Results

A CD62L-specific scFv was derived from the antibody clone 145/15. Its sequence was fused to either the MV H protein or the NiV G protein, to the latter via a short and a long [(G4S)3] linker (L3). All three constructs were equally well expressed at the surface of transfected HEK-293T cells (Suppl. Fig. 1). For production of CD62L-targeted LVs, HEK-293T producer cells were transfected with two envelope plasmids (one encoding MV H or NiV G fused to the targeting moiety, the other the fusion protein MV F or NiV F), the lentiviral packaging plasmid and the transfer vector encoding gfp. Vector particles harvested as unconcentrated supernatant were used for transduction of target cells (HT1080CD62L and HT1080αHis) and non-target cells (HT1080). Notably, HT1080αHis cells are applicable target cells due to the presence of a His-tag at the C-terminal part of the CD62L-scFv fused to the NiV G and MV H protein.
While MV‑L362L-LV was highly active in transducing both target cell types, both NiV glycoprotein-based LVs were inefficient in gene delivery, especially on HT1080CD62L cells (Suppl. Fig. 2). Hence, MV‑L362L-LV (hereafter called 62L-LV) was chosen for further investigation. For all following experiments, a second generation αCD19-CAR covering the 4-1BB costimulatory domain and the CD3ζ-signaling domain as well as a truncated LNGFR (ΔLNGFR) reporter protein was packaged into LV particles. This vector was produced at large scale, purified and concentrated over a sucrose cushion. Vector stocks contained 2.6 – 7.9x1011 particles/mL, which were on average 142 ± 7 nm in size (Fig. 1A). Transduction efficiency was even higher on primary human PBMC than on HT1080αHis cells (Fig. 1B). CAR gene delivery was strictly dependent on CD62L expression, since 62L-LV transduced HT1080CD62L cells, but not the parental HT1080 cells (Fig. 1C).
Transduction of activated primary human PBMC obtained from various donors resulted in efficient gene transfer into CD4+and CD8+ T lymphocytes (Fig. 2A). Gene transfer rates were substantially enhanced by the addition of Vectofusin-1, resulting in more than 70% CD4+CAR+ T cells and 50% CD8+CAR+ T cells. They were thus comparable to those obtained with VSV-LV (Fig. 2A and Suppl. Fig. 3). Interestingly, even after cultivation of these cells for several days, CAR T cells generated with 62L-LV contained significantly higher numbers of less differentiated cells than CAR T cells generated with VSV-LV, as indicated by the higher percentage of CD62L+ cells (Fig. 2B).
To demonstrate specificity of 62L-LV on primary human PBMC, a blocking experiment with the parental CD62L antibody (145/15) or an unrelated antibody against CD45 was performed. Incubation of activated PBMC with increasing concentrations of either anti-CD62L or anti-CD45 resulted in a gradual increase in cell staining intensity for both antibodies (Fig. 3A). CD62L staining peaked at a concentration of 2.2 ng/mL while for anti-CD45 saturation was only about to be reached for the highest concentration applied, although 100% of the cells were positive for CD45 also at lower antibody concentrations (Fig. 3A and Suppl. Fig. 4). Addition of 62L-LV vector particles to antibody pre-incubated cells showed that 62L-LV particle binding to cells decreased with increasing concentrations of anti-CD62L, while the unrelated antibody CD45 did not influence vector binding (Fig. 3B). Notably, vector binding onto PBMC could be reduced close to background levels already at an anti-CD62L concentration of 2.2 ng/mL, demonstrating that 62L-LV binds specifically to CD62L on primary cells.
During T cell activation and differentiation, CD62L is shed from the T cell surface. This has two consequences. First, CD62L levels on T cells are strongly fluctuating in cell culture. It is therefore difficult to correlate CAR gene and CD62L expression to prove the selectivity of 62L-LV after transduction of primary human PBMC. Second, shed CD62L (sCD62L) may bind to vector particles and reduce their gene transfer activity. Whether sCD62L hinders transduction by sequestering vector particles was subsequently analyzed in a binding experiment. As expected, accumulation of sCD62L in the supernatant of activated PBMC was observed for up to 10 days (Fig. 4A). Supernatant from day six, containing on average 64 ng/mL sCD62L, was used to pre-incubate 62L-LV particles prior to T cell binding. Interestingly, pre-incubation of 62L-LV with either fresh or frozen supernatants containing sCD62L did not influence binding of the vector particles to PBMC. Similar staining intensities of the reporter protein were detected regardless whether vector particles were incubated with sCD62L containing supernatants or fresh medium, indicating that sCD62L molecules present in cell culture supernatants did not alter binding of 62L-LV to T cells (Fig. 4B). Along this line, pre-incubation of vector and sCD62L did not impact the transduction efficiency of 62L-LV particles (Suppl. Fig. 5).
After having demonstrated that 62L-LV can specifically transfer CAR genes into the genome of CD62L-positive T cells, we investigated the functionality of those CAR T cells in an in vivo setting compared to CAR T cells generated with a conventional VSV-G pseudotyped LV. CAR T cells were short term generated within 3 days by 24h incubation of activated PBMC with either 62L-LV or VSV-LV and subsequently administered to NSG mice via tail vein injection. Notably, a higher amount of vector bound T cells were present in the infused product of 62L-LV-treated T cells compared to VSV-LV-treated ones, refelecting that 1.3-fold more vector particles were applied in the 62L-LV group (Suppl. Fig. 6). Untransduced PBMC served as control. To demonstrate their functionality, Nalm6 cells (luciferase-encoding CD19-positive target cells) were intravenously injected into the mice three days later and tumor growth was followed by bioluminescence imaging (BLI). A schematic time line of the experimental set-up is presented in Figure 5A. Tumor growth was clearly constrained in both vector groups, while a steady increase of tumor mass, reflected by a more than 100-fold increase in luciferase signal, was observed in all control animals (Fig. 5B). Quantification of signals revealed that tumor load in both vector groups remained at or slightly above the background level over all days of analysis (Fig. 5C). Notably, signals in animals that received VSV-LV-treated T cells were slightly reduced compared to those receiving 62L-LV-treated cells, but this difference was not significant. At day 17 post-adoptive cell transfer, no tumor cells were detected in blood, bone marrow, liver and spleen of the sacrificed mice of both vector groups, while tumor cells were present in various organs of all control animals (Fig. 5D). Along this line, proliferation of CAR T cells was observed in blood over time of animals having received 62L-LV- or VSV-LV-incubated PBMC (Fig. 6A). Interestingly, a higher proportion of CAR T cells and human CD45+ cells in general was observed for the VSV-LV group in spleen, blood and bone marrow at day 17 (Fig. 6B-D). Similar trend was observed for liver as well (Fig. 6E). These finding might reflect the fact that a nearly 6.5-fold increased MOI was used for PBMC transduction in the VSV-LV group. In conclusion, in the applied animal model functional CAR T cells can be generated with 62L-LV by short termex vivo exposure to vector particles, which are similarly potent as VSV-LV derived CAR T cells.