Adhesion of T Cells to Endothelial Cells Facilitates Blinatumomab-Associated Neurologic Adverse Events

Neurologic adverse events (NAE) are common side effects of therapies that exploit activated T cells to destroy malignant B cells (1–6). Blinatumomab is a CD19/CD3-bispecific T-cell engager (BiTE) immuno-oncology therapy that activates patients' own T cells to target CD19⁺ B cells, leading to T-cell proliferation and B-cell apoptosis (7–9). Blinatumomab treatment has shown clinical efficacy in patients with both relapsed/refractory (r/r) B-cell precursor acute lymphoblastic leukemia (ALL; ref. 2) and indolent as well as aggressive r/r non-Hodgkin lymphoma (NHL; refs. 10, 11). Blinatumomab-associated grade ≥3 NAEs include encephalopathy, headache, altered state of consciousness, aphasia, ataxia, confusional state, nervous system disorder, tremor, neurotoxicity, and seizure (1, 3, 12). Most NAEs occur early during the first treatment cycle (i.e., 12–48 hours after infusion start or dose step) and are transient and fully reversible (1). Grade ≥3 NAEs occurring in patients with r/r ALL receiving blinatumomab are managed by withholding blinatumomab until resolution of NAEs to grade ≤1 and then restarting at the initial (lower) blinatumomab dose (13), as higher doses of blinatumomab correlate with a higher incidence of NAEs (10). In patients with r/r NHL, dose steps [i.e., starting at a low dose and then increasing to multiple higher dose(s); refs. 10, 11], prophylactic steroids (10, 11), and pentosan polysulfate (PPS) coadministration (10) have been explored as potential ways to reduce the incidence or severity of blinatumomab-associated NAEs. Dose steps have the potential to limit the efficacy of blinatumomab in some patients as may high-dose steroids such as dexamethasone; however, a clear benefit of these two approaches to mitigate NAEs has yet to be proven (11). The effect of PPS, an antiadhesive agent, on blinatumomab-associated NAEs was evaluated in a small subset of patients (n = 3) with r/r NHL at high risk of NAEs; no severe NAEs were observed in these patients who received PPS during start of blinatumomab infusion and again during dose step (10). It was hypothesized that PPS may interfere with T-cell migration into the central nervous system (CNS) and mitigate blinatumomab-associated NAEs. The purpose of this study is to determine the mechanism(s) of blinatumomab-associated NAEs in patients with r/r ALL and r/r NHL, using a combination of in vitro T-cell rolling experiments and in vivo pharmacodynamic/pharmacokinetic assays.

Data from five clinical studies with blinatumomab were analyzed: a phase I study and a phase II study in adults with r/r NHL (NCT00274742, NCT01741792; refs. 10, 11), a phase Ib/2 study in pediatric patients with r/r ALL (NCT01471782; ref. 14), and two phase II studies in adults with r/r ALL (NCT01209286, NCT01466179; refs. 1, 13). Individual study designs and primary results have been described in detail elsewhere (1, 10, 11, 13, 14) and are briefly summarized in Supplementary Information (Supplementary Table S1). Different target doses of blinatumomab were used for ALL and NHL, with dose levels of 15 μg/m²/day (equivalent to 28 μg/day for an average adult) for ALL and 60 μg/m²/day (equivalent to 112 μg/day) for NHL. To mitigate adverse events, especially cytokine release syndrome (CRS), step dosing of blinatumomab is recommended (15). In addition to the clinical trials, a pediatric patient case from compassionate use of blinatumomab is described (16). The patients selected for specific analyses are summarized in Table 1.

Study protocols were approved by the independent ethics committee of each institution, and all patients provided written informed consent. Clinical studies were conducted in accordance with the provisions of the Declaration of Helsinki and with Good Clinical Practice guidelines.

Lymphocyte analysis was performed as described previously (17, 18). Briefly, peripheral blood mononuclear cells (PBMC) were isolated from blood samples of patients with r/r NHL (NCT00274742). PBMC samples were collected prior to start of infusion (day 1), and at 45 minutes, 2, 6, 24, 30, 48, and 168 hours during continuous intravenous (cIV) infusion of blinatumomab (week 1); and following dose step(s) [day 8 (and day 15)] at the same time intervals [week 2 (and week 3)]. Cerebrospinal fluid (CSF) samples were taken from an adult with r/r NHL (NCT00274742) on day 2 after treatment discontinuation due to NAEs and a compassionate-use pediatric patient with r/r ALL on day 15 and day 43 of cIV infusion. PBMC and CSF samples were analyzed by flow cytometry on a FACSCanto II system (BD Biosciences). T and B cells were stained by dye-labeled antibodies against CD3 (BD Biosciences) and CD19 (Agilent), respectively. T-cell adhesiveness was assessed by binding of a human intercellular adhesion molecule (ICAM)-1/Fc chimera protein (Bio-Techne) to the intermediate-affinity conformation of lymphocyte function-associated antigen (LFA)-1 on T cells and subsequent detection by dye-labeled antibodies against human Fc (Dianova). Absolute numbers of lymphocytes, monocytes, and platelets were determined by differential blood counts.

At time points described above (leukocyte analysis), plasma concentrations of angiopoietin (Ang)-2, an endothelial activation biomarker, were measured for 10 patients with r/r NHL (NCT00274742) using the Human Angiopoietin-2 Quantikine ELISA Kit (Bio-Techne) according to the manufacturer's instructions. Briefly, 100 μL of assay diluent followed by 50 μL of standard, control, or prediluted sample were added to each well of a precoated microplate. After incubation at room temperature for 2 hours and subsequent four-time washing, 200 μL of conjugate were added. The plate was incubated at room temperature for 2 hours and washing steps were repeated before 200 μL of substrate were added. After light-protected incubation at room temperature for 30 minutes, 50 μL of stop solution were added. Finally, the plate was read within 30 minutes at 450/570 nm on a PowerWaveX Select instrument (BioTek).

IHC was performed with a biotin/avidin/peroxidase technique as previously described in detail (19). Paraffin blocks of formaldehyde-fixed brain autopsy tissues from frontal cortex and white matter, cerebellum, and hippocampus were available for analysis. Serial paraffin sections were stained by IHC with human-specific antibodies against T-cell markers CD3 (Thermo Fisher Scientific) and CD8 (Agilent).

Blinatumomab serum and CSF concentrations were quantified using a validated T-cell activation assay as described previously (14, 17). Briefly, Raji cells (human B-cell line) were incubated with HPB-ALL cells (human T-cell line) in the presence of serial blinatumomab dilutions in serum ranging from 200 ng/mL to 3 pg/mL. A blank sample was included to measure background signal. To avoid any matrix effects of CSF on the assay, purified rat anti-blinatumomab antibodies were spiked into one half of the CSF sample to simulate a predose (i.e., blank) sample whereas the other half remained unspiked. After incubation, T cells were labeled with mouse anti-human CD69 IgG1 FITC (BD Biosciences) and analyzed on a FACSCanto II system (BD Biosciences). The activation marker CD69 was expressed on T cells in a blinatumomab concentration-dependent manner.

Because of multiple correlating factors potentially leading to the development of NAEs, data of blinatumomab-treated patients with r/r NHL or r/r ALL from various studies (NCT00274742, NCT01741792, NCT01209286, NCT01466179; n = 188 at data cutoff) or of the r/r ALL subset (n = 92 at data cutoff) were pooled and subjected to multivariate analyses to identify any associated or predictive risk factors. Covariates assessed in the regression analysis were sex, age, B/T-cell ratio, disease severity, steroid pretreatment, maximum dose received, cytokine levels, and laboratory values (e.g., monocytes).

Human brain microvascular endothelial cells (HBMEC) were purchased from ScienCell Research Laboratories (#1000) and human umbilical vein endothelial cells (HUVEC) were purchased from PromoCell (#C-12200). Both cell lines were confirmed to be negative for Mycoplasma by Venor GeM Advance Mycoplasma Detection Kit, and were not analyzed further for validation. Culturing of HBMECs or HUVECs under constant unidirectional flow conditions was done using the Ibidi Pump System (Ibidi) according to the manufacturer's instructions until passage 11. Briefly, HBMECs or HUVECs were seeded into μ-slide I^0.4 Luer (Ibidi; Collagen IV or ibiTreat, respectively) and cultured at 37°C, 5% CO₂ for 48 hours in RPMI1640 medium (Biochrom; supplemented with Nu-Serum IV, FBS, Minimal Essential Medium, l-glutamine, sodium pyruvate, heparin, Endothelial Cell Growth Supplement) at shear stress of 5 dyn/cm², or in endothelial cell growth medium (PromoCell; supplemented with FBS) at 10 dyn/cm², respectively. Preincubation of endothelial cells with 200 μg/mL PPS SP54 (bene-Arzneimittel) was initiated 24 hours after start of cell culturing under constant unidirectional flow conditions.

PBMCs were isolated from fresh blood of healthy donors as described previously (17, 18). Subsequently, T cells were purified from PBMCs using the human Pan T Cell Isolation Kit (Miltenyi Biotec) according to the manufacturer's instructions. Preincubation of T cells with 50 μg/mL minocycline (Triax Pharmaceuticals) or with 1 μg/mL natalizumab (Elan Pharma International) was done at 37°C, 5% CO₂ for 2 hours in DPBS or for 10 minutes in RPMI1640, respectively. T cells were resuspended in RPMI1640 medium before rolling experiments.

T-cell interactions with endothelial cells under hydrodynamic conditions were visualized using an in vitro flow chamber system. This system consisted of the pump system connected to a μ-slide I^0.4 Luer containing confluently grown endothelial cells, a heating system, and a CO₂ gas incubation unit (all Ibidi). This allowed application of a constant unidirectional flow at physiologic conditions (37°C, 5% CO₂) within the channel of the μ-slide ensuring viability of both T cells and endothelial cells during long-term (≤2 hours) assays. Rolling experiments utilized HBMECs, whereas firm adhesion was only measurable to HUVECs. T-cell rolling and adhesion were closely monitored using a microscopic system consisting of an inverse microscope (Nikon), a digital camera (Hamamatsu Photonics), and an imaging software (NIS-Elements AR; Nikon). Experiments were run at a T-cell concentration of 1 × 10⁶/mL ±10 ng/mL blinatumomab, and additionally ±200 μg/mL PPS, ±50 μg/mL minocycline, or ±1 μg/mL natalizumab at shear stress of 1 dyn/cm² for up to 2 hours. Individual T-cell rolling velocities and absolute numbers of adhering T cells were determined by automatic or manual tracking of single T cells at 0 and 45 minutes of continuous rolling on endothelial cells using the NIS-Elements AR software. Subsequently, the mean T-cell rolling velocity (±SD) of trackable T cells was calculated for each experimental condition.

After T-cell rolling on endothelial cells, μ-slides were fixed with 4% paraformaldehyde solution (Sigma-Aldrich) and blocked with an avidin/biotin blocking reagent (Dianova) prior to immunostaining of P-selectin, ICAM-1, and vascular cell adhesion molecule (VCAM)-1. P-selectin surface expression on endothelial cells was visualized by staining with a mouse anti-human P-selectin IgG1 (15 μg/mL; Bio-Techne) and goat anti-mouse IgG secondary antibody, Alexa Fluor 488 (1:100; Thermo Fisher Scientific). Staining of ICAM-1 was done with rabbit anti-human ICAM-1 IgG, biotin (10 μg/mL; Abcam) and streptavidin, Cy3 (1:100; Dianova). VCAM-1 expression was assessed with rabbit anti-human VCAM-1 IgG (5 μg/mL; Abcam) and goat anti-rabbit IgG secondary antibody, DyLight 350 (20 μg/mL; Thermo Fisher Scientific). Stained HBMECs were analyzed by fluorescence microscopy using ultraviolet light, and image acquisition with the NIS-Elements AR software.

Three patients with r/r NHL (NCT00274742) and at high risk of developing NAEs due to their low B/T-cell ratios (Table 1) received PPS coadministration at 100 mg by bolus intravenous infusion 1 to 3 hours before start of blinatumomab infusion and dose step followed by intravenous perfusion of 300 mg/day for 48 hours thereafter (10). Three additional patients with r/r NHL from the same phase I study with similar baseline characteristics and a comparable blinatumomab step-dosing regimen but no PPS coadministration were selected as a comparator group based on the absence of peripheral B cells at start of infusion or dose step, which is a prerequisite to demonstrate an effect of PPS on T-cell redistribution kinetics.

Statistical interpretation of in vitro T-cell rolling velocity data was performed by one-way ANOVA combined with Tukey post hoc test using the Prism software (GraphPad Software). A P <0.05 was regarded as statistically significant.

Qualified researchers may request data from Amgen clinical studies. Complete details are available at the following link: http://www.amgen.com/datasharing

The pharmacodynamic effects of blinatumomab, more extensively analyzed at start of infusion in a subset of 10 patients with r/r NHL (NCT00274742) were consistent with previous studies (Fig. 1; refs. 18, 20). Infusion of blinatumomab and modulation of its steady-state serum concentration (i.e., dose step) induced rapid T-cell redistribution characterized by T cells disappearing from peripheral blood within 2 to 6 hours, followed by their subsequent return to baseline levels within the following 7 days (Fig. 1A). Coincident with T-cell disappearance, the adhesion molecule LFA-1 on T cells shifted from a low to an intermediate affinity conformation (21) as evidenced by its increased binding to an ICAM-1/Fc chimera protein (Fig. 1B). Because this increased T-cell adhesiveness suggested interactions with blood vessel endothelium, we further investigated the release of Ang-2 by endothelial cells, a marker of endothelial cell activation. Ang-2 is stored in Weibel–Palade bodies within endothelial cells and is released into peripheral blood upon activation after interactions with T cells (22). Consistent with T-cell redistribution kinetics, Ang-2 serum concentrations increased markedly within 6 hours after start of infusion before slowly returning to baseline within the following 7 days (Fig. 1C). Moreover, monocytes (Fig. 1D) and platelets displayed redistribution patterns similar to T cells, although they are not directly engaged by blinatumomab. Effects on redistribution and endothelial activation similar to those shown in Fig. 1A–D were observed at dose step(s) and in patients with minimal residual disease (MRD) or r/r ALL.

Histopathologic examination of brain sections and analyses of CSF from patients treated with blinatumomab identified T cells at the brain microvascular endothelium and in CSF (Fig. 2); moreover, blinatumomab was detectable in CSF.

Histopathology of brain sections revealed CD3⁺ and CD8⁺ T cells at the luminal surface of capillary endothelium (Fig. 2A) and in the underlying perivascular or leptomeningeal space of a 24-year-old patient with r/r ALL (NCT01209286) who was treated at reduced dose of 5 μg/m²/day due to recurring NAEs in cycles 1 and 2. Blinatumomab infusion was permanently discontinued due to infection in cycle 3, and the patient died 4 days later due to a brain stem infarction caused by a fungal plaque.

A 17-year-old compassionate-use patient with r/r ALL was treated at a dose of 15 μg/m²/day for 6 weeks without evidence of NAEs or an open blood–CSF barrier, as monitored by CSF albumin concentration (day 15 and 43). Blinatumomab concentration in CSF was measured and below the limit of detection throughout treatment. At day 15 (under cIV infusion), a low level of predominantly CD8⁺ T cells was found in CSF (Fig. 2B); no B or natural killer cells were observed in CSF. At day 43 (immediately after end of infusion), a marked increase in T-cell counts throughout all T-cell subpopulations was detected (Fig. 2B).

A 55-year-old patient with r/r NHL (NCT00274742) who was treated at a dose of 60 μg/m²/day and permanently discontinued blinatumomab infusion at day 2 of an additional treatment cycle due to NAEs had quantifiable blinatumomab of 20 pg/mL in CSF. FACS analysis of CSF taken immediately after infusion termination revealed not only an increased white blood cell count of 6/μL, but also the presence of CD3⁺ T cells and CD19⁺ B cells or B-cell debris (Fig. 2C).

Blinatumomab concentration in CSF was also systematically evaluated in children and adolescents with r/r ALL treated with the standard 5 to 15 μg/m²/day dose-step regimen (NCT01471782). At day 15 (under cIV infusion), a mean (±SD) blinatumomab concentration in CSF of 18.2 (±26.2) pg/mL was measured; the mean (±SD) CSF/serum concentration ratio was 0.036 (±0.061). Although most of these patients presented with none to only mild disturbance of their blood–CSF barrier, a maximal blinatumomab concentration in CSF of 94.0 pg/mL was detected in a patient with an open blood–CSF barrier.

The multivariate analyses identified the ratio of B cells to T cells in peripheral blood before the start of infusion (B/T-cell ratio) as the most consistent risk factor for developing NAEs. In the full analysis set (n = 188), 74% (57/77) of patients with a B/T-cell ratio <1/8 developed NAEs throughout blinatumomab treatment compared with 51% (56/111) of patients with a B/T-cell ratio ≥1/8 (P = 0.0001; Supplementary Table S2). In the r/r ALL subset, 82% (23/28) of patients with a B/T-cell ratio <1/8 had NAEs compared with 52% (33/64) of patients with a B/T-cell ratio ≥1/8 (P = 0.006; Supplementary Table S2). A low B/T-cell ratio seems to be associated with a higher risk of developing NAEs early during the first treatment cycle, whereas in subsequent cycles, the B/T-cell ratio seems to have less significance. Interestingly, the incidence rate of grade ≥3 NAEs was lower in patients with MRD ALL (13%; ref. 23) compared with patients with r/r NHL (30%; ref. 10), despite very low peripheral B-cell counts before start of infusion in patients with MRD ALL (18).

Leukocyte extravasation is a complex multistep cascade of events including initial rolling on, subsequent firm adhesion to, and final transmigration through the blood vessel endothelial cell layer (24). We established an in vitro flow chamber system mimicking capillary hydrodynamic conditions, which allowed us to measure blinatumomab-mediated effects on T-cell rolling and adhesion to HBMECs and HUVECs, respectively (in the absence of target B cells). Freshly isolated T cells from healthy volunteers were circulated for 45 minutes on a monolayer of flow-cultivated HBMECs and T-cell rolling velocities were measured. No difference in mean (±SD) T-cell rolling velocities was seen between the beginning [344 (±58) μm/second] and end of the flow experiment [323 (±78) μm/second; Fig. 3A]. When blinatumomab (10 ng/mL) was added to the flow chamber system, the mean (±SD) T-cell rolling velocity after 45 minutes of continuous circulation was significantly reduced to 209 (±40) μm/second (∼1.5-fold reduction; Fig. 3A), suggesting increased T-cell interactions with HBMECs. However, this effect was not immediate [342 (±69) μm/second at t = 0 minute; Fig. 3A]. Moreover, the addition of blinatumomab also increased the absolute number of T cells adhering to HUVECs after 45 minutes of continuous circulation compared with the adhesion of T cells to HUVECs in the absence of blinatumomab. In addition, immunofluorescence analysis of HBMECs after T-cell rolling in the presence of blinatumomab showed pronounced upregulation of endothelial adhesion molecules: ICAM-1, P-selectin, and VCAM-1 (Fig. 3B). Thus, blinatumomab addition to the in vitro flow chamber system reduced T-cell rolling velocity and increased T-cell adhesion, coinciding with upregulation of relevant adhesion molecules on HBMECs.

We analyzed the potential of substances with antiadhesive properties to reverse the effects of blinatumomab on T-cell rolling and adhesion. HBMECs were preincubated with or without the semisynthetic heparinoid PPS, an inhibitor of P-selectin expressed on the luminal surface of endothelial cells (25). Addition of PPS to the rolling experiments increased the mean (±SD) T-cell rolling velocity of freshly isolated T cells from 399 (±153) μm/second to 515 (±159) μm/second (Fig. 4A). Moreover, addition of PPS in the presence of blinatumomab resulted in a mean (±SD) T-cell rolling velocity of 465 (±116) μm/second (Fig. 4A), which was comparable with the mean T-cell rolling velocity without blinatumomab. In line with these observations, immunofluorescence staining of HBMECs after T-cell rolling showed that PPS treatment markedly decreased P-selectin signals compared with untreated cells with or without blinatumomab (Fig. 4B).

Another agent with potential antiadhesive properties is the tetracycline antibiotic minocycline, which interferes with LFA-1 binding to ICAM-1 by chelating Ca²⁺ ions and by downregulating LFA-1 on T cells (26). T cells preincubated with minocycline had a significantly higher rolling velocity than untreated cells, demonstrating minocycline as an effective antiadhesive agent (Fig. 4C). As before, addition of blinatumomab markedly reduced the mean (±SD) T-cell rolling velocity [127 (±41) μm/second]; however, this attenuated effect was blocked when T cells were preincubated with minocycline [217 (±92) μm/second; Fig. 4C]. Addition of minocycline also reversed the blinatumomab-induced increase of firm T-cell adhesion to HUVECs (Fig. 4D). A third compound with antiadhesive properties is the humanized mAb natalizumab, which binds to very late antigen (VLA)-4 on T cells and blocks VLA-4 interaction with VCAM-1 on endothelial cells (27). Results comparable with PPS and minocycline were obtained (Fig. 4E). In summary, these three distinct antiadhesive agents prevented blinatumomab-induced reduction of mean T-cell rolling velocities and adhesion to vascular endothelial cells in vitro.

The initial T-cell kinetics differed for the 3 patients who received PPS compared with those who did not (Fig. 5). In contrast to the usually measured 20% to 60% decline in T-cell counts 45 minutes after start of blinatumomab infusion or dose step in the absence of PPS (Fig. 5, top), T-cell counts remained at or above baseline levels at 45 minutes in samples from the 3 patients treated with PPS (Fig. 5, bottom). T-cell redistribution was delayed in patients who received PPS. T-cell redistribution started 2 to 6 hours after start of infusion or dose step and was prolonged relative to patients without PPS coadministration. Interestingly, only one of the 3 patients who received PPS coadministration had moderate NAEs 48 hours after the blinatumomab dose step (when the PPS IV perfusion had been stopped) despite their low B/T-cell ratios (Table 1). These observations are consistent with PPS transiently preventing T cells from entering the perivascular space (PVS), where they could be activated by (rare) target B cells and trigger NAEs through the release of cytokines and other proinflammatory factors.

The clinical and nonclinical data reported here build on previous observations in patients with r/r NHL or r/r ALL and suggest that NAEs are linked to blinatumomab-induced T-cell activity in the CNS. Changes of blinatumomab steady-state serum concentrations, such as at start of infusion and dose step, were associated with T-cell redistribution characterized by rapid T-cell disappearance from peripheral blood and subsequent gradual return to baseline levels within 1 week. T-cell redistribution coincided with a conformational shift of LFA-1 on T cells from a low to an intermediate-affinity state, as well as an increase in the serum concentration of the endothelial activation marker Ang-2 (22). Although not directly engaged by blinatumomab, monocyte and platelet redistribution were also observed.

In a single-patient case study, T-cell adhesion to the luminal surface of brain microvascular endothelium was observed followed by potential T-cell transmigration into the PVS through an intact blood–CSF barrier. These observations suggest that blinatumomab might facilitate T-cell influx into CSF despite an intact blood–CSF barrier. One can hypothesize that massive leukocyte adhesion to blood vessel endothelial cells (including brain microvascular endothelium) might result in transient disturbance of microcirculation and subsequent transient focal cerebral hypoxia (28), potentially contributing to the reversible clinical symptoms of blinatumomab-associated NAEs. Blinatumomab was detected in CSF of several patients with r/r ALL and of a patient with r/r NHL. In addition, B-cell debris and an influx of (effector) T cells were detected in CSF of two patients immediately after end of infusion; in one case, infusion was terminated due to a grade ≥3 NAE. However, we were unable to see a direct relationship between CSF levels of blinatumomab and neurotoxicity due to small sample size and low concentrations of blinatumomab. Similar to our observations with blinatumomab, higher numbers of CD19 CAR T cells were seen in CSF of patients who experienced NAEs compared with those who did not (29). Finally, a low B/T-cell ratio (i.e., a low B-cell count in peripheral blood and bone marrow) has been identified as a potential risk factor for the development of NAEs in patients with r/r NHL or r/r ALL. The lower incidence of NAEs in children and adolescents with r/r ALL may be explained by the higher number of both peripheral B cells and bone marrow blasts due to a required blast count of >25% before start of blinatumomab treatment. In contrast, the respective inclusion criterion for adults with r/r ALL was >5% blasts in bone marrow. As patients with MRD ALL had frequently received standard intrathecal chemotherapy prophylaxis before and throughout blinatumomab treatment, which might have reduced or even eliminated residual target B cells in the brain, the risk of developing severe NAEs may also have been reduced irrespective of the B/T-cell ratio. In addition, patients with MRD ALL received blinatumomab as first- or second-line treatment, whereas patients with r/r NHL may have had increased residual target B cells in the brain due to their advanced disease state.

On the basis of our pharmacodynamic and clinical observations, we propose the following model for the development of blinatumomab-associated NAEs: (i) start of blinatumomab infusion or dose steps increase the adhesiveness of circulating T cells to blood vessel endothelium, including brain endothelium forming the blood–CSF barrier (Fig. 6A). (ii) Adhering T cells activate the endothelium and begin to extravasate into the PVS. The activated endothelium on its part attracts other circulating leukocytes, such as monocytes (30) and platelets (Fig. 6B). (iii) It is possible that in the absence of peripheral B cells (i.e., low B/T-cell ratio) extravasated T cells first encounter (rare) target B cells in the CNS where T cells get activated by blinatumomab to secrete cytokines and chemokines triggering transient local neuroinflammation, including transmigration of monocytes (Fig. 6C). (iv) Although it is speculation, transmigration of monocytes, attracted non-T cells, and released factors into the CNS may disturb the blood–CSF barrier by enhancing transient local neuroinflammation, leading to local micro-edema in the brain, and may aggravate NAEs (Fig. 6D; ref. 31). Interestingly, some aspects of our model have also been described in the context of CD19 CAR T-cell–associated neurotoxicity. Particularly, cytokine-induced CNS endothelial cell activation was implicated in the early pathophysiology of NAEs, with the authors reporting disruption of the blood–brain barrier and cerebral edema in patients with fatal neurotoxicity (32). Of note, no consistent pattern of blood–CSF barrier disruption or post-baseline changes in CNS MRI have been observed with blinatumomab. In another study of CD19 CAR T cells, intracranial edema and CAR T-cell influx into the CSF and cerebral CRS were associated with the development of NAEs (33). However, although CD19 CAR T-cell–associated neurotoxicity may be favored by systemic CRS (32, 34), blinatumomab-associated NAEs occur independent of cytokine serum concentrations, especially in patients with r/r NHL where no CRS has been reported (10, 11). On the basis of the B/T-cell ratio, we hypothesize that the location of initial T-cell activation and subsequent T-cell–mediated cytokine release (i.e., peripheral blood vs. CSF) is more critical for the development of blinatumomab-associated NAEs than maximum systemic cytokine levels, which are relatively low compared with those observed with CAR T-cell therapies.

According to our model, blinatumomab-induced T-cell adhesion to brain microvascular endothelial cells is the first necessary but not sufficient step for the development of blinatumomab-associated NAEs. With T cells largely disappearing from circulation within 2 to 6 hours after start of infusion and dose step(s), the onset of most NAEs shortly followed this T-cell redistribution and coincided with peak T-cell adhesiveness and endothelial activation. Thus, transiently interfering with leukocyte adhesion at start of blinatumomab infusion and dose step(s) seems to be a promising approach for mitigating NAEs. Our results with the in vitro flow chamber system are consistent with blinatumomab-induced pharmacodynamic effects in vivo (e.g., T-cell redistribution and endothelial activation); however, these findings should be replicated in patient-derived T cells. These pharmacodynamic findings support the proposed mode of action of PPS, and no apparent negative impact on either safety or efficacy of blinatumomab treatment was found with PPS coadministration in a very limited patient sample. Thus, PPS coadministration with blinatumomab treatment may be a beneficial mitigation strategy in patients with low B/T-cell ratios or other NAE risk factors.

In conclusion, coadministration of antiadhesive agents with blinatumomab to mitigate NAEs is the first mechanistic-based approach to reduce the risk of developing NAEs. Mitigation of this risk would prevent treatment interruptions and/or discontinuations, allow higher (starting) doses of blinatumomab, and thus increase efficacy for patients with r/r NHL where NAEs (and not CRS) are the dose-limiting adverse events. Transient coadministration of antiadhesive agents that interfere with T-cell adhesion to blood vessel endothelium should be explored in clinical studies as a potential mitigation strategy for blinatumomab-associated NAEs to further improve the safety and efficacy of blinatumomab therapy.

M. Klinger is a principal scientist and has ownership interest (including patents) in Amgen, Inc. G. Zugmaier is an executive director at Amgen, Inc. and has ownership interest (including patents) in 9688760, 20190300609, 20150071928, 8840888, 20140228316, 20140227272, 20130323247, 20130287778, 20130287774, 20110262440, 20100112603, 7700299, 20190142846, 20070037228, 20190127465, 10130638, 20170327581, 20170122947, 9486475, 20160208001, and 9192665. V. Nägele is a senior scientist and has ownership interest (including patents) in Amgen Inc. M.-E. Goebeler reports consulting or advisory roles for ROCHE Pharma AG, Novartis, and Gemoab and received travel grants from BMS, Pfizer, Janssen Cilag, and Gilead. C. Brandl is a principal scientist at Amgen, Inc. M. Stelljes has a consulting or advisory role at Pfizer, Jazz Pharmaceuticals, Gilead Sciences, MSD, and Amgen; has received honararia from Speakers' Bureau from Pfizer, Medac, MSD, and Incyte; reports receiving research funding from Pfizer (Inst); and reports travel, accommodations, and expenses from Medac and Neovii Biotech. A. von Stackelberg is an advisory board member of Morphosys and Roche; has received speakers bureau honoraria from Amgen, Novartis, Shire, and Miltenyi; and is a consultant/advisory board member of Pfizer. R.C. Bargou has received speakers bureau honoraria from Amgen and has ownership interest (including patents) in patent blinatumomab. P. Kufer is an executive director at BiTE Technology and has ownership interest (including patents) in Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Klinger, G. Zugmaier, R.C. Bargou, P. Kufer

Development of methodology: M. Klinger, V. Nägele, C. Brandl

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Klinger, V. Nägele, M.-E. Goebeler, C. Brandl, M. Stelljes, H. Lassmann, A. von Stackelberg, R.C. Bargou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Klinger, G. Zugmaier, V. Nägele, M.-E. Goebeler, C. Brandl, H. Lassmann, R.C. Bargou, P. Kufer

Writing, review, and/or revision of the manuscript: M. Klinger, G. Zugmaier, V. Nägele, M.-E. Goebeler, C. Brandl, M. Stelljes, H. Lassmann, A. von Stackelberg, R.C. Bargou, P. Kufer

Study supervision: R.C. Bargou, P. Kufer

The authors would like to thank Prof Klaus-Michael Müller, MD, Department of Pathology, and Prof Tanja Kuhlmann, MD, Department of Neuropathology, University Hospital Münster, Münster, Germany, for performing the initial brain autopsy. Editorial support was provided by Allison Saviano and Kathleen Raulin of Sephirus Communications Inc., funded by Amgen Inc., and Beatrice Chiang and Julie Gegner, employees of Amgen Inc. The study was funded by Amgen Research (Munich) GmbH and Amgen Inc.

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.