A Novel Recombinant Anti-CD22 Immunokinase Delivers Proapoptotic Activity of Death-Associated Protein Kinase (DAPK) and Mediates Cytotoxicity in Neoplastic B Cells Free

Functional screens with antisense libraries had enriched death-associated protein kinases (DAPK) as a prominent death-protective cluster in the presence of cytotoxic signals (1, 2). Knowledge on DAPKs has evolved since then with DAPK1 established as the prototypic serine-threonine kinase of a family of 5 proapoptotic proteins. They encompass DAPK2 (DAPK1-Related Protein 1 = DRP-1), zipper-interacting protein kinase (ZIPK = DAPK3), DAP-related apoptotic kinase (DRAK) 1, and DRAK2 (1–4). They share a 50% to 80% homology of their catalytic domains. Loss-of-function of the DAPK tumor suppressors is implicated as a central tumor-initiating and metastasis -promoting event in a broad range of cancers, that is, of breast, lung, head and neck, gastrointestinal, and the hematopoetic system (1, 5, 6). The causes of tumor-associated DAPK deficiency are primarily epigenetic mechanisms (promoter hypermethylation), but also genomic deletions with subsequent LOH are observed (1, 7, 8).

The DAPK proteome of multimeric complexes involves phosphorylated substrates that transmit apoptotic or autophagic signals in response to cellular stresses including those by redox burden or oncogenes (1, 2, 9). Robust protective DAPK function promotes p19ARF-mediated p53 activity (10). Loss of DAPK proficiency is tumorigenic in the context of oncogenic stress, that is, by Myc- or E2F1, as this attenuates safeguarding p53-conferred cell death. Lost DAPK can also no longer retain oncogenic ERK1/2 in the cytoplasm (11). The operational downstream pathways (caspase- and p53-dependent vs. -independent apoptosis; ref. 1) and cell biologic outcomes [i.e., apoptosis (1, 10) versus autophagic preservation (1) vs autophagic cell death (12, 13) vs. necroptosis (14)] appear cell context specific and to be dictated by the nature of the upstream stressor and the executing DAPK family member (1, 2, 15).

Reduced DAPK1 mRNA transcription due to promoter hypermethylation and aberrant distribution of affected CpG islands was demonstrated in virtually all cases of sporadic chronic lymphocytic leukemia (CLL; refs. 6, 16). In addition, a CLL haplotype with allele-specific imbalances of DAPK1 expression indicates that recurring germline single-nucleotide variants and promoter methylation can be considered as predisposing to (familial) CLL (16, 17). DAPK1 promoter hypermethylation was also shown to be associated with more aggressive disease and adverse clinical outcome in diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma (FL; refs. 5, 18).

Given their central role in tumorigenesis and clonal sustenance, DAPKs have been intensely investigated as therapeutic targets. Strategies employ gene therapy, development of functional DAPK activators, or demethylating agents to reestablish DAPK expression (19, 20). This also extents (reversely) to DAPK inhibitors in autoinflammatory disease and graft rejection (15, 20–22). We explored here if reinstating DAPK activity would perturb the high apoptotic threshold in DAKP-deficient neoplastic B cells that are burdened by incurring (oncogenic) stress. We further hypothesized that an efficient immunoligand-based targeting would potentiate that selectivity. Our DAPK2′-CD30L immunokinase already showed encouraging in vitro (23) and in vivo (24) activity in systems of Hodgkin lymphoma.

A novel CD22-targeting construct containing a constitutively active DAPK1 variant showed here efficient binding, internalization, and cell death induction in malignant B cells. This immunokinase revealed a favorable in vitro profile of induced apoptosis and autophagy, including efficacy in high-risk CLL. Its novel design and high-yield purification protocols are described with the advantages and feasibility aspects encouraging further exploration toward a clinical application. This concept of immunoligand-based toxic reconstitution (not mere rescue replenishment) of a tumor suppressor function would supplement mAbs, small-molecule inhibitors, and T-cell–based therapies (25) whose implementation into the therapeutic approaches in B-cell lymphomas, particularly CLL, has ushered a new era of potentially chemotherapy-free targeted treatments (26).

Peripheral blood samples were obtained from 31 CLL patients (diagnosed according to iwCLL criteria; ref. 27) that were either treatment-naïve or had a minimum interval from any last therapy of 4 weeks. After informed written consent, sampling was performed in accordance with the Declaration of Helsinki and the guidelines at the University of Cologne (IRB approval #01-163). CLL B cells, healthy donor–derived B cells (purities as per flow cytometry; for CLL >95%), and healthy donor peripheral blood mononuclear cells (PBMC) were isolated as described previously (28) or, for tonsillar B cells, followed a MACS protocol (Supplementary Data). The cell lines REH, Nalm6, SP53, Mino, Granta, BC-1, MEC1, JVM3, BJAB, Namalwa-PNT, Daudi, Raji, Ramos, DoHH2, L428, KMS-12-BM, RPMI-8226, HEK293T, and U937 (Supplementary Table S1) were originally acquired between 2012 and 2014 from the DSMZ and from the ATCC. Only original stock propagated immediately upon arrival for 2 to 3 passages was picked for studies and cultures terminated after the 10th round of passaging (5–7 weeks). Upon thawing for experimentation between 2012 and 2015, all lines were authenticated by flow cytometry confirming their characteristic immunophenotype, including Ig composition and TCL1 levels (28–30). The same applied to the FL/DLBCL lines LP, FN, CJ, LR, MS, DBr, DS, EJ established and provided by Dr. R.J. Ford (UT M.D. Anderson Cancer Center, Houston, TX). Their identity has been described previously (29, 31). Further details are indicated in the Supplementary Data. Experimentations were done in suspension cultures under conditions as instructed and published (28, 29).

Protocols and reagents for immunoblots (DAPKs and apoptotic proteins), IHC (DAPK1), and flow cytometry are outlined in the Supplementary Data.

Via manual search of literature and data repositories, we obtained available primary data out of 8 reports on array-based gene expression profiling (GEP) of CLL samples with specific controls. Details and further meta-analyses on B-cell subsets and other B-cell lymphomas are found in the Supplementary Data. All datasets were separately background corrected and annotated using the BioConductor packages “affy” and “biomaRt” in R-3.1.0. The datasets were quantile-normalized and replicates were combined by mean. Sample names were assigned from the GEO (32) entry to test for differential expression via t test or Wilcoxon/Mann–Whitney test. Only the highest fold change (FC; P < 0.1) was obtained and visualized in the summarizing heatmap.

The transient transfection of HEK293T cells employed a liposome-based protocol (33). Purification of recombinant protein containing supernatants was done in a three-step procedure. First, an immobilized nickel-nitrilotriacetic (NiNTA) acid metal affinity chromatography (IMAC; His-Trap; GE Lifesciences) based fast protein liquid chromatography (FPLC) was performed. After buffer exchange, the eluted fractions were applied on a CM sepharose column for cation exchange chromatography (CEC). Peptide analysis was carried out using an ESI Q-Tof-2 MS (Waters Micromass) mass spectrometer (23). Sequences of individual peptides were identified using the Mascot algorithm (Matrix Science).

Specific binding of DK1KD-SGIII and all other generated proteins to primary cells and cell lines were determined by flow cytometry. Competitive binding and internalization assays as well as confirmatory fluorescence microscopy are described in detail in the Supplementary Data. Briefly, after DK1KD-SGIII incubation and washing, cells were labeled with monoclonal Alexa Fluor 488 anti-penta histidine antibodies (AHA; Qiagen). The competition assays were based on saturation by an APC-labeled anti-human CD22 antibody RFB-4 (Invitrogen). For flow cytometry–based internalization assays, DK1KD-SGIII and AHA were added to the target cells. After 30-minute incubation, surface proteins were removed by addition of proteinase K for 45 minutes (Qiagen) and levels of internalized AHA-labeled protein were determined. Measurements were performed on FACSCalibur (Beckton Dickinson) and Gallios (Beckman Coulter) cytometers. Antibodies against human CD5, CD19, IgG2a, and IgG1 were from Biolegend. Data were analyzed with the Cyflogic software (Cyflo).

In vitro kinase activity of recombinant DAPK1 was determined using the PKLight Assay Kit (Lonza) according to the manufacturer's instructions. Instead of protein kinase A (PKA), recombinant myosin light chain II (MLCII; ProSpec) served as a substrate for DK1KD.

The colorimetric 3-[4,5-dimethylthiazol-2-yl-2,5-diphenyl] tetrazolium bromide (MTT) assay (Promega) assessed cell viability by measuring metabolic activity in duplicates per sample as described previously (28). Averaged absorbances of the experimental conditions were normalized to the control conditions. Apoptosis rates were measured by flow cytometry on a Gallios cytometer (Beckmann Coulter) with 5 × 10⁵ cells stained with AnnexinV/7AAD (Becton Dickinson). Preincubations with 20 μmol/L of the pan-caspase inhibitor ZVAD-fmk (Promega) were done for 60 minutes.

Parametric Student t test estimations (either paired or nonpaired) were performed in GraphPad Prism. P values <0.05 were considered significant. Data are presented as means with indicated error bars as SEM.

Further information about sources of DNA, vectors, cloning of DAPK1 and DAPK2 mutants, cloning of the most suitable scFv, antibodies, transfections, purification, and mass spectrometry as well as in silico meta-analyses are given in the Supplementary Data.

The scarce literature on DAPK expression in inflammatory cells implicates T cells and macrophages as the main source (15). Alternative splicing creates various DAPK1 isoforms. A kinase-less truncated s-DAPK1 variant can mediate protein degradation of full-length DAPK1 (34). Here, we evaluated the protein expression of full-length DAPK1 and DAPK2 in CLL patient samples, in multiple B-cell tumor lines, and in B cells from healthy donors. We conclude that in the context of a generally low-level DAPK expression in the B-cell lineage, particularly CLL cells, lack noticeable DAPK1. In detail, prominent DAPK1 signals in normal lymph nodes originated from perivascular macrophages. In reactive tonsils, only a minority of large centroblastic cells of follicular germinal centers revealed cytoplasmic signals (Supplementary Fig. S1A). Pan-CD19 isolates from tonsils (N = 3; purities 98%–99%) showed weak DAPK1 expression. The CD19-positive fraction derived from healthy donor peripheral blood samples (N = 4) revealed signals below those from DAPK transfectants, but above those from CLL (Supplementary Table S2; Fig. 1 and Supplementary Fig. S1). Given their purities of 88% to 89%, contaminating non-B cells might have contributed to the signal from this pool of naïve B cells. Among non-CLL B-cell lymphomas, unequivocal DAPK1 expression was most frequently seen in DLBCL lines. For the vast majority of CLL, there was an absence of DAPK1 and DAPK2 proteins. DAPK1 was not detected in 10 of 11 cases (91%) and 7 of 9 (78%) cases lacked noticeable DAPK2 levels (Fig. 1; Supplementary Table S2). Similarly, the CLL-derived lines MEC1 and JVM-3 expressed no or negligible amounts of DAPK1/2. Beyond initial reports of lost DAPK1 mRNA in CLL (16, 17), this pattern had not been shown for DAPK1 and DAPK2 proteins before.

The predominant lack of DAPK1 expression in CLL alongside a low-level retention in other B-cell lymphomas and normal B-cell subsets was confirmed by comparative meta-analyses of DAPK expression from publicly available raw data of array-based gene expression profiles (GEP). For CLL, we performed this from 9 technically comparable datasets comprising comparisons of CLL and normal B cells from 8 reports (protocol in online Supplementary Data). Therein, most of the DAPK genes revealed a downregulation in CLL-to-normal comparisons (Fig. 1C), but with ZIPK often showing an opposite directionality to the rest of the gene family. Data extracted from the EBI/Atlas RDF repository (35) confirmed these findings (status 07/20/2015; online Supplementary Data). The overall subtle (−4 to +4) fold changes in these CLL versus B-cell comparisons implicated what had been reported before, namely that normal B cells include sizeable DAPK-silenced populations as compared with non-B cells (36). In CLL, scenarios of treatment seem to be associated with upregulation of DAPK(1) or selection for less DAPK(1)-hypermethylated subclones. Additional metacomparisons revealed that the already repressed levels of DAPK1 show very low variation across normal B-cell subsets (Supplementary Fig. S1B) and that DAPK1 levels in most non-CLL B-cell lymphomas exceed those in normal B cells (Supplementary Fig. S1C).

The tumor-suppressive function of DAPK1 likely relies on its high proapoptotic/-autophagic activity (1). Accordingly, the observed loss of DAPK expression might contribute to the prominent prosurvival phenotype of CLL. Therefore, we postulated that constitutively active DAPK mutants would reconstitute DAPK signaling and may reduce the leukemia-specific apoptotic threshold. The function of wild-type DAPK1 is tightly regulated by an autoinhibitory Ca²⁺/Calmodulin (CaM) element located next to its kinase domain (Fig. 2A). DAPK1 further consists of a cytoskeleton binding domain and a death domain. Removal of the CaM sequence renders this truncated DAPK1 constitutively active (23, 37–39). Therefore, we generated different mutants derived from DAPK1 and -2, named DAPK1KDCyt, DK1KD, DAPK2ΔCaM, and DK2KD, all lacking the regulatory CaM element (Fig. 2A). Overexpressed in Raji lymphoma and JVM-3 CLL-like cells (pcDNA3.1 vector), DK1KD was the most efficient of these constructs (including in comparison with full-length DAPK1 and -2) in inducing cell death (Supplementary Fig. S2) and was hence selected for further analysis.

We aimed for CLL-specific delivery of the constitutively active DAPK mutant DK1KD via fusion to an immunoglobulin scFv. The initial screening for the most suitable binding domains for such an immunoligand-based fusion protein entailed preselection of antibodies and antibody fragments based on criteria of reasonable antigen specificity, antigen density, low immunogenicity, slow degradation (this excluded CD5; ref. 40), and capacity of internalization. Given these criteria and based on published data for similar constructs as well as considering the surface marker profile of CLL, we chose CD19, CD22, and CD40 (41–43). Several mAb and/or scFv for each of these epitopes as well as a CD40L construct were tested (for sources and details, see Acknowledgements and Supplementary Data). We first selected 3 most suitable constructs.

The specific antibody anti-CD19 HD37 as well as the scFv's anti-CD22 SGIII and anti-CD40 G28-5 were cloned into the pMS vector, expressed recombinantly, and tested for binding to Raji lymphoma cells and primary CLL isolates. To exclude unspecific binding, these ligands were also tested on CD19-negative, CD22-negative, and CD40-deficient myeloid/monocytic U937 cells. Specific binding was detected for CD22 (SGIII) and CD40 (G28-5) in Raji and CLL cells (Fig. 2B). In CLL, receptor binding also often mediates prosurvival signaling, rescuing primary CLL cells from their pronounced spontaneous apoptosis in vitro. Hence, survival/viability (AnnexinV/7AAD flow cytometry/MTT assays) of CLL cells (3 patient samples) was assessed before (0h) and 48 hours after incubation with the binding domain candidates. It revealed that the anti-CD40 scFv G28-5 decreased spontaneous in vitro apoptosis by almost 20%, whereas CLL cell survival/viability were not affected by the other candidate ligands, including SGIII. HD37 was also excluded from further studies. Its binding to the B-cell receptor coreceptor CD19 on positive target cells was highly heterogeneous (an example of HD37 binding failure in a CLL sample is given in Fig. 2B) and misfolding in the HEK293T system was suspected. Consequently, the anti-CD22 scFv SGIII was selected for further work as part of the delivery approach of constitutive DAPK1. SGIII is a humanized scFv derived from the murine anti-CD22 mAb RFB-4 by specificity grafting into frameworks based on scaffolds preselected for stability from phage display libraries. SGIII exhibits superior antigen binding and stability (44). Recent data, including from clinical trials, confirm the suitability of CD22 as a target of mAb-based therapies (45). We finally screened various B-cell lymphoma lines for CD22 expression with the original RFB-4 clone. It corroborated the results of the SGIII binding studies and revealed that target receptor densities are largely independent of major histogenetic differentiation stages except for the known CD22-low/absent entities of Hodgkin lymphoma and multiple myeloma (Supplementary Table S1b).

The selected components “DK1KD” and “SGIII” as the best performing DAPK kinase mutant and scFv binding ligand, respectively (above), were cloned into the pMS vector (Supplementary Fig. S3) for expression of a fused recombinant immunoligand “DK1KD-SGIII”. The pMS enables expression and purification of the target protein through an Igκ leader sequence promoting immediate secretion of the (fusion) protein into the supernatant (46). Therefore, DK1KD did not execute its cytotoxic effects in the HEK293T producers. A 6xHis-tag remains part of the secreted protein and facilitates its later detection. An IRES-eGFP creates a bicistronic transcript for detection of producer cells expressing the fusion protein without an eGFP labeled secretion product.

HEK293T cells were transfected with the pMS versions (Supplementary Fig. S3) followed by a novel multi-step up-scalable purification protocol resulting in an average yield of 2 mg/L. The first step included the protein purification from the supernatants in a NiNTA-driven IMAC-based FPLC (detailed description in online Supplementary Data). The various fractions were assessed by Coomassie staining after SDS gel separation (Fig. 2C). In the second step, after buffer exchange via dialysis, the protein-containing fractions were purified by cation exchange chromatography (CEC). Contaminating proteins eluted from CEC were found in fractions 3–8 (e.g., fraction 5 in Fig. 2D) while the target protein eluted later and was found in the second elution peak (fractions 35–38 in Fig. 2D). These fractions revealed only one band exclusively at the expected 62–70 kDa region by both Coomassie stains and anti-His Western blot analysis (Fig. 2E, left). To further confirm the identity of the purified protein from this elution fraction, the 62–70 kDa band was excised from the Coomassie-stained gel, analyzed via electron spray ionization tandem mass spectrometry (ESI-MS/MS), and compared with the predicted sequence of the fusion protein in the Mascot protein databank. Peptide mass/charge profiles identified DK1KD-SGIII with sequence coverage of 49.8% and an overall score of 1016.1 (Fig. 2E, right).

Catalytic activity of the DAPK1 mutant DK1KD in fusion to SGIII would be necessary to induce cell death in CLL cells upon delivery. Therefore, catalytic activity of DK1KD-SGIII and the respective controls DK1KD and SGIII were investigated by an in vitro kinase activity assay. DAPK1′s potent proapoptotic/autophagic capabilities are mediated by phosphorylation of its various targets, among which myosin light chain II (MLCII) is one of the best established. Consequently, in vitro kinase activity assays with the exogenous recombinant substrate MLCII were performed. Protein kinase A with its substrate kemptide (luciferin) served as the positive control system. The underlying principle of this assay is a competition for ATP between luciferase and the kinase DK1KD-SGIII (DK1KD). Therefore, luciferase activity, measured in relative light units (RLU) by a luminescence reader, is inversely proportional to DK1KD kinase activity with higher luciferase quenching reflecting higher DK1KD activity (Supplementary Fig. S4). In the presence of 5 nmol/L DK1KD or 5 nmol/L DK1KD-SGIII, bioluminescent signals dropped significantly as compared with the kinase-free control, indicating DK1KD dependent kinase activity. In contrast, 5 nmol/L of SGIII did not affect light emission, hence did not elicit nonspecific kinase activity (Fig. 2F, left). In addition, no significant difference of catalytic activity was identified between DK1KD and DK1KD-SGIII demonstrating that the fusion of DK1KD to SGIII did not affect DK1KD's enzymatic activity. Moreover, kinase activity was dependent exclusively on the concentration of DK1KD as part of the DK1KD-SGIII fusion protein. Different concentrations of DK1KD-SGIII (5–50 nmol/L), unlike respective equimolar concentrations of SGIII, led to dose-dependent increases in kinase activity (Fig. 2F, right). Taken together, the SGIII ligand-tethered kinase DK1KD demonstrated specific and potent catalytic activity by phosphorylating DAPK1′s physiologic substrate MLCII.

For DK1KD-SGIII as an immunotherapeutic, highly specific delivery into target cells is necessary. We first analyzed the surface binding of DK1KD, DK1KD-SGIII, and SGIII to primary CLL cells and various CD22 positive (+) or negative (−) cell lines via isotype-controlled flow cytometric analysis with Alexa Fluor 488 anti-His antibodies (AHA). This demonstrated that SGIII mediated selective binding of the construct to human CD22(+) cells. We documented strong binding of DK1KD-SGIII and SGIII, but not of DK1KD, to B-cell lymphomas, that is, with increased intensities in JVM-3, Raji, MEC1. No binding was detected in CD22(−) U937 cells. The binding of DK1KD-SGIII to CLL cells was uniform and dose dependent (Fig. 3A and Supplementary Figs. S5 and S6). Confirmatory fluorescence microscopy of CLL cells incubated with DK1KD-SGIII recombinant protein and AHA visualized its strong accumulation on the cell surface (Fig. 3B).

Epitope specificities of DK1KD-SGIII as a selective CD22-scFv–based construct were addressed in more detail in competitive binding assays. Binding of DK1KD-SGIII to CLL cells as detected by AHA was increasingly prevented by preincubation with increments of the APC-coupled α-CD22 mAb RFB-4, while incubation with nonspecific IgG2a reduced DK1KD-SGIII binding only slightly (Fig. 3C). Successful binding of the preincubated RFB-4-APC was measured simultaneously in a different channel. Reciprocally, preincubated SGIII-based constructs at 20-fold excess largely prevented detection of RFB-4-APC signals.

The recombinant immune kinase needs to be internalized by the targeted CLL cells to exhibit its cytotoxic function. In flow cytometry–based internalization assays, cells were either treated with proteinase K, which removed extracellular proteins including CD22, followed by incubation with 50 nmol/L DK1KD-SGIII and AHA or vice versa. Cells exclusively treated with proteinase K or DK1KD-SGIII served as negative or positive controls of the specificity of the fluorescent signal, respectively. As expected, preincubation of CD22(+) Ramos lymphoma cells with proteinase K and subsequent incubation with DK1KD-SGIII revealed hardly any binding of the CD22-targeted immune conjugate. The proteinase-K–mediated loss of surface CD22 was confirmed by RFB4-APC. In contrast, strong fluorescence signals were detected when DK1KD-SGIII was added prior to proteinase K (Fig. 3D). As proteinase K removed any extracellular receptor–bound protein–AHA complexes, it had to be concluded that this fluorescent signal originated from already internalized CD22–DK1KD–SGIII–AHA complexes. This uptake of DK1KD-SGIII occurred within 45 minutes. Confirmatory fluorescence microscopy also demonstrated a rapid (≤1 hour) intake of DK1KD-SGIII into primary CLL cells (Fig. 3E).

We next addressed the question whether DK1KD-SGIII, mimicking reinstated DAPK1, is able to resemble DAPK1′s cytotoxic potential. Cell death induction of DK1KD-SGIII in CD22(+) and CD22(−) cells was determined by AnnexinV/7AAD flow cytometry and MTT assays after 48-hour incubation with increasing concentrations of the immune ligands. Treatment of CD22(+) CLL-like B-cell lines (JVM3, MEC1) and lymphoma lines (Raji, Ramos, DOHH2) with DK1KD-SGIII resulted in cell death/reduced viability in a dose-dependent fashion (Fig. 4A). IC₅₀ values of CD22-low JVM-3 cells were approximately 8-fold higher (extrapolated: 1,300 nmol/L) as compared with CD22-high MEC1 or DOHH2 cells (170 nmol/L and 175 nmol/L, respectively). Further supporting a CD22/compound–response relationship, both CD22 expression as measured by DK1KD-SGIII binding and IC₅₀ values for DK1KD-SGIII in Raji cells (500 nmol/L) were in between those of MEC1 and JVM-3 cells (Fig. 4A and Supplementary Fig. S5). No cytotoxicity was observed in CD22(−) U937 cells. Neither incubation of purified SGIII nor DK1KD with CD22(+) or with CD22(−) cell lines increased cell death/reduced viability (Supplementary Fig. S7). Immunoblots for the distal apoptotic executioner PARP showed cleavage of PARP upon treatment with 500 nmol/L DK1KD-SGIII in all CD22(+) cell systems, whereas no processed PARP was detected in CD22(−) U937 cells (Fig. 4B). The purine analogue fludarabine (chemotherapeutic backbone of most CLL therapies) induced apoptosis (cPARP) irrespective of CD22 status.

Importantly, DK1KD-SGIII showed an advantageous high selectivity as compared with the classical cytostatic fludarabine. The pronounced cell death induction by the complete immunokinase DK1KD-SGIII in CLL (n = 4 cases), but not by DK1KD or SGIII alone (Fig. 4C), contrasted the absence of effects of such antileukemic concentrations of DK1KD-SGIII on healthy donor PBMC (3 donors; Fig. 4D), while fludarabine was cytotoxic in both cell systems. Cell survival as means ± SEM at 100 nmol/L DK1KD-SGIII: 1.075 ± 0.029 in PBMC versus 0.750 ± 0.100 in CLL; at 500 nmol/L: 1.027 ± 0.038 in PBMC versus 0.573 ± 0.076 in CLL, P = 0.005; for fludarabine at 5 μmol/L: 0.733 ± 0.061 in PBMC versus 0.395 ± 0.139 in CLL; at 25 μmol/L: 0.710 ± 0.061 in PBMC versus 0.2850 ± 0.124 in CLL (P = 0.04). This was further corroborated by exclusive DK1KD-SGIII binding (AHA assay, previous paragraph) to and killing of B cells within healthy donor PBMC (CD5⁻/CD19⁺) and CLL samples (CD5⁺/CD19⁺) while sparing T cells (CD5⁺/CD19⁻) or other non B/T leukocytes (Fig. 4E and F).

Despite highly efficient therapeutic options in CLL (26), relapsed/refractory or even upfront resistant disease remains a problem. We tested whether DK1KD-SGIII was capable of killing CLL cells ex vivo from chemoimmunotherapy-resistant disease. Incubation of freshly isolated CLL cells with either DK1KD-SGIII or fludarabine (48 hours) was followed by cell death/viability analysis by AnnexinV/7AAD flow cytometry/MTT assays. We found 13 of 15 CLL samples responding (control-corrected cytotoxicity of >20%) to DK1KD-SGIII (Table 1), providing further proof-of-concept for the DAPK1 immunoligand approach. Correlation with available clinicopathologic data including previous treatment of patients revealed similar efficacy of DK1KD-SGIII across the subsets defined by high-risk strata. Figure 5A–C show representative cases, for example, a clinically fludarabine-resistant patient (in vitro IC₅₀ not reached) in which DK1KD-SGIII (IC₅₀: 868.7 nmol/L) induced cell death dose dependently (Fig. 5C). Clinical pretreatment (Fig. 5B) or the presence of poor risk cytogenetic aberrations (i.e., del11q23, Fig. 5A) also did not influence the in vitro performance of DK1KD-SGIII.

Genuine DAPK1 regulates the execution of predominantly type I (apoptosis) and type II (autophagy) programmed cell death (PCD; refs. 4, 14, 39). Therefore, we finally investigated the mechanisms employed by DK1KD-SGIII to induce B-cell lymphoma/leukemia cell death and to overcome the elevated apoptotic threshold of CLL. Analysis of nuclear morphology of DK1KD-SGIII–treated CLL cells by fluorescence microscopy revealed a high prevalence of fragmentation characteristic for apoptotic cells (Fig. 5D). Furthermore, DK1KD-SGIII treatment resulted in time-dependent increases in AnnexinV+/7AAD− (12 hours, 24 hours) and AnnexinV+/7AAD+ (48 hours) cells as a typical apoptotic sequel. In contrast, an AnnexinV−/7AAD+ cell population, indicating a necroptotic route, was not detected. This largely excluded this type of PCD as a prominent mode of action of DK1KD-SGIII (Fig. 5E).

Immunoblot analysis for the apoptotic execution-phase protein caspase-3 revealed marked increases in proteolytically processed (cleaved) caspase-3 products upon exposure of CLL cells to DK1KD-SGIII, which (including PARP processing) was (sub)totally abrogated by the pan-caspase inhibitor ZVAD-fmk (Fig. 5F). In conjunction with the AnnexinV/7AAD data, this implicated that the DK1KD-SGIII–evoked PARP-mediated cell death was to a large part via caspase-dependent classical apoptosis. However, incubation with DK1KD-SGIII also induced the expression of the autophagy marker LC3B which is known to correlate with the number of autophagosomes (47). It indicates that, as for native DAPK1, autophagy might represent a route alternative to conventional apoptosis in mediating DK1KD-SGIII-induced viability reduction (Fig. 5F). This is particularly relevant, as this mostly p53-independent program might be responsible for the preserved efficacy of DK1KD-SGIII noted in 2 of the 3 tested CLL cases with cytogenetics (del11q23, del17p) indicating defects in classical p53-mediated apoptosis. Fittingly, the p53-mutated MEC1 cells were efficiently killed by DK1KD-SGIII as well (Fig. 4A).

We present a novel and efficient strategy to reconstitute the activity of proapoptotic DAPK1 in malignant B cells, which we confirm to lack expression of this important tumor suppressor, especially in the majority of cases of CLL. The concept was an immunoligand-based delivery of a constitutively active DAPK1 kinase domain DK1KD, which is stripped off its autoregulatory element. After rounds of receptor selection, DAPK1 was fused to SGIII, a humanized anti-CD22 scFv. This novel immunokinase (not a classical immunotoxin) DK1KD-SGIII showed superior stability, robust binding to CD22, and rapid internalization into target cells. We established a protocol for the high-yield purification of the fusion protein via FPLC-based IMAC and CEC. DK1KD-SGIII selectively induced apoptosis and autophagy in B-cell lymphoma lines and primary CLL, in the latter irrespective of clinical therapy resistance.

DAPK proteins, as best established for DAPK1, represent important integrative nodes between apoptosis and autophagic cell death pathways (1, 2). Mounting evidence implicates their dysregulation, mostly via (epigenetic) suppression, as a failed safeguarding p53-activating mechanism, hence, a central tumorigenetic step in a broad spectrum of cancers (1, 2). We hypothesized that the proleukemogenic role of DAPK1 and its tumor-associated loss would provide an actionable vulnerability and by that a strong therapeutic rationale for the unique concept of reinstatement of a tumor suppressor. Important evidence for a preserved relevance of deficient DAPK at the tumor stage derived from functional antisense-based screens (1).

Antibody–drug conjugates (ADC) hold great promise towards more targeted cancer therapies. These chimeric proteins of ever-optimized chemistries are generally composed of a classical cytotoxic moiety and a binding domain connected by a linker (48, 49). The ideally tumor-specific affinity of ADC results from an advantageous tumor-to-normal distribution of the targeted surface epitope. Stability, affinity, internalization kinetics, and cell-intrinsic processing are further determinants of ADC performance. Prominent examples of successful clinical application and marketing approval are conventional immunotoxins, that is, brentuximab-vedotin (anti-CD30 mAb with monomethyl auristatin E, for refractory or relapsed Hodgkin lymphoma and systemic anaplastic large cell lymphoma), gemtuzumab-ozogamicin (anti-CD33 mAb with a calicheamicin derivative, for relapsed AML), or trastuzumab-emtansine (anti-HER2 mAb with DM1, for inoperable/metastasized HER2⁺ breast cancer; ref. 50). Other immunoligands such as those with bispecific antibody constructs or natural ligands are designed to attract components of the immune system to harness their antitumor potential (51). Our novel immunokinase format (not an immunotoxin) has already shown promising activity in Hodgkin lymphoma xenografts as a DAPK2′-CD30L construct (24).

In the process of ligand-epitope selection, crucial in determining the fate of the whole construct (52), we arrived at anti-CD22 based on several of its advantages over other B-cell–specific mAb or scFv with available sequence information. First, the binding of the humanized anti-CD22 scFV SGIII was consistent and stable (e.g., in contrast to the anti-CD19 mAb HD37) and it did not induce apoptosis as compared with the CD40 antibody fragment G28-5. Furthermore, unlike the CLL-associated targets for “naked” antibodies CD20 or CD52, the transmembrane receptor CD22 is rapidly internalized (42, 45). CD22 is a member of the SIGLEC family of lectins and emerges as one of the most intensely pursued B-cell–specific targets in formats of immunoligands (e.g., inotuzumab ozogamicin) or CAR T-cell–directed therapies (45). Finally, we had chosen here an anti-CD22 scFv version over a mAb construct for various reasons of immediate consequence or for subsequent clinical application: easier cloning and recombinant production, facilitated endocytosis as well as lower immunogenicity, and lack of unspecific uptake by Fc receptor carrying cells, both due to the missing Fc region. Although CD22 regulates apoptotic signaling (45), engagement with our SGIII component without the DK1KD part did not noticeably influence survival.

It remains speculative to which degree the levels of DAPK1 expression (its tumor- or B-cell-of-origin associated loss; Fig. 1 and Supplementary Fig. S1; Supplementary Tables S1A and S2) affect the response towards DK1KD-SGIII. Likely, the constitutively active DK1KD overrides the regulated (stress-dependent) operation of natural DAPK1; hence (residual), DAPK1 expression would not pose major restrictions towards DK1KD-SGIII efficacy. In fact, the specifics of lymphoma-associated dysregulated survival pathways and high (oncogenic) stress levels would provide some selectivity. We postulate that the levels of CD22 expression (a key factor for binding and internalization) more profoundly determine the effect of our construct. Fittingly, there was a trend of higher CD22 expression with higher sensitivity to DK1KD-SGIII. The near-complete correlation of SGIII scFv binding with the signals from the original anti-CD22 RFB-4 antibody implicates the reliability of this standard diagnostic read-out in the screening for tumors suitable for targeting by DK1KD-SGIII. We consider a certain degree of toxicity towards normal CD22(+) B cells (Fig. 4F) as an acceptable trade-off.

The stages of development of this novel immunokinase included thorough evaluations of functionality. The DK1KD component showed robust catalytic activity, the DK1KD-SGIII fusion protein was stable, and it revealed specific binding to CD22(+) cells followed by a rapid endocytosis. As expected, on the basis of our experiments of genetic DK1KD overexpression, the active immunoligand DK1KD-SGIII induced marked cell death in the 5 CD22(+) lymphoma cell lines (2 Burkitt, 1 FL/DLBCL, 2 CLL-derived; IC₅₀s ranging from 170 to 1,300 nmol/L) and in 13 of 15 freshly isolated primary CLL samples (IC₅₀s ranging from 275 to 875 nmol/L). No cytotoxicity was observed in the non-B cells of healthy donor or CLL samples, most likely as they are CD22(−). The fusion protein also bound the CD22(+) B cells of healthy donors, but further experiments need to show how differentially DK1KD-SGIII acts in death induction of normal B cells versus lymphoma/CLL cells.

Seminal work demonstrated that silencing of the tumor suppressor DAPK1 is a key feature of familial and sporadic CLL, implicating it in early leukemogenesis (16, 17). We confirm here the aberrant expression pattern at the protein level. Our studies of protein expression and in silico GEP meta-analysis implicate that DAPK1 levels are reduced in normal B cells as well (Fig. 1 and Supplementary Fig. S1; Supplementary Table S1A). This is in accordance with previous findings of marked DAPK promoter hypermethylation in a higher percentage of normal B cells than in non-B lineage cells (36). The differential size of DAPK1 hypermethylated subsets within the B-cell fraction of healthy individuals defined by IgM expression (36) or otherwise (Supplementary Fig. S1) indicates that there might be a predestined DAPK-deficient precursor compartment of particular lymphomagenic susceptibility (53).

In CLL, characterized by a genetically and environmentally instructed high apoptotic threshold, it is desirable to therapeutically circumvent conventional und p53-dependent forms of PCD. In fact, an attractive feature of DK1KD-SGIII is that it not only induces classical caspase- and PARP-mediated apoptosis, but also triggers the expression of autophagic LC3B in CLL cells. The latter clearly distinguishes the immunokinase from fludarabine (Fig. 5F). As caspase inhibition was associated with near-complete abrogation of DK1KD-SGIII–induced PARP cleavage without affecting LC3B levels, further studies have to investigate whether the construct induces actual autophagic cell death and whether this is PARP independent, or if an autophagic self-preservative program is activated in a subset of CLL cells. This is of particular interest as fludarabine resistance had been associated with autophagy addiction (54).

Although showing low to intermediate efficacy in some cell lines, further legitimate hope for a clinically meaningful advantageous profile of cell death induction by DK1KD-SGIII is based on our observation that it is equiefficacious across CLL of distinct clinical risk features. Importantly, the immunokinase sustained its high activity in chemoimmunotherapy refractory CLL and required a low IC₅₀ in p53 mutated and fludarabine resistant (55) MEC1 cells. Studies on larger cohorts of risk-stratified CLL need to solidify these preliminary findings.

Overall, in a proof-of-principle, the strategy of reinstating B-cell lymphoma associated lost tumor-suppressive DAPK1 by delivering its constitutive kinase domain via a CD22-specific immunoligand is highlighted by encouraging biotechnologic feasibility, superior selectivity, and notable in vitro efficacy. The cytotoxicity of this prototype will have to be increased in subsequent steps of preclinical optimization, for example, including evaluations of valency (up to 10-fold increase; ref. 56), introduction of a translocation-driving protein domain (up to 20-fold increase; ref. 57) and removal of the 6xHis-tag, or using SGIII as a diabody (9- to 48-fold increase; ref. 56). DK1KD-SGIII carries high potential for intensified preclinical testing. Subject of such additional studies should be the detailed modes of cell death execution, immunogenicity, the role of cell-intrinsic mechanisms of immunotoxin resistance (58), or the in vivo performance of species-adapted constructs to assess the penetration to sanctuary sites and milieu compartments. We envision such optimized immunokinase variants to broaden the armamentarium of antilymphoma/CLL therapies in notoriously hard-to-treat cases ideally in the context of low residual tumor burden.

No potential conflicts of interest were disclosed.

Conception and design: N. Lilienthal, M.K. Tur, S. Barth, M. Herling

Development of methodology: N. Lilienthal, S. Barth

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Lilienthal, G. Lohmann, S. Zittrich, P. Mayer, C.D. Herling

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Lilienthal, G. Lohmann, G. Crispatzu, E. Vasyutina, C.D. Herling, M. Herling

Writing, review, and/or revision of the manuscript: N. Lilienthal, G. Lohmann, E. Vasyutina, S. Zittrich, M. Hallek, S. Barth, M. Herling

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Zittrich, P. Mayer, M.K. Tur, M. Hallek, G. Pfitzer, M. Herling

Study supervision: P. Mayer

Other (purification of protein): G. Pfitzer

The authors thank R.J. Ford, M. Binder, and L. Frenzel for provision of cell lines and D. Beutner for human tonsils. DNA for the α-CD19 antibody HD37 was a kind gift from Prof. M. Little (Affimed), DNA for the α-CD22 scFv SGIII was a kind gift from Dr. M. Arndt (National Center for Tumor Diseases, University Hospital of Heidelberg, Heidelberg, Germany), and DNA for the α-CD40 scFv G28-5 was a kind gift from Dr. Tanja de Gruijl (Division of Immunotherapy, Department Medical Oncology, Vrije Universiteit Medical Center, Amsterdam, the Netherlands).

This work was supported by The German Cancer Aid Max-Eder award (to M. Herling), the German Research Foundation (DFG; HE-3553/3-1, to M. Herling; and SCHW-1711/1-1, to C.D. Herling) as part of the collaborative research group KFO-286, by a grant from the CLL Global Research Foundation (to M. Herling), the German Jose-Carreras leukemia foundation (DJCLS R 12/08), and by the local CECAD initiative (to M. Herling). N. Lilienthal received a stipend by a joint pharmacology graduate program between the University of Cologne and the Bayer Health Care AG. E. Vasyutina was supported by the local Köln-Fortune program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.