Administration of B7-H3 targeted chimeric antigen receptor-T cells induce regression of glioblastoma

Dear Editor, Nowadays, glioblastoma (GBM) was the most common and lethal form of primary intracranial tumor. Despite standard-of-care therapy, GBM still exhibited a poor prognosis with 5-years survival rate less than 5%. Recent years, adoptive CAR-T therapy came to be a novel immunotherapy in treating malignant tumors. Great progress has been made by CD19 targeted CAR-T cells against refractory B cell cancers. Recent studies also reported about the clinical potential of CAR-T therapy targeting IL13Ra2 and EGFRvIII in treating GBM. However, limited numbers of therapeutic targets in GBM may preclude it from progress and being popularized. B7-H3 (CD276) has been found to be overexpressed by many tumors and tumor-infiltrating dendritic cell. Our previous studies suggested the potent anti-tumor effect of B7-H3 targeted CAR-T cells against GBM in preclinical models. Here we presented our clinical experience with one patient to evaluate the therapeutic potential of B7-H3 targeted CAR T-cell therapy in treating recurrent GBM. In this case, a 56-year-old woman presented with recurrent GBM in the left frontal and parietal lobe (Supplementary Fig. 1). The patient has received twice craniotomy and standard-of-care with chemoradiation in the last 2 years. Pathologic study of tumor resection showed 50% expression of Ki67 and a high but heterogeneous B7-H3 expression, with a histochemistry score evaluated as 110 (0–300) (Supplementary Fig. 2a). Flow cytometry assay of tumor primary cells also confirmed the high B7-H3 expression (Supplementary Fig. 2b). In the preclinical study, we identified the specific tumor-lysis ability of autologous B7-H3 targeted CAR-T cells. The structure of B7-H3-targeted CAR was shown in Supplementary Fig. 2c. Flow cytometry results indicated CAR-T cells displayed memory T cell markers (CD45RO and CD62L), and had relative low levels of or were negative for effector T-cell markers (CD69 and CD25) and PD-1/Tim-3 (Supplementary Fig. 2d). In a real-time monitoring of cytotoxicity assay, B7-H3 targeted CAR-T cells induced specific anti-tumor effect in tumor primary cells (Supplementary Fig. 2e). Enzyme-linked immunosorbent assay (ELISA) results also indicated an activation effect of the CAR-T cells when cocultured with tumor primary cells (Supplementary Fig. 2f). Three weeks after the craniotomy, tumor recurrence was found in the surgical resection site by magnetic resonance imaging (MRI). The patient received weekly intracavitary infusions of B7-H3 targeted CAR-T cells. The first two round infusion was following a dose-escalating principle (Fig. 1a). The CAR-T cells was delivered by an Ommaya device (Supplementary Fig. 1b). After the first-round infusion, we observed a dramatic reduction of recurrent tumor by MRI. Remarkably, the enhanced part of the recurrent tumor was significantly reduced, compared to the signal before infusion (Fig. 1b). The clinical response was sustained for about 50 days after the initiation of CAR-T cells infusion. Unfortunately, this patient appeared in drowsiness and altered consciousness in cycle 6 and 7 and MRI revealed tumor recurrence. Finally, the patient dropped out of the clinical study after the 7 cycles infusion. Although there were no toxic effects of grade 3 or higher associated with the CAR-T cells infusion, the patient suffered from headache during cycles 1–5, which could not be completely alleviated by giving oral analgesic therapy. The headache first appeared at 3 h after cycle 1 infusion. The symptom was more obvious and repeated attack in the follow-up infusion. Treatment was paused for three days until the remission of headache in cycle 2. We thus maintained a dose of 1 × 10 in the next few infusions for security. Remarkably, the lasting time of headache seemingly correlated to the infusion dose of CAR-T cells at the first-round treatment (Supplementary Table 1). In the last 3 cycles, this symptom was less obvious, in spite of higher doses of infusion (dose: 1.5 × 10 and 2 × 10). For evaluating the physical condition of the patient, multiple serum biochemical indexes were continuedly monitored during the treatment and results revealed no significant changes before and after local administration of the CAR-T cells (Supplementary Table 2). After intracranial administration of CAR-T cells, evaluation of nucleated cells indicated a significant expansion of T cells in cerebrospinal fluid (CSF) samples obtained from the infusion device, especially in cycle 3 (Fig. 1c). Moreover, cell count of CSF sample collected from lumbar puncture on day 3 of cycle 3 show the existence of CAR-T cells and expansion of T cells (Supplementary Fig. 3). Further, 16 inflammatory cytokines were measured for evaluating immunologic changes in CSF and periphery blood before and after each cycle infusion. As a result, levels of 10 cytokines increased by a factor of more than 5 from pre-infusion baseline levels in CSF, and the cytokines level decreased between weekly treatment cycles. Of interest, IL2 and IL6 especially IL6 level increased significantly in periphery blood (increased by a factor of more than 5), though the extent was less obvious than that in CSF (Fig. 1d). The measured levels for 10 cytokines were provided in Supplementary Table S3. In this subject, although B7-H3 targeted CAR-T cells mediated a short-term anti-tumor response in situ. However, the tumor became resistant to the therapy despite higher doses of CAR-T cells in later cycles. One of the possible reasons for tumor resistance was target antigen heterogeneity. Pre-therapy IHC result indicated a heterogenous expression in tumor specimen obtained before infusion. Analysis of CSF showed that the expansion of T cells was limited in the later cycles. Combined with the inflammation cytokines changes result, we supposed that CAR-T cells were not capable to eliminate all the tumor cells completely, especially B7-H3 tumor cells. These tumor cells resisted to the therapy and relapsed. Since the patient dropped out of the clinical study, the post-therapy analysis was limited. Yet, such phenomenon of antigen heterogeneity was also detected in CAR-T therapy of GBM targeting EGFRvIII and IL13Ra2. The heterogeneous expression of the two


Manufacture of B7-H3 targeted CAR-T cells
The CAR sequence was codon optimized and contains a human CD8α leader peptide, an anti-B7-H3 single chain variable fragment (scFv), a human CD8α hinge domain, a human CD8α transmembrane domain and the cytoplasmic domain of 4-1BB/CD3ζ. A truncated CD19 (CD19t) used for CAR detection was separated by a P2A ribosome skip sequence from the CAR sequence. The anti-B7-H3 scFv sequence mentioned above were derived from a highly specific monoclonal antibody (mAb) against B7-H3 (clone: mAb-J42) generated by our group using standard hybridoma technique.
For lentivirally transduced CAR-T cells manufacturing, peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation using Lymphoprep (Greiner Bio-One) on the day of venous blood collection. PBMC were then incubated with anti-CD25 and anti-CD45RA microbeads (Miltenyi Biotec) for negative selection to remove naïve T and Tregs (since these Tregs are known to express high levels of CD25/IL-2 receptor) according to the operation instrutions. For T cell expansion, the cells were cultured in TexMACS GMP Medium (Miltenyi Biotec) and stimulated with GMP grade OKT3 (CD3 mAb: 600ng/mL, Novoprotein) and anti-CD28 antibody (CD28 mAb: 300ng/mL, Novoprotein) with addition of recombinant human (rh) IL2 (100units/mL, PeproTech) and rh-IL15 (5ng/mL, Miltenyi Biotec) in a 37°C, 5% CO2 environment. Two days after PBMC isolating, activated T cells were transduced with lentivirus (MOI=1) by recombinant fibronectin fragment (CH-296, Novoprotein) in the presence of rh-IL2 and rh-IL15. T cells were harvested for cryopreservation after culturing 8-10 days. For cryopreservation, CAR-T cells were harvested, washed in phosphate buffer solution (PBS) containing 2% HSA, and resuspended in Serum-Free Cell Freezing Medium (BioLife Solutions). The whole CAR-T cell manufacturing process was completed within about 2 weeks. Quality control result of the manufactured B7-H3 targeted CAR-T cells were shown supplementary Table  4.

Infusion process of CAR-T cells
Participant was pre-medicated with acetaminophen P.O. (per os) and diphenhydramine I.V. (intravenous injection) or P.O. 30 minutes before CAR-T cell infusion according to a protocol of glioblastoma CAR-T cell therapy [1]. CAR-T cell infusions were administered manually via Ommaya device in a 1.0 mL volume over approximately 10 minutes using a 21-gauge butterfly needle, followed by a 1.0 mL normal saline flush over 5 minutes. After CAR-T cell infusion, research participants were required for being monitored for 3 hours.

Clinical Imaging
MRI scans of the cerebral post-gadolinium T1-weighted sequences were acquired on a 3.0 T clinical MRI instrument (Siemens Trio Tim). The CT scans of the Ommaya device were acquired on a Philips Brilliance 16 row helical scanner. Regions of contrast-enhancing recurrent tumor and Ommaya device were confirmed by a radiologist. All the scans were analyzed on a Siemens Syngo imaging system.

Immunohistochemistry analysis
The resected tumor specimens were collected from the Neurosurgery Department of West China Hospital. For all IHC analysis, the samples were performed on a pathological section of paraffinembedded specimens. The tumor regions were confirmed by a clinical neuropathologist. All of samples were dealt with 10% formalin and embedded by paraffin, dried for 90 minutes at 65℃, then blocked with distilled water containing 3% H2O2 and PBS containing 10% normal goat serum (Boster) at room temperature in turn. Slices were stained with corresponding mono-antibody at 4℃ overnight. For B7-H3 expression analysis, anti-B7-H3 mAb (clone: D9M2L, Cell Signaling Technology, CST) was used to calculate for defined tumor areas by DAB detection system (ZSGB-Bio).

Flow cytometry analysis
For all flow cytometry analysis, cells were washed twice with PBS containing 0.5% bovine serum albumin (BSA) and incubated with corresponding antibodies for 30 minutes at 4℃ in the dark. BD LSRFortessaTM (BD Biosciences) instrument was used to analyze the cells and percentages of positive cells were calculated via FlowJo-V10 software for analysis. For the analysis of CSF, all samples were stored and transported at cryogenic temperatures, then processed with ACK buffer (Sigma) and stained with the corresponding antibody. For CAR-T cell detection shown in Figure S3, anti-CD3 and anti-CD19 specific fluorescein-conjugated antibody (Biolegend) was used. The total cell count was measured by Counter-Star instrument after lysis of red blood cells using ACK buffer.

Multiple cytokines detection
Peripheral blood samples were collected in tube without EDTA and CSF samples were obtained from the reservoir of Ommaya delivery device. All samples were centrifuged at 3000 g for 10 minutes, and the supernatants were collected and frozen at −80°C. Cytokines profile was measured using a MILLIPLEX multiplex immunoassay (Millipore) and analyzed with a Luminex FlexMap 3D system (Luminex Corporation).

CAR expression analysis by quantitative real-time reverse transcriptase polymerase chain reaction (RT-PCR)
Total RNA was prepared from the nucleated cell obtained from CSF using an Eastep® Super Total RNA Extraction kit (Promega). The purity was confirmed with an A260/A280 ratio greater than 2.0. The complementary DNA (cDNA) was synthesized using a GoScript TM Reverse Transcription Mix (Promega). RT-PCR was performed by using the Bio-Rad CFX96 TM Real-Time PCR system and SsoFast EvaGreen® Supermix Mix (Bio-Rad). The reaction solution consisted of 2 µL of the template, 10 µL of the Supermix Mix (Bio-Rad), 1 µL of upstream/downstream primer (10 µM), and 6 µL of distilled water. The PCR conditions used were Enzyme activation for 30 sec at 95°C, followed by 40 PCR cycles of denaturation at 95°C for 5 s, annealing/extension at 58°C for 5 sec. CAR expression levels were normalized to that of β-actin in each sample using the ΔΔCT method.

Real-Time Monitoring of Cytotoxicity (RTCA)
The cytotoxic ability of B7-H3 targeted CAR-T cells was determined using the xCELLigence® real-time cells analyzer (ACEA Bioscience, Inc. xCELLigence RTCA SP). To start the real-time cell analysis, background readings was obtained from 100 μL of media added in each well of the E-plate® 96. In order for cell attachemnt, the primary tumor cells (10 4 cells per well) were cultured in E-plate 96 (ACEA Bioscience) for about 15 hours before CAR-T cells were added (E:T=5:1). Three replicates were available for each well. Cell index measurements were performed at 15 min intervals for 72 h. The data were acquired and analyzed using the manufacturers protocols (ACEA Bioscience, Inc. RTCA Software 2.1). The expression levels of CAR were normalized to that of β-Actin in each sample using the ΔΔCT method. The sequences of the primers used to detect β-Actin mRNA and CAR mRNA were as follows: β-Actin forward primer (5'-GGACCTGACTGACTACCTCAT-3'), β-Actin reverse primer (5'-CGTAGCACAGCTTCTCCTTAAT-3'), CAR forward primer(5'-GAAGCCTCTGGATTCACTTT-3'), CAR reverse primer(5'-TAACCATCCCAATGTCTTGC -3').

Enzyme-linked immunosorbent assay (ELISA)
CAR-T cells (1×10 5 ) were co-cultured with primary tumor cells (5×10 4 ) in 48-well plates without addition of exogenous cytokines. 12 hours after coculture, supernatants were collected, and cytokines (IFNγ and IL2) were measured using corresponding ELISA kits (Thermo-Fisher Scientific) following manufacturer's instructions. In this assay, student's t test was used for single comparisons. The experiments were repeated twice. A P value less than 0.05 was considered significant.