CD30 on extracellular vesicles from malignant Hodgkin cells supports damaging of CD30 ligand-expressing bystander cells with Brentuximab-Vedotin, in vitro

Abstract

The goal of targeted immunotherapy in cancer is to damage both malignant and tumor-supporting cells of the microenvironment but spare unaffected tissue. The malignant cells in classical Hodgkin lymphoma (cHL) selectively express CD30. They release this receptor on extracellular vesicles (EVs) for the tumor-supporting communication with CD30 ligand (CD30L)-positive bystander cells. Here, we investigated how CD30-positive EVs influence the efficacy of the CD30 antibody drug conjugate (ADC) Brentuximab Vedotin (SGN-35). The malignant cells and the EVs expressed the active sheddase ADAM10. ADAM10 cleaved and released the CD30 ectodomain (sCD30), causing a gradual depletion of SGN-35 binding sites on EVs and creating a soluble competitor of the ADC therapy. In a 3D semi-solid tumor microenvironment model, the EVs were retained in the matrix whereas sCD30 penetrated readily into the surrounding culture medium. This resulted in a lowered ratio of EV-associated CD30 (CD30EV) to sCD30 in the surrounding medium in comparison to non-embedded cultures. A low percentage of CD30EV was also detected in the plasma of cHL patients, supporting the clinical relevance of the model. The adherence of CD30EV but not sCD30 to CD30-/CD30L+ mast cells and eosinophils allowed the indirect binding of SGN-35. Moreover, SGN-35 damaged CD30-negative cells, provided they were loaded with CD30+ EVs.

Keywords: ADAM10; Brentuximab-Vedotin; CD30; Hodgkin lymphoma; extracellular vesicle.

Figure 1. ADAM10 activity on EVs from Hodgkin lymphoma cell lines

(A) Indicated cells (4 × 10⁶/mL) were cultivated for 2 h in serum-free medium. Under these conditions they showed no loss of viability. The cell supernatant was precleared by a sequence of centrifugation steps before it was subjected to ultracentrifugation for 2 h at 100000 × g. The pellet was suspended in 1 ml of PBS and tested by nanoparticle tracking analysis (NTA). The graph shows an overlay of 5 independent determinations of the L540 EVs. The mean diameters of EVs from all tested cell lines are shown in the summarizing graph. (B) The EVs from 8 × 10⁷ KM-H2, L1236 or L540 cells were immobilized at 4.5 μm-microspheres. Aliquots of the microspheres were incubated with ADAM10 antibody (red line) or isotype control (filled histogram). The beads were labeled with fluorescence-labeled anti-mouse IgG and evaluated by flow cytometry. (C) ADAM10 was determined by Western blotting in Triton X-100 lysate from 1 × 10⁵ L540 cells (cells) or EVs from 8 × 10⁷ L540, KM-H2 (KM) or L1236 (L12). (D and E) Aliquots of EVs from 4 × 10⁷ cells or 200 ng of recombinant ADAM10 or ADAM17 were suspended in 25 mM Tris-HCl, pH 8 containing 6 × 10⁻⁴ % Brij-35 in the presence or absence of 3 μM BB3644 or 3 μM GI254023X. Then, the aliquots were incubated with the fluorescent substrates PEPDAB010 (D) or PEPMCA001 (E) (BioZyme Inc., Apex, NC) in black microtiter plates at 37°C. Fluorescence was determined in a kinetic study at 530 nm (D) or 393 nm (E) as indicated. The data show means of two independent experiments minus background fluorescence without EVs.

Figure 2. Ectodomain shedding on Hodgkin cell-derived EVs

Isolated EVs from the supernatants of 8 × 10⁷ L1236 and KM-H2 cells were cultivated in EV-depleted medium for 18 h at 37°C ± BB3644 (3 μM, red line and red/shaded bar). Then, EVs were again centrifuged at 100000 × g. (A) The pellet was washed once in PBS and immobilized at 4.5 μm-microspheres. Aliquots of microspheres were incubated with Ki-2 mAb (CD30ecto), Ki-12 mAb (CD30endo) or isotype control (filled histogram). Aliquots for CD30endo determination were permeabilized with 0.1% Triton X-100 before the addition of the antibodies (TX perm). Samples were labeled with fluorescence-labeled anti-mouse IgG for flow cytometry. (B) After ultracentrifugation, the pellets and the supernatants were evaluated by the commercial ELISA (CD30ecto) and the CD30endo ELISA as indicated. The mean fluorescence intensity (MFIs) of 4–5 independent experiments were evaluated by arbitrarily setting the inhibited aliquots as 100%. The percentages of the non-inhibited aliquots were statistically evaluated by a two-tailed, nonparametric t-test (Mann-Whitney) (ns = not significant, * = P < 0.05, **P < 0.01).

Figure 3. Release of CD30 in 3D microenvironment model

(A, B) L540 cells (2 × 10⁵) were embedded in a 24-well plate in 100 μL of a semi-solid gel containing equal volumes of growth factor-reduced matrigel and RPMI-1640 with 20% EV-depleted FCS ± BB3644 (3 μM) (embedded). After hardening of the matrix, 900 μL of culture medium containing 10% of EV-depleted FCS (± 3 μM BB3644) was added. As control, cells were cultivated in suspension on top of 100 μL of cell-free semi-solid matrigel (suspended). After 24 h, the culture supernatants, without matrigel, were removed and pre-centrifugated to remove cells and cell debris. Then, the supernatant was ultracentrifugated at 100000 × g for 120 min to sediment the EVs (•). The ultracentrifugation supernatant was isolated and the EV pellet was suspended in PBS and adjusted to the same volume as the supernatant. (A) The CD30 ectodomain ELISA was used to determine CD30EV in the PBS-suspended EVs and sCD30 in the ultracentrifugation supernatant. The results show U/mL as means ± SE for four independent experiments. From these experiments the percentage of CD30EV was calculated (% of total released CD30). (B) EVs were immobilized at 4.5 μm-microspheres and aliquots of microspheres were incubated with antibodies as indicated and investigated by flow cytometry. The mean fluorescence intensity (MFI, black graph) was determined and compared with the BB3644-inhibited aliquots (red tinted histograms). The inhibited samples were arbitrarily set as 100%. (C) The percentage of CD30EV (% of total released CD30) was determined the plasma of cHL patients (N = 6).

Figure 4. CD30-positive EVs target SGN-35 to cells of the microenvironment

(A) CD30L-DsRed2-transfected HMC 1.1 (red) and CD30-eGFP-transfected HD-MyZ (green) were co-cultured in growth factor-reduced matrigel and incubated for 2 h at 37°C, 5% CO2. Two consecutive confocal images (Δ = 1.219 μm) of co-cultivated cells are shown. Arrowheads indicate release and binding of CD30⁺ vesicles to CD30L⁺ HMC 1.1 cells. The circle indicates internalized CD30. (B) Confocal image of a tissue section of a lymph node infiltrated by cHL of mixed cellular subtype was stained with NASDCL (red) and with a CD30 primary antibody (Ber-H2) and an ALEXA488-conjugated secondary antibody (green). Bars indicate 30 μm. Confocal images were taken with laser scanning microscopy (Zeiss Meta 510, Zeiss, Germany) using a 40× oil objective with NA 1.3 and the appropriate filters and analyzed with Imaris 7.0 software. (C) L540 cells (5 ml of 2 × 10⁶/mL) were cultivated for 3 h at 37°C in serum-free medium with biotin-labeled SGN-35 (1 μg/mL). Supernatants were harvested, precentrifuged and EVs were finally pelleted at 100000 × g. The pull-down of SGN-35 on pelleted EVs was investigated by Western Blot under reducing conditions. Streptavidin-coupled peroxidase was used to detect the heavy and light chain of the biotinylated SGN-35 (EVs). Biotinylated SGN-35, directly applied to the Western Blot served as loading control. (D) Determination of CD30L on HMC1.2 cells by flow cytometry. HMC1.2 cells (5 × 10⁵/mL) were incubated for 1 h on ice with CD30Fc or an anti-CD30L antibody (left) or with the indicated amounts of sCD30, CD30EV or without CD30 (tinted curve) in the presence of 0.1 μg/mL of FITC-labeled SGN-35.

Figure 5. SGN-35 damaged CD30L⁺ immune cells through CD30EV

EOL-1 cells (5 × 10⁵/mL) were cultivated for 96 h with SGN-35 (0, 1 or 5 μg/mL) and ± 1000 U/mL sCD30 or 1000 U/mL CD30 on EVs (CD30EV) from L540 cells. Cells were stained with annexin V-coupled ALEXA647 and propidium iodide (PI). They were analyzed by flow cytometry and double-positive cells were gated (Q2). The percentage of cells/gate is indicated. (A) The image shows one representive experiment of four. (B) The mean percentage in Q2 ± SD of four independent experiments is depicted for some conditions as indicated. The significance was determined by two-tailed, non-parametric, Mann-Whitney U test (* = P < 0.05, > 0.01).

Figure 6. Proposed model for the role of EVs and CD30 shedding for the immunotargeting with SGN-35

The malignant cells in cHL reside in lymphoid tissue, surrounded by a microenvironment of extracellular matrix and proinflammatory cells. They selectively express the receptor CD30. The CD30 antibody-drug conjugate SGN-35 binds to the CD30⁺ tumor cells, is internalized and the toxic compound is cleaved and activated by lysosomal proteases. The malignant cells not only expose CD30 on the surface, they also release CD30 on EVs (CD30EV), either by membrane blebbing or release of exosomes from multivesicular endosomes. EVs also bind SGN-35 and SGN-35-loaded EVs migrate away from the cancer cell. The loading of EVs might also occur within the tumor cell. After SGN-35 internalization, the drug might target to multivesicular endosomes instead of lysosomes. Apoptotic blebs of damaged tumor cells might contribute to the release of SGN-35-loaded EVs. Mast cells and eosinophils support the cHL tumor growth. These cells express the natural CD30 ligand and bind SGN-35-loaded CD30EV. These cells might internalize the EVs and are damaged by SGN-35 in a CD30EV-dependend manner. Cells and EVs express the CD30 sheddase ADAM10, which gradually cleaves the ectodomain. In the microenvironment, sCD30 is quickly drained whereas EVs are retained. Monomeric sCD30 is a competitor of SGN35 binding to cells and EVs. This hypothesis suggests an elevated ratio of membrane-associated CD30 within the matrix and an elevated ratio of sCD30 outside of the matrix. SGN-35 exploits this mechanism to preferentially target cancer and bystander cells in the tumor microenvironment. The CD30-depleted EVs and the high level of competing sCD30 in the circulation might explain the minute side-effects of the drug.