Discussion
For any CAR T cell therapy, generating a product with high safety and
efficacy in terms of longevity, engraftment and antitumor-effector
function is the ultimate goal. CAR T cell design and cellular
composition of the CAR T cell product are essential parameters defining
these key therapeutic features. Parameters affecting CAR T cell function
are e.g. the choice of costimulation, ratio between
CD4+ and CD8+ CAR T cells, CAR T
cell differentiation status and amount of exhausted CAR T cells. This
paper describes a novel gene transfer vector, termed 62L-LV, which
specifically transduces CD62L-positive cells, thus offering the
potential to preferentially generate CD62L+ CAR T
cells without the need of preselection of defined T cell subsets.
Importantly, the newly generated 62L-LV vector could be robustly
produced with regard to particle size, concentration and functional
titer. With an average size of 142 nm and 1011particles/mL, size and concentrations of 62L-LV stocks lay in the
expected ranges of previously established RT-LVs. Functional titers of
concentrated 62L-LV batches encoding the αCD19-CAR were on average above
1x106 t.u./mL on HT1080αHis cells and
about one log higher on PBMC. This difference in gene transfer activity
illustrates that titer determination depends on the particular
experimental conditions including the cell type, used transgene and
transduction condition. Functional titers can therefore not be compared
to those of other vector types. Gene transfer into primary human PBMC
with 62L-LV was as efficient as with VSV-LV using same amount of vector
volume (Suppl. Fig. 3), yet still resulted in a significantly higher
proportion of less differentiated CAR T cells upon long term
cultivation.
As CD62L is a T cell differentiation marker, its expression changes
throughout T cell life time and activation status. CD62L is regulated by
transcriptional shutdown of the CD62L gene as well as shedding from the
cell surface upon T cell activation. Therefore, direct proof for the
selectivity of 62L-LV on primary cells is difficult, since transduced
CD62L+ cells might have turned negative for CD62L when
gene expression becomes detectable. Yet, T cells transduced with 62L-LV
contained significantly higher proportions of CD62L+CAR T cells than those generated with VSV-LV. While this already
indicated that CD62L was used as entry receptor, we provide further
evidence for its selectivity from transduction of engineered CD62L
expressing cells as well as vector particle binding assays to primary T
cells. Binding of 62L-LV to primary T lymphocytes was specifically
blocked by the parental CD62L-specific antibody from which the targeting
domain of 62L-LV was derived.
An interesting finding of our study was that 62L-LV particles were not
blocked by shed CD62L. This was not expected, as it is known that
binding capacities of CD62L to target molecules are retained after
cleavage. Various reasons might be causative for this finding. The
concentration of sCD62L in cell culture supernatants were lower (50
ng/mL) than those in serum of healthy individuals (0.8 – 2.3 µg/mL). In
addition, sCD62L is known to aggregate which further reduces the amounts
of molecules available for binding of 62L-LV. Even more relevant, it has
been suggested that sCD62L is conformationally different from the
membrane-associated full-length CD62L as a monoclonal antibody directed
against an epitope in the EGF-like domain of CD62L was able to bind to
the cell-surface associated CD62L but not the soluble form. The same may
hold true for the 145/15 antibody. As a consequence, 62L-LV would be
specific for CD62L but not the conformational different sCD62L.
Regardless of the exact mechanism, we have proven that 62L-LV transduces
T lymphocytes also in presence of sCD62L.
Currently approved CAR T cell products available on the US and EU market
include Yescarta, Kymriah, Abecma, Tecartus and Breyanzi. Genetic
modification of those products is accomplished by transduction with
VSV-LV or γ-retroviral vectors. According to information provided on the
companies’ homepages, between 2-5 weeks are required for CAR T cell
production and release. To reduce production times, shorter T cell
cultivation and expansion could be beneficial. In this regard, we show
here that CAR T cells generated within three days of ex vivohandling control the tumor burden in a mouse model. This result is well
in line with the previous observation of Ghassemi and colleagues, who
have shown that functional CAR T cells cannot be only generated within
three days, but also outperform conventionally generated CAR T cells in
xenogeneic mouse tumor models. In difference to the published results,
we stimulated our cells with IL-7 and IL-15 instead of IL-2, activated
the PBMC for only two days with αCD3 and αCD28 and administered the
cells already 24 hours after vector incubation. Additionally, using
62L-LV less differentiated T cells were directly targeted. It remains to
be analyzed in more sophisticated mouse models whether targeting these
cells provides a survival-advantage over VSV-LV incubated CAR T cells.
While shortening the manufacturing time for CAR T cells appears feasible
and desirable, certain safety concerns arise with this procedure. During
conventional CAR T cell manufacturing, transduced cells undergo several
washing and expansion steps reducing the amounts of residual vector
particles to negligible concentrations. In contrast, it can be assumed
that particle uptake and gene transfer are not completed for CAR T cell
products injected as early as 24 or 48 hours after vector incubation.
Vector particles still bound to the T cells may transduce other cells
upon infusion. This risk is expected to be higher for VSV-G pseudotyped
vectors that have a broad cell tropism. That such a scenario can be
fatal was demonstrated 2018 in a clinical trial investigating the CAR T
cell product Kymriah. In this trial, an accidental transfer of a
CD19-CAR into a single leukemic cell during manufacturing led to relapse
and death of a patient. Causative for this event was that a CAR
construct present in tumor cells can bind in cis to the CAR specific
epitope on the surface of the tumor cell, in this case CD19, mask the
epitope from recognition by CAR T cells, conferring resistance to the
CAR T cell product and enabling its proliferation. In contrast to
VSV-LV, RT-LVs, like 62L-LV, have a more restricted tropism which is
controlled by the specificity of the used targeting domain.
Theoretically, 62L-LV might bind and transduce all kinds of
CD62L+ cells, like B lymphocytes, neutrophils,
monocytes, eosinophils, hematopoietic progenitor cells, immature
thymocytes and a subset of NK cells. Yet, it has to be assessed whether
those cells will be modified by 62L-LV in vivo and what
consequence the potential CAR expression in these CD62L-positive cell
types could have. Importantly, it has to be ensured that tumor cells do
not express CD62L. In order to reduce this potential safety concern, the
exact time-point of completed transduction after short-term incubation
should be investigated and additional washing steps could be implemented
to remove residual particles from the cells prior to adoptive transfer.
Taken together, the newly established 62L-LV has shown great potential
for the generation of less differentiated CAR T cells without the need
of prior or later T cell subtype selection. It is thus a suitable
alternative to VSV-G pseudotyped LV vectors. One immediate application
may be its use for short-term generated CAR T cells within few days,
which may substantially simplify CAR T cell production. Although
promising, this approach will need further investigation with regard to
safety concerns and scalibility of receptor targeted lentiviral vector
production before implemented into clinical studies.