patients. To DLBCL DLBCL DLBCL DLBCL DLBCL DLBCL evaluate this, we determined the CD79B mutation status in 35 primary DLBCL samples from patients treated in both trials. We N 77 0 59 1 identified two patient samples harboring either a CD79B Y196H or ORR 3 2 0 4 5 0 a CD79B Y196C mutation, respectively. In contrast, 33 patients had CR 2 2 0 1 0 0 PR 1 0 0 3 5 0 two wild-type alleles for CD79B. As the two patients with CD79B mutations were treated with anti-CD22-MMAE (both achieved a Abbreviations: ABC, activated B cell; CR, complete remission; DLBCL, CR), we cannot evaluate the in vivo efficacy of the anti-CD79B- diffuse large B-cell lymphoma; GCB, germinal center B cell; MMAE, monomethyl auristatin E; N, number of patients; ORR, overall response MMAE treatment. However, our in vitro data suggest that anti- rate; PR, partial remission. CD79B-MMAE should be active in both patients with wild-type or mutated CD79B.

Figure 3. Predictors of response to anti-CD22-MMAE and anti-CD79B-MMAE treatment. (a) BCL2, BCL-XL, MCL1 and MYC expression determined using western blot analysis in DLBCL cell lines that are either sensitive or resistant to anti-CD22-MMAE and anti-CD79B-MMAE. Lysates from OCI-Ly7 cells were used as internal control on both blots. (b) In vitro sensitivity of DLBCL cell lines to anti-CD22-MMAE is correlated to CD22 surface expression measured using ﬂow cytometry. (c) In vitro sensitivity of DLBCL cell lines to anti-CD79B-MMAE is correlated to CD79B surface expression measured using ﬂow cytometry. (d) Response to anti-CD22-MMAE and CD22 expression determined by immunohistochemistry in DLBCL patients do not correlate (P = 0.2; χ2-test). CR: complete remission; PR: partial remission; SD: stable disease; PD: progressive disease. (e) Response to anti-CD79B-MMAE and CD79B expression determined by immunohistochemistry in DLBCL patients do not correlate (P = 0.8; χ2-test). (f) In vitro sensitivity of DLBCL cell lines to anti-CD22-MMAE is not correlated to sensitivity to MMAE. (g) In vitro sensitivity of DLBCL cell lines to anti-CD79B-MMAE is not correlated to sensitivity to MMAE.

Leukemia (2015) 1578 – 1586 © 2015 Macmillan Publishers Limited Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1583 Anti-CD22-MMAE and anti-CD79B-MMAE induce cell death arrest were induced. Treatment of the GCB DLBCL cell lines OCI- in DLBCL Ly7 and RL with anti-CD22-MMAE resulted in a signiﬁcant increase To obtain a better understanding of the growth-inhibitory effects in the sub-G1 fraction after 72 h, indicating an increase in cell of anti-CD22-MMAE and anti-CD79B-MMAE treatments observed death (Figure 2d and Supplementary Figure 3a). Similarly, anti- in DLBCL cell lines, we investigated whether cell death or cell cycle CD79B-MMAE treatment induced cell death as measured by a sub-

Anti-CD22-MMAE/anti-CD79B- Anti-CD22- Anti-CD79B- Not MMAE MMAE MMAE sensitive sensitive sensitive sensitive OCI-Ly10 RIVA U2932 BJAB DB OCI-Ly4 SUDHL-6 OCI-Ly7 OCI-Ly7 Farage HS445 HT HBL1 OCI-Ly3 Pfeiffer SUDHL-2 WSU-NHL BCL2 BCL-XL

MCL1

MYC

Tubulin

r = -0.75 r = -0.68 -6 -4 p = 6x10 p = 1x10

101 101

Anti-CD22- 100 Anti-CD79B- 100 MMAE MMAE IC50 IC50

10-1 10-1

10-2 10-2

0 20 40 60 80 100 120 140 160 0 50 100 150 200 250 CD22 surface expression (MFI) CD79B surface expression (MFI)

Response to anti-CD22-MMAE Response to anti-CD79B-MMAE

300 300

200 200 CD22 CD79B H-score H-score 100 100

0 0

CR PR SD PD CR PR SD PD

101 101

Anti-CD22- 0 0 MMAE 10 Anti-CD79B- 10 MMAE IC50 IC50

10-1 10-1

-2 r = 0.25 -2 r = 0.30 10 p = 0.2 10 p = 0.1

10-1 100 10-1 100

MMAE IC50 MMAE IC50

© 2015 Macmillan Publishers Limited Leukemia (2015) 1578 – 1586 Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1584 G1 peak increase in OCI-Ly19 and RL cells (Figure 2e and CD79B expression and response to either anti-CD22-MMAE or anti- Supplementary Figure 3b). We also observed an increase in the CD79B-MMAE is related to the applied technique (flow cytometry G2/M peak in RL and OCI-Ly7 cells treated with anti-CD22-MMAE versus immunohistochemistry), we determined CD22 and CD79B and in RL cells treated with anti-CD79B-MMAE (Supplementary expression in our panel of cell lines additionally using immunohis- Figures 3a and 3b), indicative of a mitotic arrest, consistent with tochemistry (Supplementary Table 6). Subsequently, we correlated the mechanism of action of a tubulin-targeted agent. response to treatment in the cell lines with the CD22 and CD79B To assess the contribution of the respective antibody moieties H-score. For CD79B we could not detect a significant correlation to the ADC’s mechanism of action, we measured viability of DLBCL between expression and response to anti-CD79B-MMAE treatment cell lines treated with the unconjugated anti-CD22 and anti-CD79B any more (r = − 0.21; P = 0.3), whereas for CD22 only a weak antibodies. The unconjugated antibodies did not induce any correlation between CD22 expression and response to anti-CD22- − changes in cell viability or cell cycle, suggesting that the toxic MMAE was detectable (r =0.53;P =4×10 3; Supplementary Table 5 effect of these ADCs is mediated solely by MMAE (Figures 2d and e and 6). These data suggest that immunohistochemistry is less and Supplementary Figures 3a and 3b). To confirm the effect of sensitive in accurately determining CD22 and CD79B expression MMAE on cell viability, we treated our panel of cell lines with levels compared with flow cytometry. MMAE in the absence of the antibody. Indeed, MMAE significantly Finally, we investigated whether sensitivity to MMAE correlated reduced viability in all investigated cell lines (Supplementary Table 5), with sensitivity to either anti-CD22-MMAE or anti-CD79B-MMAE demonstrating the sensitivity of these cell lines to the payload measured by the corresponding IC50 values in our panel of DLBCL drug MMAE. cell lines. For anti-CD22-MMAE (r = 0.25; P = 0.2; Figure 3f) and for Co-expression of MYC and BCL2 is associated with adverse anti-CD79B-MMAE (r = 0.30; P = 0.1; Figure 3g) we did not detect a survival in DLBCL patients and expression of BCL2 family members significant correlation, most likely because of the fact that virtually such as MCL1 has been shown to affect response to conventional all cell lines were extremely sensitive to the MMAE treatment chemotherapy in DLBCL.30,33,34 Therefore, we analyzed whether (Supplementary Table 5). expression of either MYC and/or the BCL2 family members BCL2, BCL-XL or MCL1 is associated with either response or resistance to DISCUSSION anti-CD22-MMAE and anti-CD79B-MMAE. To this end, we determined the expression pattern of these proteins using western blot ADCs represent a promising novel approach for the treatment of analysis in cell lines that are characterized by either response to malignant lymphomas. Recently, the anti-CD30 ADC brentuximab both agents or by response to only one agent or that are vedotin was approved for the treatment of relapsed CD30-positive characterized by resistance to both drugs (Figure 3a). We detected Hodgkin’s lymphoma and ALCL by the US Food and Drug BCL2 and MCL1 expression in only a fraction of the investigated cell Administration (FDA) and the European Medicines Agency (EMA). lines, whereas BCL-XL and MYC were expressed in virtually all lines, Previous work indicated that anti-CD22-MMAE and anti-CD79B- MMAE are toxic to preclinical models derived from various albeit at different levels (Figure 3a). We could not detect a 18–20 preferential expression pattern in either one of the response groups malignant B-cell lymphomas. However, these studies did (Figure 3a). not investigate whether different molecular subtypes or lymphomas with specific, somatically acquired mutations respond preferentially to these agents. In the present study we demon- CD22 and CD79B surface expression and sensitivity to MMAE as strated that the viability of the vast majority of ABC and GCB predictors of response in DLBCL DLBCL cell lines was significantly impaired following anti-CD22- Next, we investigated whether the degree of cell surface MMAE and anti-CD79B-MMAE treatment. This was validated in expression of CD22 and CD79B correlates with response to anti- two clinical studies, as both ABC and GCB DLBCL patients CD22-MMAE and anti-CD79B-MMAE in DLBCL cell lines. Indeed, responded to these agents when treated for their relapsed or we detected a significant correlation between antigen surface refractory disease: seven out of 12 (58%) ABC and seven out of 16 expression and response to both anti-CD22-MMAE (r = − 0.75; (44%) GCB DLBCL patients responded to treatment with an ADC. − 6 P =6×10 ; Figure 3b) and anti-CD79B-MMAE (r = − 0.68, However, one has to be very cautious in drawing definitive − 4 P =1×10 ; Figure 3c). However, several cell lines responded to conclusions from our phase-I studies, given the small patient treatment with anti-CD22-MMAE and anti-CD79B-MMAE despite numbers. Nevertheless, the observation of responses to single low surface target antigen expression (for example, OCI-Ly1 and agent anti-CD22-MMAE and anti-CD79B-MMAE suggests that both NUDUL-1 for anti-CD22-MMAE treatment and OCI-Ly10, OCI-Ly3, DLBCL subtypes could be responsive to these agents. Our data are OCI-Ly2, OCI-Ly4 and WSU-DLCL2 for anti-CD79B-MMAE encouraging, especially as relapsed and primary refractory DLBCL treatment; Supplementary Table 2 and 5). These data suggest patients typically experience dismal outcome.13 Similarly, CD79B that using a predefined cutoff level of surface target expression mutations that predominantly occur in ABC DLBCL patients who might not accurately predict which patients may respond to anti- are associated with inferior survival did not impair response to CD22-MMAE or anti-CD79B-MMAE treatment. anti-CD79B-MMAE in our in vitro models. This is interesting, as In order to explore this possibility, we evaluated whether previous data in mouse B cells suggested that mutations in the response to treatment in primary DLBCL patients correlates with CD79B ITAM tyrosine residues elevate surface BCR expression by CD22 and CD79B expression levels using immunohistochemistry. inhibiting receptor internalization.21 In our analyses, the inter- Response evaluation and immunohistochemistry were available nalization rate of the anti-CD79B antibody was similar in CD79B- for 18 anti-CD22-MMAE-treated and 18 anti-CD79B-MMAE-treated mutated TMD8 cells compared with two CD79B wild-type cell patients (dose level ⩾ 1.8 mg/kg). The response to both ADCs was lines. These results were underscored as the CD79B ITAM not correlated to the expression of the targets CD22 and CD79B, mutations did not decrease the efficacy of anti-CD79B-MMAE in respectively (P = 0.2 for CD22 and P = 0.8 for CD79B; χ2-test; Figures HBL1 and TMD8 cells. However, results of future studies in patients 3d and e). A low H-score did not predict resistance to these with mutated CD79B treated with anti-CD79B-MMAE need to be agents. Thus, the in vivo data supported our hypothesis that using awaited to confirm these data in a clinical setting. a predefined CD22 or CD79B expression cutoff level for treatment The finding that anti-CD22 and anti-CD79B ADCs can successfully selection might exclude patients who may still benefit from be utilized therapeutically in both ABC and GCB DLBCL patients is effective ADC therapies. clinically relevant. In contrast to anti-CD22 and anti-CD79B ADCs, To investigate whether the discrepant results in cell lines and various novel compounds such as the Bruton’s tyrosine kinase primary patient samples regarding the correlation of CD22 and inhibitor ibrutinib have been implicated to be preferentially active

Leukemia (2015) 1578 – 1586 © 2015 Macmillan Publishers Limited Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1585 in ABC DLBCL patients.6 Similarly, the immunomodulatory drug REFERENCES lenalidomide had activity predominantly in relapsed or refractory 1 Nogai H, Dorken B, Lenz G. Pathogenesis of Non-Hodgkin's Lymphoma. J Clin ABC DLBCL patients in a recent phase-II trial.7 Thus, for the optimal Oncol 2011; 29: 1803–1811. use of these agents an accurate molecular diagnosis and 2 Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A et al. Distinct types assignment into cell of origin subgroups presumably will become of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 403 – a prerequisite. However, despite recent improvements using novel 2000; : 503 511. fi 3 Wright G, Tan B, Rosenwald A, Hurt EH, Wiestner A, Staudt LM. A gene expression- techniques such as the NanoString technology, the classi cation based method to diagnose clinically distinct subgroups of diffuse large B cell into ABC and GCB DLBCL is still not routinely made in clinical lymphoma. Proc Natl Acad Sci USA 2003; 100: 9991–9996. 35,36 practice. Thus, having agents such as anti-CD22-MMAE and 4 Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RCT et al. Diffuse large anti-CD79B-MMAE that are equally effective in the treatment of B-cell lymphoma outcome prediction by gene-expression profiling and super- both ABC and GCB DLBCL patients and for which molecular vised machine learning. Nat Med 2002; 8:68–74. fi 5 Monti S, Savage KJ, Kutok JL, Feuerhake F, Kurtin P, Mihm M et al. Molecular classi cation is not absolutely required might be advantageous. fi fi Interestingly, the anti-CD22 and anti-CD79B antibodies alone pro ling of diffuse large B-cell lymphoma identi es robust subtypes including one characterized by host inflammatory response. Blood 2005; 105: did not induce any toxicity in the investigated DLBCL cell lines. 1851–1861. These data indicate that the toxic effects of the anti-CD22 and 6 Wilson W, Gerecitano J, Goy A, de Vos S, Kenkre V, Barr P et al. The Bruton's anti-CD79B ADCs are caused solely by MMAE. However, sensitivity tyrosine kinase (BTK) inhibitor, Ibrutinib (PCI-32765), has preferential activity in to MMAE did not correlate with sensitivity to anti-CD22-MMAE or the ABC subtype of relapsed/refractory de novo diffuse large B-cell lymphoma anti-CD79B-MMAE, suggesting that additional features or the (DLBCL): interim results of a multicenter, open-label, phase 2 study. Blood 2012; combination of different factors determine sensitivity to these 120: Abstract 623. agents. Potentially, both target gene expression as well as 7 Hernandez-Ilizaliturri FJ, Deeb G, Zinzani PL, Pileri SA, Malik F, Macon WR et al. Higher response to lenalidomide in relapsed/refractory diffuse large B-cell lym- sensitivity to MMAE are of major importance in dictating response phoma in nongerminal center B-cell-like than in germinal center B-cell-like to anti-CD22 and anti-CD79B ADCs. phenotype. Cancer 2011; 117: 5058–5066. We detected a significant correlation between sensitivity to 8 Dunleavy K, Pittaluga S, Czuczman MS, Dave SS, Wright G, Grant N et al. anti-CD22-MMAE and anti-CD79B-MMAE and CD22 and CD79B Differential efficacy of bortezomib plus chemotherapy within molecular subtypes surface expression determined using flow cytometry in DLBCL of diffuse large B-cell lymphoma. Blood 2009; 113: 6069–6076. cell lines. However, antilymphoma responses of the ADC 9 Lenz G, Wright G, Dave SS, Xiao W, Powell J, Zhao H et al. Stromal gene signatures 359 – treatment were observed in DLBCL cell lines with expression in large-B-cell lymphomas. N Engl J Med 2008; : 2313 2323. 10 Coiffier B, Lepage E, Briere J, Herbrecht R, Tilly H, Bouabdallah R et al. levels across the entire dynamic range of surface expression. In CHOP chemotherapy plus rituximab compared with CHOP alone in elderly addition, in our panel of cell lines CD79B expression determined patients with diffuse large-B-cell lymphoma. N Engl J Med 2002; 346:235–242. using immunohistochemistry and response to anti-CD79B-MMAE 11 Pfreundschuh M, Schubert J, Ziepert M, Schmits R, Mohren M, Lengfelder E et al. did not correlate, whereas for CD22 only a weak correlation Six versus eight cycles of bi-weekly CHOP-14 with or without rituximab in elderly between expression and response to anti-CD22-MMAE was patients with aggressive CD20+ B-cell lymphomas: a randomised controlled trial detectable. Even more importantly CD22 and CD79B expression (RICOVER-60). Lancet Oncol 2008; 9: 105–116. determined with immunohistochemistry did not correlate with 12 Habermann TM, Weller EA, Morrison VA, Gascoyne RD, Cassileth PA, Cohn JB et al. Rituximab-CHOP versus CHOP alone or with maintenance rituximab in response to anti-CD22-MMAE and anti-CD79B-MMAE in primary older patients with diffuse large B-cell lymphoma. J Clin Oncol 2006; 24: DLBCL patients, indicating that the target expression levels 3121–3127. detected using immunohistochemistry may not reliably predict 13 Gisselbrecht C, Glass B, Mounier N, Singh Gill D, Linch DC, Trneny M et al. Salvage patient response upfront. Even patients with no detectable levels regimens with autologous transplantation for relapsed large B-cell lymphoma in of CD22 by immunohistochemistry responded. A similar phe- the rituximab era. J Clin Oncol 2010; 28: 4184–4190. nomenon has previously been observed in patients with various 14 Leslie LA, Younes A. Antibody-drug conjugates in hematologic malignancies. Am Soc Clin Oncol Educ Book 2013, 108–113. CD30-positive lymphoma subtypes treated with brentuximab 37 15 Hamblett KJ, Senter PD, Chace DF, Sun MM, Lenox J, Cerveny CG et al. Effects of vedotin. In these patients no correlation was observable drug loading on the antitumor activity of a monoclonal antibody drug conjugate. 37 between CD30 expression and response. Our data suggest Clin Cancer Res 2004; 10: 7063–7070. that the expression levels required for activity of CD22 ADC may 16 Younes A, Bartlett NL, Leonard JP, Kennedy DA, Lynch CM, Sievers EL et al. be below the level of detection of the immunohistochemistry Brentuximab vedotin (SGN-35) for relapsed CD30-positive lymphomas. NEnglJ assay used in our studies. Thus, our collective data indicate that Med 2010; 363: 1812–1821. patients with DLBCL should not be excluded from therapy with 17 Pro B, Advani R, Brice P, Bartlett NL, Rosenblatt JD, Illidge T et al. Brentuximab vedotin (SGN-35) in patients with relapsed or refractory systemic anaplastic these agents based on low CD22 or CD79B expression using large-cell lymphoma: results of a phase II study. J Clin Oncol 2012; 30: 2190–2196. immunohistochemistry. 18 Polson AG, Yu SF, Elkins K, Zheng B, Clark S, Ingle GS et al. Antibody-drug In summary, anti-CD22-MMAE and anti-CD79B-MMAE represent conjugates targeted to CD79 for the treatment of non-Hodgkin lymphoma. Blood active agents for the therapy of different DLBCL subtypes. Despite 2007; 110: 616–623. the small number of treated patients, we detected responses in 19 Dornan D, Bennett F, Chen Y, Dennis M, Eaton D, Elkins K et al. Therapeutic both ABC and GCB DLBCL patients recapitulating the activity seen potential of an anti-CD79b antibody-drug conjugate, anti-CD79b-vc-MMAE, for 114 – in vitro with representative cell lines. the treatment of non-Hodgkin lymphoma. Blood 2009; : 2721 2729. 20 Polson AG, Williams M, Gray AM, Fuji RN, Poon KA, McBride J et al. Anti-CD22- MCC-DM1: an antibody-drug conjugate with a stable linker for the treatment of CONFLICT OF INTEREST non-Hodgkin's lymphoma. Leukemia 2010; 24:1566–1573. 21 Gazumyan A, Reichlin A, Nussenzweig MC. Ig beta tyrosine residues contribute to BZ, HK, RM, AC, TS, Y-WC, AGP and KEM are employees of Genentech. MCP-W is an the control of B cell receptor signaling by regulating receptor internalization. employee of Seattle Genetics. GL received research funding from Genentech. J Exp Med 2006; 203: 1785–1794. 22 Li D, Poon KA, Yu SF, Dere R, Go M, Lau J et al. DCDT2980S, an anti-CD22- monomethyl auristatin E antibody-drug conjugate, is a potential treatment for ACKNOWLEDGEMENTS non-Hodgkin lymphoma. Mol Cancer Ther 2013; 12:1255–1265. This work was supported by research grants to GL from Genentech and the Deutsche 23 Advani R, Chen AI, Lebovic D, Brunvand MW, Goy A, Chang JE et al. Final results of Krebshilfe. In addition, this work was supported by the Deutsche Forschungsge- a phase I study of the anti-CD22 antibody-drug conjugate (ADC) DCDT2980S with meinschaft, DFG EXC 1003 Cells in Motion—Cluster of Excellence, Münster, Germany, or without rituximab (RTX) in patients (Pts) with relapsed or refractory (R/R) B-cell as well as by a Doctoral Scholarship to MG from the Philipps-University Marburg. Non-Hodgkin's Lymphoma (NHL). Blood 2013; 122: Abstract 4399.

© 2015 Macmillan Publishers Limited Leukemia (2015) 1578 – 1586 Anti-CD22 and anti-CD79B ADCs in DLBCL subtypes M Pfeifer et al 1586 24 Palanca-Wessels MC, Salles GA, Czuczman MS, Assouline SE, Flinn IW, Sehn LH 31 Weilemann A, Grau M, Erdmann T, Merkel O, Sobhiafshar U, Anagnostopoulos I et al. Final results of a phase I study of the anti-CD79b antibody-drug conjugate et al. Essential role of IRF4 and MYC signaling for survival of anaplastic large cell DCDS4501A in relapsed or refractory (R/R) B-cell Non-Hodgkin Lymphoma (NHL). lymphoma. Blood 2015; 125: 124–132. Blood 2013; 122: Abstract 4400. 32 Davis RE, Ngo VN, Lenz G, Tolar P, Young RM, Romesser PB et al. Chronic active 25 Pfeifer M, Grau M, Lenze D, Wenzel SS, Wolf A, Wollert-Wulf B et al. PTEN loss B-cell-receptor signalling in diffuse large B-cell lymphoma. Nature 2010; 463:88–92. defines a PI3K/AKT pathway-dependent germinal center subtype of diffuse large 33 Green TM, Young KH, Visco C, Xu-Monette ZY, Orazi A, Go RS et al. Immuno- B-cell lymphoma. Proc Natl Acad Sci USA 2013; 110: 12420–12425. histochemical double-hit score is a strong predictor of outcome in patients with 26 Nogai H, Wenzel SS, Hailfinger S, Grau M, Kaergel E, Seitz V et al. IkappaB-zeta diffuse large B-cell lymphoma treated with rituximab plus cyclophosphamide, controls the constitutive NF-kappaB target gene network and survival of doxorubicin, vincristine, and prednisone. J Clin Oncol 2012; 30: 3460–3467. ABC DLBCL. Blood 2013; 122: 2242–2250. 34 Johnson NA, Slack GW, Savage KJ, Connors JM, Ben-Neriah S, Rogic S et al. 27 Zheng B, Fuji RN, Elkins K, Yu SF, Fuh FK, Chuh J et al. In vivo effects of targeting Concurrent expression of MYC and BCL2 in diffuse large B-cell lymphoma treated CD79b with antibodies and antibody-drug conjugates. Mol Cancer Ther 2009; 8: with rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone. 2937–2946. J Clin Oncol 2012; 30: 3452–3459. 28 Johnston SR, Saccani-Jotti G, Smith IE, Salter J, Newby J, Coppen M et al. 35 Scott DW, Wright GW, Williams PM, Lih CJ, Walsh W, Jaffe ES et al. Determining Changes in estrogen receptor, progesterone receptor, and pS2 cell-of-origin subtypes of diffuse large B-cell lymphoma using gene expression in expression in tamoxifen-resistant human breast cancer. Cancer Res 1995; 55: formalin-fixed paraffin-embedded tissue. Blood 2014; 123: 1214–1217. 3331–3338. 36 Masque-Soler N, Szczepanowski M, Kohler CW, Spang R, Klapper W. Molecular 29 Vallejo-Gracia A, Bielanska J, Hernandez-Losa J, Castellvi J, Ruiz-Marcellan MC, classification of mature aggressive B-cell lymphoma using digital multiplexed Ramon y CS et al. Emerging role for the voltage-dependent K+ channel Kv1.5 in gene expression on formalin-fixed paraffin-embedded biopsy specimens. Blood B-lymphocyte physiology: expression associated with human lymphoma malig- 2013; 122:1985–1986. nancy. J Leukoc Biol 2013; 94: 779–789. 37 Bartlett NL, Sharman JP, Oki Y, Advani RH, Bello CM, Winter JN et al. A phase 2 30 Wenzel SS, Grau M, Mavis C, Hailfinger S, Wolf A, Madle H et al. MCL1 is study of brentuximab vedotin in patients with relapsed or refractory CD30- deregulated in subgroups of diffuse large B-cell lymphoma. Leukemia 2013; 27: positive non-Hodgkin lymphomas: interim results in patients with DLBCL and 1381–1390. other B-cell lymphomas. Blood 2013; Abstract 848.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)