Abstract
Allogeneic cellular immunotherapy exhibits promising efficacy for
cancer treatment, but donor cell rejection remains a major barrier.
Here, we systematically evaluate human leukocyte antigens (HLA) and
immune checkpoints PD-L1, HLA-E, and CD47 in the rejection of
allogeneic NK cells and identify CD8^+ T cells as the dominant cell
type mediating allorejection. We demonstrate that a single gene
construct that combines an shRNA that selectively interferes with HLA
class I but not HLA-E expression, a chimeric antigen receptor (CAR),
and PD-L1 or single-chain HLA-E (SCE) enables the one-step construction
of allogeneic CAR-NK cells that evade host-mediated rejection both in
vitro and in a xenograft mouse model. Furthermore, CAR-NK cells
overexpressing PD-L1 or SCE effectively kill tumor cells through the
upregulation of cytotoxic genes and reduced exhaustion and exhibit a
favorable safety profile due to the decreased production of
inflammatory cytokines involved in cytokine release syndrome. Thus, our
approach represents a promising strategy in enabling “off-the-shelf”
allogeneic cellular immunotherapies.
Subject terms: Cancer immunotherapy, Synthetic biology, Cancer therapy
__________________________________________________________________
The use of donor-derived CAR-NK cells is limited by CD8 T cell-mediated
allorejection. Here, the authors describe a one-step approach, based on
selective HLA knockdown and overexpression of PD-L1, that allows
allogeneic modified CAR-NK cells to escape rejection by the host immune
system while exhibiting enhanced anti-tumor activity and safety in
preclinical mouse models.
Introduction
The recent approvals of CD19 and BCMA targeting chimeric antigen
receptor (CAR)-T cell therapies are significant breakthroughs in the
field of cancer immunotherapy and have stimulated the development of
genetically modified cell therapies for both hematological malignancies
and solid tumors^[54]1–[55]3. To date, all six approved CAR-T cell
therapies use patients’ own (autologous) T cells^[56]2. However,
autologous CAR cell therapies have major drawbacks, including high cost
and relatively long vein-to-vein time due to individualized
manufacturing^[57]4,[58]5 and low quality and/or quantity of source
immune cells due to prolonged exposure to immunosuppressive tumor
microenvironment and multiple lines of previous cytotoxic
treatment^[59]5,[60]6. Expressing CARs in healthy donor T cells could
potentially overcome these drawbacks; however, the rejection of
allogeneic CAR cells by the host immune system and the potential to
cause graft versus host disease (GVHD) with allogeneic CAR-T cells
remain major challenges^[61]7–[62]9. Unlike T cells, natural killer
(NK) cells do not mediate GVHD, making them attractive for allogeneic
CAR approaches^[63]10–[64]13. The use of allogeneic off-the-shelf
umbilical cord blood (UCB)-derived CAR-NK cells has been very safe and
associated with promising efficacy in early clinical
trials^[65]14,[66]15. However, allogeneic CAR-NK cells can be
recognized and rapidly rejected by the recipient’s immune system.
Allo-rejection is primarily mediated by CD8^+ T cells through T cell
receptor (TCR) recognition of peptide-major histocompatibility
complexes (pMHC)^[67]16. To avoid allo-rejection by T cells, the
standard approach is to suppress MHC expression in allogeneic cells by
either CRISPR-mediated knockout or shRNA-mediated knockdown of
transporter associated with antigen processing (TAP) or
β2-microglobulin (B2M)^[68]17–[69]19, the common light chain for both
the classical MHC class I molecules, such as human leukocyte antigen
(HLA)-A, B, and C, and the nonclassical MHC class I molecules, such as
HLA-E and HLA-G^[70]20,[71]21. However, the general suppression of pMHC
expression, including HLA-E, in allogeneic NK cells makes them
susceptible to the host NK cell-mediated killing due to the lack of
NKG2A-mediated inhibition^[72]22–[73]24. Thus, additional modifications
of allogeneic NK cells are needed to overcome host NK cell-mediated
rejection.
The use of immune checkpoint inhibitors has significantly improved
cancer treatment, enabling the possibility of long-term survival of
patients with metastatic tumors^[74]25–[75]27. Recently, immune
checkpoints, including PD-L1, HLA-E, and CD47, have been overexpressed
in immune and non-immune cells to prevent their rejection and allow
longer persistence after adoptive transfer in an allogeneic
setting^[76]28–[77]32. Although PD-L1 is best known as a checkpoint for
T cells, PD-L1 expression is induced in activated NK cells and has been
associated with improved tumor control and trafficking^[78]33,[79]34.
HLA-E is the ligand for the inhibitory receptor NKG2A, which is
expressed by NK cells and a subset of CD8^+ T cells, and the
HLA-E/NKG2A axis contribute to immune escape in several
malignancies^[80]3,[81]35,[82]36. CD47 is best known as a checkpoint
for macrophages, which can mediate allo-rejection through
phagocytosis^[83]37. Whether the expression of PD-L1, HLA-E, and CD47
can inhibit allo-rejection of CAR-NK cells, especially following
knockout or knockdown of pMHC expression, has not been systematically
evaluated.
Here, we comprehensively evaluate the effects of combining the
suppression of pMHC expression with exogenous expressions of PD-L1
and/or HLA-E in overcoming the rejection of allogeneic CAR-NK cells by
both host T and NK cells. We identify a shRNA that knocks down human
HLA-A, HLA-B, and HLA-C (collectively referred to as HLA-ABC)
expression without decreasing the expression of human HLA-E, and show
that knockdown of HLA-ABC leads to a significant abrogation of host
CD8^+ T cell- and NK cell-mediated rejection. A combination of
shRNA-mediated HLA-ABC knockdown and exogenous expression of PD-L1
and/or HLA-E in NK cells dramatically reduces allo-rejection by the
host immune system in vitro and in vivo. Furthermore, we integrate the
shRNA, PD-L1, or single-chain HLA-E (SCE) and a CAR construct into the
same lentivector for one-step construction of allogeneic CAR-NK cells,
without needing multiple steps of genetic modifications. We show that
these allogeneic CAR-NK cells effectively control tumor growth, evade
allo-rejection by host T and NK cells, and exhibit a better safety
profile in vivo. Unexpectedly, exogenous PD-L1 and HLA-E expression
significantly enhances the anti-tumor activity of NK and CAR-NK cells,
likely due to reduced exhaustion and the upregulation of genes involved
in cytotoxicity and proliferation. We believe our approach could also
be applied to other allogeneic cell-based products and can aid the
design of “off-the-shelf” allogeneic therapies.
Results
Identification of a potent shRNA for specific knockdown of HLA-ABC without
affecting HLA-E expression
To develop allogeneic CAR-NK cells that can escape rejection by host T
cells, NK cells, and macrophages, we used a tiered approach
(Supplementary Fig. [84]1). We first developed shRNA to knockdown
HLA-ABC to suppress T cell-mediated rejection as well as to easily
integrate the shRNA into the same CAR lentivector for one-step
construction of allogeneic CAR-NK cells. We then evaluated expression
of HLA-E and/or PD-L1 to further suppress rejection of allogeneic NK
cells by host T and NK cells. Interestingly, human peripheral blood NK
cells expressed a high level of CD47, and the majority of them also
expressed SIRPα (Supplementary Fig. [85]2a, b). Over-expression of CD47
in K562 cells, a human lymphoblast cell line with no expression of
HLA-ABC and HLA-E, significantly inhibited killing by human NK cells
(Supplementary Fig. [86]2c, d), suggesting a significant role of
CD47-SIRPα interaction in inhibiting cytotoxicity of NK cells.
We targeted multiple genes to knock down HLA-ABC by RNA interference,
including the HLA-ABC heavy chain, B2M, and TAP1 and TAP2. To
specifically knock down HLA-ABC without affecting HLA-E, we aligned the
coding sequences of 119 most prevalent classical HLA class I alleles
across global populations (Supplementary Fig. [87]3a), including 38
HLA-A alleles, 56 HLA-B alleles, 25 HLA-C alleles, as well as two
nonclassical HLA-E alleles (E*01:01 and E*01:03), and designed shRNAs
that had no more than one mismatch with HLA-ABC alleles but had at
least two mismatches with HLA-E alleles (Supplementary Fig. [88]3b–f).
We constructed lentivectors expressing shRNAs under the control of the
U6 promoter, transduced human Jurkat T cells, and assayed HLA-ABC
expression by flow cytometry 48 h later. Among the 15 shRNAs tested,
shRNA #1 and #2 (targeting HLA-ABC heavy chain), #13 and #14 (targeting
B2M) were most potent in inhibiting surface HLA-ABC expression
(Supplementary Fig. [89]4). These four shRNAs also significantly
knocked down surface HLA-ABC expression in NK cells from four donors
(Fig. [90]1a, b). Due to their high potency, shRNA #1 and #13 were used
in the subsequent studies.
Fig. 1. HLA-ABC-reduced allogeneic NK cells evade killing by host CD8^+ T
cells.
[91]Fig. 1
[92]Open in a new tab
a, b Comparison of HLA-ABC knockdown in human NK cells by shRNA #1, #2,
#13, and #14. Shown are representative flow histograms showing
knockdown of surface HLA-ABC expression in NK cells from one donor (a)
and summary of HLA-ABC MFI on NK cells from four donors (b). Data in
(b) are mean ± s.d. (n = 4). Statistical comparisons are between the
indicated samples and the control. c shRNA #1 reduces surface HLA-A*02
expression in human NK cells. The black and red histograms show
untransduced and transduced NK cells stained with anti-HLA-A*02
antibody, respectively. The numbers indicate MFI. d, e shRNA #1 does
not inhibit surface HLA-E expression. Jurkat T cells with or without
(control) IFNγ stimulation were transduced with lentivectors expressing
either shRNA #1 or #13 and stained for HLA-E 48 h post-transduction.
Representative flow histograms (d) and summary of MFI of HLA-E (e).
Data in (e) (mean ± s.d.) shows one of three representative data in
four technical replicates. f shRNA #1 does not inhibit surface HLA-E
expression in K562 cells stably expressing SCE. Shown are
representative flow histograms comparing HLA-E levels in parental K562
cells (control) and K562 cells stably expressing SCE without (black) or
with expression of shRNA #1 (red) or #13 (green). g Experimental
protocol of T cell and NK cell killing of allogeneic NK cells. h
HLA-ABC-reduced allogeneic NK cells evade host CD8^+ T cell killing. i
HLA-ABC-reduced allogeneic NK cells are sensitive to host NK cell
killing when CD47-SIRPα axis is abrogated. Data in (h and i) are
mean ± s.d. from three independent experiments with 7 (h) or 6 (i)
independent healthy donors each in three technical replicates. Each
circle represents on donor. Two-way ANOVA with Tukey’s multiple
comparisons was performed between different groups to determine the
statistical difference. The p-values are shown for the indicated
comparisons. Figure 1g is created in Biorender.
shRNA #1 has one nucleotide mismatch with the prevalent HLA-A*02
alleles and two nucleotide mismatches with HLA-E alleles (Supplementary
Fig. [93]3). To investigate if shRNA #1 is still effective to knock
down HLA-A*02 expression, we isolated NK cells from HLA-A*02^+ donors,
transduced them with lentivector expressing shRNA #1, and assayed
HLA-A*02 expression by flow cytometry. Compared to untransduced NK
cells, HLA-A*02 expression was reduced by ~13-fold in transduced NK
cells (Fig. [94]1c), demonstrating effective knockdown of HLA-A*02
expression by shRNA #1 despite one nucleotide mismatch.
We also determined whether shRNA #1 interferes with HLA-E expression.
IFNγ treatment of Jurkat T cells significantly induced HLA-E expression
as indicated by an increase in MFI from 510 to 980 (Fig. [95]1d, e).
Expression of shRNA #1 in Jurkat T cells did not significantly reduce
the IFNγ-induced surface HLA-E expression, whereas shRNA #13, which
targets B2M, completely inhibited the IFNγ-induced surface HLA-E
expression. Alternatively, we constructed a SCE where HLA-E*01:03 heavy
chain, B2M, and the peptide from the HLA-G leader sequence were
connected by two linkers into a single polypeptide (see Supplementary
Fig. [96]7a, b). Transduction with the lentivector expressing SCE into
K562 cells led to a high level of HLA-E expression (Fig. [97]1f), which
was not reduced by shRNA #1 but reduced by shRNA #13. These results
show that shRNA #1 and #13 can potently knock down surface HLA-ABC
expression, and shRNA #1 specifically knocks down surface HLA-ABC
expression without affecting HLA-E expression.
HLA-ABC-reduced allogeneic NK cells can evade killing by host CD8^+ T cells
We investigated whether surface HLA-ABC-reduced allogeneic NK cells can
evade killing by host CD8^+ T cells and NK cells. Purified NK cells
were transduced with lentivector expressing shRNA #1 or shRNA #13, and
HLA-ABC-reduced NK cells were purified by cell sorting (based on
reduced HLA-ABC expression), expanded and used as targets for host
CD8^+ T cell and NK cell-mediated killing. CD8^+ T cells from
HLA-mismatched donors were first primed with mitomycin C-pretreated
allogeneic NK cells (with normal HLA-ABC expression) in the presence of
IL-2 for 7 days. The primed CD8^+ T cells were then cocultured with
prelabeled allogeneic NK cells with either normal or reduced HLA-ABC
expression overnight at E:T ratios of 0, 1, 10, 20, and 50, and the
lysis of allogeneic NK cells was determined by flow cytometry
(Fig. [98]1g). As shown in Fig. [99]1h, CD8^+ T cells killed allogeneic
NK cells with normal HLA-ABC expression in a dosage-dependent manner,
reaching 80% killing at E:T ratio of 50. In contrast, CD8^+ T cell
killing of allogeneic NK cells with reduced HLA-ABC expression was
reduced to ~10% at E:T ratio of 50. Primed CD4^+ T cells did not kill
allogeneic NK cells appreciably (Supplementary Fig. [100]5).
To test whether surface HLA-ABC-reduced allogeneic NK cells are
sensitive to host NK cell killing, we cocultured NK cells with
prelabeled allogeneic NK cells with either normal or reduced HLA-ABC
expression at E:T ratios of 0, 5, and 10 (Fig. [101]1g). Because NK
cells express high levels of both CD47 and SIRPα, whose interaction
blocks NK cell killing (Supplementary Fig. [102]2), anti-CD47 antibody
and human FcR blocking antibodies were added into the coculture to
inhibit CD47-SIRPα axis and antibody-dependent cellular cytotoxicity.
Allogeneic NK cells with normal HLA-ABC expression were not killed
appreciably, whereas allogeneic NK cells with reduced HLA-ABC
expression were killed in a dose-dependent manner, reaching ~40% at E:T
ratio of 10 (Fig. [103]1i). Together, these results show that the
knockdown of HLA-ABC expression in allogeneic human NK cells by both
shRNA #1 and #13 enables them to evade CD8^+ T cell killing but
sensitizes them to NK cell killing when CD47-SIRPα interaction is
blocked.
Exogenous expression of PD-L1 in allogeneic NK cells moderately inhibits host
T cell responses
We determined the effect of PD-1-PD-L1 interaction on T cell responses
to allogeneic NK cells (Fig. [104]2a). NK cells were transduced with
lentivector expressing PD-L1 and PD-L1^+ NK cells were purified by cell
sorting and expanded (Fig. [105]2b). CD3^+ T cells were isolated from
PBMC and cocultured with autologous or allogeneic NK cells with or
without PD-L1 expression at a ratio of 1:1. CD69 and CD25 expression by
T cells was measured 12 h post-coculture. Fewer than 10% of T cells
expressed CD69 when they were cocultured with autologous NK cells with
or without exogenous PD-L1 expression (Fig. [106]2c and Supplementary
Fig. [107]6a, b). Between 35% and 46% of T cells were induced to
express CD69 when they were cocultured with allogeneic NK cells without
exogenous PD-L1 expression. This percentage was reduced to ~25% when
cocultured with allogeneic NK cells with exogenous PD-L1 expression
(Fig. [108]2c). Similarly, <3% T cells expressed CD25 when cocultured
with autologous NK cells, whereas 11%–15% T cells expressed CD25 when
cocultured with allogeneic NK cells without exogenous PD-L1 expression
and the percentage was decreased to ~7% when cocultured with allogeneic
NK cells with exogenous PD-L1 expression (Fig. [109]2d). T cells did
not proliferate significantly when cocultured with autologous NK cells
with or without exogenous PD-L1 expression, whereas 10–18% T cells
underwent proliferation when cocultured with allogeneic NK cells
without PD-L1 expression and the percentage was reduced to ~7% when
cocultured with allogeneic NK cells with PD-L1 expression (Fig. [110]2e
and Supplementary Fig. [111]6c). When primed CD8^+ T cells were
cocultured with allogeneic NK cells without exogenous PD-L1 expression
at the E:T ratios of 10 and 20, ~40% and ~60% of untransduced
allogeneic NK cells were lysed, respectively, whereas lysis of
allogeneic NK cells expressing PD-L1 was reduced to ~30% and ~40% at
E:T ratios of 10 and 20, respectively (Fig. [112]2f). Together, these
results show that exogenous expression of PD-L1 on allogeneic NK cells
partially inhibits host T cell responses.
Fig. 2. Exogenous expression of PD-L1 in allogeneic NK cells moderately
inhibits host T cell responses.
[113]Fig. 2
[114]Open in a new tab
a Experimental design for assaying T cell responses to allogeneic NK
cells with or without exogenous PD-L1 expression. b PD-L1 expression by
NK cells 48 h following lentivector transduction (left panel) and 7
days following sorting and expansion (right panel, red histogram).
Black histogram: untransduced NK cells stained with anti-PD-L1. c, d
CD3^+ T cells were cocultured with autologous or allogeneic NK cells
without or with exogenous PD-L1 expression at 1 to 1 ratio. CD69 (c)
and CD25 (d) expression by T cells were measured at 12 h
post-coculture. e CD3^+ T cells were labeled with cell tracker CTV and
cocultured with autologous or allogeneic NK cells without or with
exogenous PD-L1 expression at 1 to 1 ratio for 5 days. T cell
proliferation as indicated by CTV dilution was assayed by flow
cytometry. f CD8^+ T cells were cocultured with prelabeled autologous
or allogeneic NK cells without or with exogenous PD-L1 expression
overnight at E:T ratios of 0, 10, and 20. Lysis of NK cells was
quantified. Data in (c–f) are mean ± s.d. from two independent
experiments with four healthy donors each in four technical replicates.
Each dot represents on donor. One-way (c–e) or two-way (f) ANOVA with
Tukey’s multiple comparisons was performed between different groups to
determine the statistical difference. P-values are shown as numbers in
(c–f). Figure 2a is created in Biorender.
Exogenous expression of single-chain HLA-E inhibits the killing of allogeneic
NK cells by host NK cells
Both NK cells and a fraction of CD8^+ T cells express the CD94:NKG2A
inhibitory receptor, which recognizes HLA-E as ligand^[115]35,[116]38.
We investigated whether expression of HLA-E by allogenic NK cells
inhibits host NK cell responses. Hanson et al. has reported that
single-chain HLA class I that incorporates HLA class I heavy chain,
B2M, and the presented peptide into a single polypeptide with two
linkers (Supplementary Fig. [117]7a), with mutation of tyrosine 84 in
the wildtype (WT) heavy chain to either alanine (Y84A) or to cysteine
(Y84C) for better fitting of the peptide into the peptide-binding
groove (Supplementary Fig. [118]7b)^[119]39. We adopted the same
approach to express three forms of SCE consisting of the peptide
derived from the HLA-G leader sequence (VMAPRTLFL), B2M, and
HLA-E*01:03 heavy chain (Supplementary Fig. [120]7b). Lentivectors
expressing the SCE variants were used to transduce K562 cells; all
three SCE variants supported surface HLA-E expression, with SCE Y84C
showing the highest level of expression (Fig. [121]3a, b) when
transduced at the same MOI.
Fig. 3. Exogenous expression of single-chain HLA-E inhibits NK cell killing
of allogeneic NK cells.
[122]Fig. 3
[123]Open in a new tab
a Experimental schema (for a–f). b Comparison of HLA-E expression in
the parental K562 cells (control) and K562 cells that were transduced
to express SCE WT, SCE Y84A, and SCE Y84C. c, d SCE expression in K562
cells inhibits NK cell activation. Shown are percentages of CD107a^+
(c) and IFNγ^+ (d) NK cells following coculture with the indicated K562
cells. e SCE inhibits NK cell killing of K562 cells. f Comparison of
SCE presenting different leader peptides in inhibiting NK cell killing.
Data in (c–f) are mean ± s.d. with 4 technical replicates. Experiments
were independently repeated for three times. g Schema of NK cell
responses to HLA-ABC-reduced and SCE expressing allogeneic NK cells.
Allogeneic NK cells were transduced with lentivector expressing shRNA
#13 alone or both shRNA #13 and SCE Y84C. Note: shRNA #13 knocks down
endogenous B2M expression but does not knock down SCE expression
because B2M nucleotide sequences were codon optimized in SCE and have 5
mismatches with shRNA #13. Transduced allogeneic NK cells
(HLA-ABC-reduced or plus SCE) were sorted and expanded. NK cells from
five different donors were cocultured with allogeneic NK cells with
normal (control) or reduced HLA-ABC expression (#13), or reduced
HLA-ABC plus SCE expression (#13 + SCE) for 6 h. h, i Percentages of
CD107a^+ and IFNγ^+ host NK cells. j Percentages of killing of
allogeneic NK cells by host NK cells at different E:T ratios in the
presence of anti-CD47 antibody and Fc blocker. Data in (h–j) are
mean ± s.d. from two independent experiments with five healthy donors
each in two technical replicates. One-way (c, d and h, i) or two-way (e
and j) ANOVA with Tukey’s multiple comparisons was performed between
different groups to determine the statistical difference. P-values are
shown as numbers in (c–e, h–j). Figure 3a, g are created in Biorender.
To evaluate the functionality of SCE variants, purified human NK cells
were cocultured with K562 cells with or without SCE expression
(Fig. [124]3a). NK cell degranulation (CD107a), IFNγ expression, and
lysis of K562 cells were measured 4 h post-coculture. Approximately 55%
of NK cells were induced to express CD107a following coculture with the
parental K562 cells (Fig. [125]3c). This percentage was significantly
reduced to ~35%, ~30%, and ~10% by K562 cells expressing SCE WT, Y84A,
and Y84C variants, respectively. Similarly, the percentages of IFNγ^+
NK cells were significantly reduced from ~40% following coculture with
the parental K562 cells to ~15–30% following coculture with K562 cells
expressing the SCE variants, with the SCE Y84C showing the most potent
inhibition (Fig. [126]3d). In addition, human NK cells effectively
killed the parental K562 cells with increasing E:T ratios
(Fig. [127]3e), whereas killing of K562 cells was significantly
inhibited by expression of the SCE variants, especially SCE Y84C (80%
vs 20% lysis at E:T ratio of 10).
HLA-E presents peptides derived from the leader sequences of the heavy
chains of HLA-ABC and HLA-G. Because of amino acid sequence variations
in the leader sequences, we tested the peptide from HLA-G and 6
peptides from 13 prevalent HLA-ABC alleles (Supplementary
Fig. [128]7c). K562 cells were transduced with lentivectors expressing
SCE Y84C presenting these different peptides, cocultured with human NK
cells for 4 h at the E:T ratios of 0.5, 1, and 3. Human NK cells
efficiently lysed the parental K562 cells, reaching ~85% at E:T ratio
of 3 (Fig. [129]3f). Expression of SCE Y84C presenting different
peptides all significantly inhibited lysis of the transduced K562
cells, but SCE presenting the HLA-G peptide was the most potent in
inhibiting NK cell lysis of K562 cells. Thus, SCE Y84C, presenting the
HLA-G peptide, was used in all subsequent studies.
We tested whether SCE Y84C expression could protect HLA-ABC-reduced
allogeneic NK cells from being killed by host NK cells. We used shRNA
#13 to knock down HLA-ABC expression because it was more potent than
shRNA #1, enabling a more stringent killing assay. Allogeneic NK cells
were transduced with lentivector expressing shRNA #13 or shRNA #13 plus
SCE Y84C, and the transduced NK cells were purified and expanded
(Fig. [130]3g). NK cells from five healthy donors were cocultured with
allogeneic NK cells with normal or reduced HLA-ABC expression, or
reduced HLA-ABC but exogenous SCE Y84C expression at E:T ratios of 0,
5, and 10 for 6 h (in the presence of anti-CD47 and Fc blocker
antibodies). NK cells were not induced to express CD107a and IFNγ
following coculture with allogeneic NK cells with normal HLA-ABC
expression, but were significantly induced to express CD107a and IFNγ
following coculture with HLA-ABC-reduced NK cells (Fig. [131]3h, i).
Exogenous expression of SCE Y84C in HLA-ABC-reduced NK cells almost
completely inhibited the induction of CD107a and IFNγ following
coculture (Fig. [132]3h, i). Similarly, NK cells did not kill
allogeneic NK cells with normal HLA-ABC expression, but killed
HLA-ABC-reduced NK cells, whose elimination was completely inhibited by
exogenous expression of SCE Y84C (Fig. [133]3j). These results show
that exogenous expression of SCE Y84C inhibits host NK cell killing of
allogeneic NK cells with reduced HLA-ABC expression.
HLA-ABC-reduced allogeneic NK cells with PD-L1 or SCE overexpression evade
host immune cell killing in vitro
We tested whether a combination of HLA-ABC knockdown and exogenous
PD-L1 and/or HLA-E expression inhibits the killing of allogeneic NK
cells by host T cells and NK cells (Supplementary Fig. [134]8).
Allogeneic NK cells were transduced with lentivectors expressing shRNA
#1, shRNA #13, SCE, PD-L1, or various combinations, and the
downregulation of HLA-ABC, and/or expression of HLA-E or PD-L1 were
verified (Supplementary Fig. [135]9). The transduced allogeneic NK
cells were sorted, expanded, and used as targets. Host CD8^+ T cells
were primed with mitomycin C pretreated untransduced allogeneic NK
cells (from the same donor as the NK cell target) in the presence of
IL-2 for 7 days. Primed CD8^+ T cells were then cocultured with labeled
untransduced (control) or transduced allogeneic NK cells at an E:T
ratio of 0, 10, and 20 overnight. Untransduced allogenic NK cells were
efficiently lysed by CD8^+ T cells from two different donors, reaching
60% lysis at E:T ratio of 20 (Supplementary Fig. [136]10a). Expression
of PD-L1 or SCE only moderately inhibited CD8^+ T cell killing of
allogeneic NK cells, with PD-L1 being more effective than SCE (40% vs
55%). Knockdown of HLA-ABC by shRNA #1 or #13 significantly reduced
CD8^+ T cell killing of allogeneic NK cells to ~10% at E:T ratio of 20.
A combination of knockdown of HLA-ABC and SCE and/or PD-L1 expression
did not further reduce CD8^+ T cell killing of allogeneic NK cells as
compared to HLA-ABC knockdown alone. In a separate killing assay with
CD8^+ T cells from four healthy donors, a combination of HLA-ABC
knockdown by shRNA #1 or shRNA #13 and SCE or PD-L1 overexpression in
allogeneic NK cells all dramatically inhibited CD8^+ T cell killing of
allogeneic NK cells (Fig. [137]4a), with the combination with PD-L1
expression performing significantly better than with SCE expression.
Thus, HLA-ABC knockdown in allogeneic NK cells is the most important
for evading host CD8^+ T cell killing in vitro, whereas overexpression
of PD-L1 and/or SCE is less effective on its own (Fig. [138]4a and
Supplementary Fig. [139]10a).
Fig. 4. Combination of HLA-ABC reduction and expression of SCE or PD-L1 in
allogeneic NK cells inhibits host T and NK cell responses.
[140]Fig. 4
[141]Open in a new tab
a Host CD8^+ T cell killing of allogeneic NK cells in vitro. b Host NK
cell killing of allogeneic NK cells in vitro. Data in (a, b) are
mean ± s.d. from three independent experiments with four healthy donors
each in 3 technical replicates. Each circle represents one doner. c
Schematic diagram of the experimental protocol. d, e Relative
percentages of donor 1 NK cells (red bar), transduced donor 2 NK cells
(black bar), and untransduced donor 2 NK cells (open bar) among total
human NK cells before transfer (BT) and day 7, 14, and 21 after
transfer in the blood (d) or in the tissues day 21 after transfer (e).
f, g Relative percentages of donor 2 NK cells (red bar), transduced
donor 1 NK cells (black bar), and untransduced donor 1 NK cells (open
bar) among total human NK cells before transfer and day 7, 14, and 21
after transfer in the blood (f) or in the tissues day 21 after transfer
(g). h, i Relative percentages of human T cells among total CD45^+
human cells before transfer and day 7, 14, and 21 after transfer in the
blood (h) or in the tissues day 21 after transfer (i). j, k Comparison
of percentages of human T cells that express CD69 before transfer and
day 7, 14, and 21 after transfer in the blood (j) or in the tissues day
21 after transfer (k). Data in (d–k) are mean ± s.d., n = 3 per group
for (d–g) and n = 6 per group for (h–k). Two-way ANOVA with Tukey’s
multiple comparisons was performed between different groups to
determine the statistical difference in (a, b). One-way ANOVA with
Dunnett’s multiple comparison was performed in (h and j). The p-values
are shown for the indicated comparisons. At least 5000 hCD45^+ cells
were collected per sample for analysis. Figure 4c is created in
Biorender.
Similarly, host NK cells were cocultured with labeled untransduced
(control) or transduced allogeneic NK cells at E:T ratio of 0, 5, and
10 for 6 h in the presence of anti-CD47 and Fc blocker. Untransduced or
SCE and/or PD-L1 expressing allogeneic NK cells were not killed by host
NK cells. In contrast, knockdown of HLA-ABC by shRNA #1 or #13 led to
effective host NK cell killing of allogeneic NK cells, reaching 50% and
60%, respectively, at E:T ratio of 10 (Supplementary Fig. [142]10b).
Expression of SCE (or plus PD-L1) completely abolished killing of
HLA-ABC-reduced allogeneic NK cells, whereas expression of PD-L1 did
not. In a separate killing assay with NK cells from four healthy
donors, expression of SCE, but not PD-L1, completely inhibited the
killing of HLA-ABC-reduced allogeneic NK cells (Fig. [143]4b and
Supplementary Fig. [144]10b).
We also evaluated the cytotoxicity of NK cells that had reduced HLA-ABC
expression but overexpressed SCE and/or PD-L1. Untransduced and
transduced NK cells were cocultured with K562 cells at E:T ratios of 0,
0.5, and 1 for 4 h. Untransduced NK cells killed K562 cells in a
dose-dependent manner (Supplementary Fig. [145]11). Knockdown of
HLA-ABC did not significantly affect NK cell killing of K562 cells.
Unexpectedly, expression of SCE or PD-L1 with or without HLA-ABC
knockdown significantly increased NK cell killing of K562 cells
(comparing 30% vs 60–70% lysis at E:T ratio of 1).
Together, these results show that in vitro (i) HLA-ABC knockdown in
allogeneic NK cells is more effective than overexpression of SCE and/or
PD-L1 in inhibiting killing by host CD8^+ T cells, (ii) overexpression
of SCE is more effective than overexpression of PD-L1 in inhibiting
host NK cell killing of HLA-ABC-reduced allogeneic NK cells, and (iii)
overexpression of SCE and PD-L1 significantly enhances NK cell killing
of K562 target cells.
Allogeneic NK cells with HLA-ABC knockdown and PD-L1 or HLA-E expression
evade host immune cell rejection in vivo
We tested whether allogeneic NK cells with HLA-ABC knockdown but PD-L1
or SCE overexpression could evade host T cell and NK cell rejection in
vivo. In the first set of experiments, purified NK cells from donor 1
(HLA-A*02:01^+) were transduced with the lentivector expressing shRNA
#13 and SCE (Fig. [146]4c). The transduced and untransduced NK cells
were mixed with equal numbers of PBMCs from donor 2 (HLA-A*02:01^−) and
adoptively transferred into NSG mice that were also deficient in both
MHC classes I and II (NSG MHC I/II DKO) and had been irradiated 2 days
earlier. Human NK cells and T cells in the cell mixture before adoptive
transfer, in peripheral blood 7, 14, and 21 days after transfer, and in
lung, liver, spleen and bone marrow 21 days after transfer were
analyzed by flow cytometry (Fig. [147]4c). Transduced NK cells from
donor 1 were identified as CD45^+, CD56^+, HLA-ABC-reduced, HLA-E^+,
and HLA-A*02:01^−; untransduced NK cells from donor 1 were identified
as CD45^+, CD56^+, HLA-ABC^+, HLA-E^−, and HLA-A*02:01^+, and NK cells
from donor 2 PBMC were identified as CD45^+, CD56^+, HLA-ABC^+,
HLA-E^−, and HLA-A*02:01^− (Supplementary Fig. [148]12). Before
adoptive transfer, the relative percentages of transduced and
untransduced NK cells from donor 1 among total NK cells were ~45% each,
whereas the percentages of donor 2 NK cells were ~10% (Fig. [149]4d,
e). The percentages of donor 1 transduced NK cells increased to ~70% 7
days post transfer and to ~80% 14- and 21-days post transfer in the
blood. Correspondingly, the percentage of donor 1 untransduced NK cells
decreased to ~10% 7 days post transfer and to <5% 14- and 21-days post
transfer, whereas the percentages of donor 2 NK cells increased to
~20%. Consistently, the percentages of donor 1 transduced NK cells were
~80% in lung, liver, spleen and bone marrow at 21 days post transfer,
whereas the percentages of donor 1 untransduced NK cells were <5% in
these tissues. There was no significant change in CD107a expression by
donor 2 NK cells and donor 1 NK cells (transduced or untransduced)
before adoptive transfer and 7, 14, and 21 days after adoptive transfer
in the peripheral blood or tissues (Supplementary Fig. [150]13). Thus,
the untransduced NK cells (HLA-ABC^+ SCE^−), but not transduced NK
cells (HLA-ABC^−/low SCE^+), from donor 1 are preferentially eliminated
in recipient mice.
We assayed the relative levels of donor 2 T cells and their activation
status as indicated by expression of CD69 in the recipient mice. The
percentages of human CD45^+ cells varied between mice and ranged from
13% to 45% in PBMCs and from 9% to 51% among leucocytes in the tissues.
The percentages of T cells within human CD45^+ cells were ~25% in the
cell mixture before adoptive transfer and few expressed CD69
(Fig. [151]4h–k). The percentages of T cells increased to 45–50% among
human CD45^+ cells in the peripheral blood on day 7, 14, and 21 post
transfer and to ~50% in the tissues. Notably, the percentages of CD69^+
T cells increased to ~60% 7 days post transfer and then decreased to
~20% and ~10% 14- and 21-days post transfer in the blood, respectively,
and to ~12% in the tissues (Fig. [152]4j–k). Thus, the kinetics of
donor 2 T cell activation (CD69 expression) is correlated with
rejection of donor 1 untransduced NK cells in the recipient mice.
We also performed a reciprocal experiment where NK cells from donor 2
were transduced, mixed with PBMCs from donor 1, and adoptively
transferred into the same recipient mice. Similar results were
observed, i.e., preferential elimination of donor 2 untransduced NK
cells (HLA-ABC^+ SCE^−), but not donor 2 transduced NK cells
(HLA-ABC-reduced SCE^+), and activation of donor 1 T cells at day 7
post transfer in the recipient mice (Fig. [153]4f, g, h–k).
In a similar reciprocal experiment, NK cells from donor 1 (or donor 2)
were transduced with a lentivector expressing shRNA #1 plus PD-L1. NK
cells, containing both transduced and untransduced cells, from donor 1
(or donor 2) were mixed with an equal number of PBMCs from donor 2 (or
donor 1) and adoptively transferred into NSG MHC I/II DKO mice. Again,
similar results were obtained, i.e., (i) allogeneic untransduced NK
cells (HLA-ABC^+ PD-L1^−) were rejected, whereas allogeneic transduced
NK cells (HLA-ABC-reduced PD-L1^+) were not, and (ii) induction of
highest percentages of CD69 expression by T cells 7 days post transfer
(Supplementary Fig. [154]14). Together, these results show that
allogeneic NK cells with HLA-ABC knockdown and exogenous expression of
either SCE or PD-L1 can evade rejection by host T cells and NK cells in
vivo.
CAR NK cells with HLA-ABC knockdown and PD-L1 or SCE expression exhibit
enhanced cytotoxicity against tumor cells in vitro
To investigate whether HLA-ABC knockdown and exogenous expression of
PD-L1 or SCE could be used in one-step production of allogeneic CAR-NK
cells, we constructed lentivectors encoding shRNA #1 or #13, CD19-CAR
or mesothelin (MSLN)-CAR, and PD-L1 or SCE (Supplementary
Fig. [155]15). CD19 CAR included a CD8α signal peptide, a single-chain
variable fragment (scFv) specific for the human CD19, CD8α hinge,
transmembrane region, and 4-1BB and CD3ζ intracellular domains. The
MSLN CAR had the same structure, except that it was scFv-specific for
the human mesothelin. NK cells were transduced with these lentivectors,
and the expression of CAR, SCE, or PD-L1, and HLA-ABC was verified. For
example, NK cells that were transduced with lentivector expressing
shRNA #1 (referred to as #1), MSLN CAR, and PD-L1 or SCE expressed MSLN
CAR and PD-L1 or SCE and had reduced HLA-ABC expression (Fig. [156]5a,
b). The transgene-expressed PD-L1 was significantly higher in
transduced NK cells than the endogenous PD-L1 expressed on a small
fraction of human NK cells (Supplementary Fig. [157]16a–c). Knockdown
of HLA-ABC and expression of CAR and PD-L1 did not affect the
expression of the killer cell immunoglobulin-like receptors (KIR),
including CD158a, CD158b, and CD158e (Supplementary Fig. [158]16d–g).
Furthermore, HLA-ABC knockdown, CAR expression, and PD-L1 or SCE
overexpression did not affect CD94, NKG2A, and NKG2C transcript levels
in NK cells (Supplementary Fig. [159]16h). To test the CAR-mediated
target cell killing, we assayed killing of OVCAR8 cells by MSLN-CAR NK
cells and Raji cells by CD19-CAR NK cells in vitro. MSLN-CAR NK cells
killed OVCAR8 cells in a dose-dependent manner, reaching 30% at E:T
ratio of 3 (Fig. [160]5c). Knocking down of HLA-ABC by shRNA #1 or #13
in MSLN-CAR NK cells did not have any significant effect on
CAR-mediated killing of OVCAR8 cells. Expression of PD-L1 or SCE in
MSLN-CAR NK cells with HLA-ABC knockdown killed OVCAR8 cells more
effectively, reaching ~50% lysis at E:T ratio of 3. Similarly,
untransduced NK cells did not kill Raji cells appreciably, whereas
CD19-CAR NK cells efficiently killed Raji cells in a dose-dependent
manner, reaching ~50% at E:T ratio of 3 (Fig. [161]5d). Knocking down
of HLA-ABC by shRNA #1 or #13 in CD19-CAR NK cells did not have any
significant effect on CAR-mediated killing of Raji cells. In contrast,
CD19-CAR NK cells that had HLA-ABC knockdown and exogenously expressed
PD-L1 or SCE killed Raji cells more effectively, reaching ~80% lysis at
E:T ratio of 3, consistent with enhanced killing of K562 cells by NK
cells that express PD-L1 or SCE (Supplementary Fig. [162]11). These
results show that shRNA knockdown of HLA-ABC and exogenous expression
of PD-L1 or SCE can be integrated into the existing CAR lentivector in
a one-step production of allogeneic CAR NK cells, and exogenous PD-L1
or SCE expression enhances CAR-NK cell killing of target cells.
Fig. 5. CD19-CAR and MSLN-CAR NK cells with HLA-ABC knockdown and PD-L1 or
SCE expression exhibit enhanced cytotoxicity against tumor cells in vitro.
[163]Fig. 5
[164]Open in a new tab
a, b Expression of MSLN CAR, HLA-ABC, and PD-L1 by NK cells transduced
with #1 + MSLN CAR + PD-L1 (a) or expression of MSLN CAR, HLA-ABC, and
HLA-E by NK cells transduced with #1 + MSLN CAR + SCE (b). Red
histograms, transduced NK cells; black histograms, untransduced NK
cells. c Comparison of lysis of OVCAR8 target cells by various MSLN-CAR
NK cells (d) or lysis of Raji target cells by various CD19-CAR NK cells
at the indicated E:T ratio. Data in (c, d) are mean ± s.d. from three
independent experiments with four healthy donors each in 3 technical
replicates. Each dot represents one doner. Two-way ANOVA with Tukey’s
multiple comparisons was performed between different groups to
determine the statistical difference in (c, d). e, f Enriched pathways
in #1 + MSLN CAR + PD-L1 (e) and #1 + MSLN CAR + SCE (f) NK cells. NK
cells isolated from 4 unrelated donors were transduced with
lentivectors expressing #1 + MSLN CAR, #1 + MSLN CAR + PD-L1, and
#1 + MSLN CAR + SCE and cocultured with OVCAR8 target cells for 18 h.
NK cells from the same donors and without transductions were used as
the controls. After coculture, NK cells were purified, and RNA was
isolated and subjected to bulk RNA sequencing. Gene set enrichment
analysis (GSEA) was performed using fgsea (v 1.29.1), and the p-values
were calculated with weighted two-sided Kolmogorov–Smirnov statistic
tests and adjusted using the Benjamini–Hochberg correction. Gene sets
were curated from the msigDB (v 7.5.1) database. Data visualization was
done with ggplot2 (v3.5.0). Cellular stress, proinflammatory,
proliferation, and homeostasis-related pathways are labeled in blue,
yellow, red, and gray, respectively.
To elucidate the mechanisms underlying the enhanced NK cell
cytotoxicity following PD-L1 and SCE overexpression, we performed
transcriptomic analysis of untransduced NK cells from 4 unrelated
donors and their counterparts transduced with lentivectors expressing
shRNA #1 + MSLN CAR, shRNA #1 + MSLN CAR + PD-L1, or shRNA #1 + MSLN
CAR + SCE. NK cells were labeled with Celltrace Violet (CTV) and
cocultured with OVCAR8 target cells or in the media alone for 18 h in
triplicate. The cocultured NK cells were purified by cell sorting and
processed for RNA-seq. The principal component analysis revealed that
NK cells cocultured with OVCAR8 target cells account for the greatest
transcriptional difference among the samples (Supplementary
Fig. [165]17). The differential gene expression analysis revealed that
the cytotoxicity-related genes (e.g., GZMB and IFNG) were upregulated
in NK cells after coculture with target cells (Supplementary
Fig. [166]18a). The non-cocultured (resting) #1 + CAR + PD-L1 NK cells
upregulated KLRC4 but downregulated anti-inflammatory cytokine IL-13
compared with the resting CAR NK cells. Similarly, the #1 + CAR + SCE
NK cells expressed higher levels of PRF1, GZMA, and IFNAR2 compared
with the resting #1 + CAR NK cells (Supplementary Fig. [167]18b, c).
Furthermore, cocultured #1 + CAR, #1 + CAR + PD-L1, and #1 + CAR + SCE
NK cells upregulated cytotoxicity-related genes (e.g., CD226, IFNG,
IFNGR1, and GZMB) compared to their resting counterparts (Supplementary
Fig. [168]18d–f), suggesting that CAR recognition of target cells is a
superior signal for NK cell activation. The pathway enrichment analysis
revealed that the proliferation and proinflammation pathways were
enriched in #1 + CAR + PD-L1 NK cells, and homeostasis pathways were
enriched in #1 + CAR + SCE NK cells after co-culture with the target
cells. In contrast, the cellular stress pathways were more enriched and
associated with #1 + CAR NK cells after co-culture with the target
cells (Fig. [169]5e, f). This data suggests that CAR NK cells may
experience cellular stress when killing target tumor cells, which is
alleviated by exogenous expression of inhibitory ligands PD-L1 or SCE.
Allogeneic CD19-CAR NK cells with HLA-ABC knockdown and PD-L1 or SCE
overexpression evade host immune cell killing and control tumor growth in
vivo
We determined whether CAR NK cells with HLA-ABC knockdown and PD-L1 or
SCE overexpression could control tumor growth and avoid rejection by
host T and NK cells in NSG mice. Purified allogeneic NK cells from
HLA-A*02:01^− donors were either not transduced (group 1) or transduced
with lentivectors expressing CD19 CAR (group 2), #1 + CD19 CAR + PD-L1
(group 3), and #1 + CD19 CAR + SCE (group 4) (Fig. [170]6a). The
allogeneic NK cells were mixed with equal numbers of PBMCs from
HLA-A*02:01^+ donors and adoptively transferred into NSG MHC I/II DKO
mice that had been injected with Raji-luciferase tumor cells two days
earlier. Tumor burden was significantly reduced in groups 3 and 4 mice
as compared to groups 1 and 2 mice, as indicated by significantly lower
bioluminescence (BLI) and longer survival (Fig. [171]6b–d). Notably,
group 3 mice showed a significantly lower BLI and slightly better
overall survival than group 4 mice (Fig. [172]6c, d).
Fig. 6. Allogeneic CD19-CAR NK cells with HLA-ABC knockdown and PD-L1 or SCE
overexpression evade host immune cell rejection and control tumor growth in
vivo.
[173]Fig. 6
[174]Open in a new tab
a Schema of the experimental protocol. Allogeneic NK cells without
transduction (group 1) or transduced with lentivectors expressing CD19
CAR (group 2), #1 + CD19 CAR + PD-L1 (group 3), or #1 + CD19 CAR + SCE
(group 4), were mixed with equal numbers of host PBMCs (4 million), and
adoptively transferred into recipient NSG MHC I/II DKO mice that had
been injected (IV) with 0.5 million Raji-luciferase two days earlier.
b, c Tumor burden was assessed on day 7, 14, and 21 by BLI (b) and
presented as the normalized intensity of BLI (c). Groups 1, 2, 3, and 4
mice are labeled in black, blue, orange, and red, respectively. Data
are mean ± s.d. (n = 4 for group 1 and 2, n = 5 for group 3 and 4). d
Kaplan–Meier curves showing survival of mice. The numbers indicate the
p-values between the indicated two groups. e, f Relative percentages of
host NK cells (red bar), allogeneic transduced NK cells (black bar),
and allogeneic untransduced NK cells (open bar) before transfer, in the
blood at day 14 after transfer (e) and in the blood and tissues at the
endpoint (f). g, h CD107a expression in host NK cells (red bar),
allogeneic transduced NK cells (black bar), and allogeneic untransduced
NK cells (open bar) at day 14 after transfer (g) and in spleen at the
endpoint (h). Data in (e–h) are mean ± s.d. from two independent
experiments, n = 9 for group 1 and 2, and n = 10 for group 3 and 4, at
day 14 n = 8 for all four groups at the endpoint. i, j CD69 expression
in host CD8^+ (open bar) and CD4^+ T (gray bar) cells in blood at day
14 after transfer (i) and in spleen at the endpoint (j). Data are
mean ± s.d. with 5 mice for each group at day 14 and at the endpoint.
At least 5000 hCD45^+ cells were collected per sample for analysis.
Two-way ANOVA with Tukey’s multiple comparisons and Log-rank test were
performed between different groups to determine the statistical
difference in (c and d), respectively. The p-values are shown for the
indicated comparisons. Figure 6a is created in Biorender.
We analyzed the proportion and activation status of NK and T cells in
the tumor-bearing recipient mice. Fourteen days post-cell transfer,
blood was collected from all four groups of mice. At the endpoint, both
blood and tissues, including spleen, liver, lung, and bone marrow, were
collected. Blood was lysed of red blood cells, and single-cell
suspensions were prepared from the tissues. Cells were stained with
fluorochrome-labeled antibodies specific for human CD45, CD3, CD4, CD8,
CD69, CD56, CD107a, HLA-A2, and HLA-ABC followed by flow cytometry.
Host CD8^+ T cells were identified as CD45^+ CD3^+ CD8^+; host CD4^+ T
cells as CD45^+ CD3^+ CD4^+; host NK cells as CD45^+ CD56^+ HLA-A2^+;
untransduced allogeneic NK cells as CD45^+ CD56^+ HLA-ABC^+ HLA-A2^−;
and transduced allogeneic NK cells as CD45^+ CD56^+ HLA-ABC^− HLA-A2^−
(Supplementary Fig. [175]19). The percentages of human CD45^+ cell
varied between mice from 8% to 36% in the PBMCs and from 7% to 45% in
leucocytes of the tissues.
In group 1 mice, unmodified allogeneic NK cells were >80% of total NK
cells before adoptive transfer. The percentage decreased to <20% in the
blood by day 14 after transfer and to <10% in the blood and tissues at
the endpoint, while percentages of host NK cells increased
proportionally, suggesting that allogenic NK cells were preferentially
eliminated in the recipient mice (Fig. [176]6e, f). Similarly, in group
2 mice, untransduced and transduced allogenic NK cells each consisted
of ~45% of total NK cells. Still, this percentage was reduced to <20%
in the blood by day 14 after transfer and to <10% in the blood and
tissues at the endpoint, suggesting their rejection in the recipients
(Fig. [177]6e, f). In group 3 and group 4 mice, while untransduced and
transduced allogeneic NK cells started at similar percentages (~45%)
before transfer, the proportion of untransduced allogeneic NK cells
decreased to ~15% in the blood 14 days after transfer and to ~5% in the
blood and tissues at the endpoint, whereas the proportion of the
transduced allogeneic NK cells increased to ~75% in the blood 14 days
after transfer and to >80% in the blood and tissues at the endpoint
(Fig. [178]6e, f). Thus, allogeneic CAR-NK cells with HLA-ABC knockdown
and PD-L1 or SCE expression evaded rejection, whereas their
untransduced counterparts in the same recipient mice were rejected.
Furthermore, allogeneic CD19 CAR NK cells, but not the untransduced
allogeneic NK cells, upregulated CD107a expression, suggesting their
response to CD19^+ Raji tumor cells. The host NK cells did not
upregulate CD107a expression, indicating that they may not play a
significant role in rejection of allogeneic NK cells (Fig. [179]6g, h
and Supplementary Fig. [180]20a–d).
We also investigated the dynamics of CD8^+ and CD4^+ T cells in total T
cells and their activation status after adoptive transfer, along with
allogeneic NK cells. Consistent with rejection of allogeneic NK cells
in group 1 and 2 mice, the relative percentages of CD8^+ T cells in
total T cells were increased from ~45% to ~60% in the blood at day 14
after transfer and at the endpoint, whereas the relative percentages of
CD8^+ T cells did not change significantly in group 3 and group 4 mice
(Supplementary Fig. [181]21a, b). Importantly, significantly higher
percentages of CD8^+ T cells expressed CD69 in group 1 and group 2 mice
than in group 3 and group 4 mice (Fig. [182]6 I, j and Supplementary
Fig. [183]20e–h). In contrast, CD4^+ T cells neither proliferated nor
upregulated CD69 expression in all four groups of mice (Supplementary
Figs. [184]20e–h and [185]21c, d). These results show that CD8^+ T
cells are the key cell type involved in rejection of allogeneic NK
cells. Furthermore, a significantly lower percentage of CD8^+ T cells
was CD69^+ in the blood 14 days post transfer (~30%) and this
percentage was reduced to baseline in the blood and tissues at the
endpoint in group 3 mice as compared to those in group 4 mice
(Fig. [186]6I, j and Supplementary Fig. [187]20e–h). These results
suggest that the combination of HLA-ABC knockdown with PD-L1 expression
is more effective than with SCE expression in inhibiting
allo-rejection, consistent with the better tumor control observed in
group 3 mice compared to group 4 mice.
Allogeneic CD19-CAR NK cells with HLA-ABC knockdown and PD-L1 overexpression
inhibit inflammatory cytokine production by host immune cells in vitro and in
vivo
During an alloreaction, a wide array of cytokines is produced by
various immune cells to mediate the immune response against the
transplanted tissue or cells, often through a positive feedback
loop^[188]40. We measured the impact of HLA-ABC knockdown and PD-L1
overexpression by allogeneic CAR NK cells on cytokine production. In
vitro, PMBCs were cocultured with unmodified allogeneic NK cells or the
same donor NK cells transduced to express shRNA #1, CD19 CAR and PD-L1
for 48 h and culture supernatants were collected to measure the levels
of a panel of 48 human cytokines. For 30 cytokines, the same levels
were detected in both cocultures (Fig. [189]7a). For 17 cytokines,
coculturing with transduced allogeneic NK cells significantly inhibited
their production, including IL-6, IL-1, IFNγ, and TNF, that mediate
cytokine release syndrome of CAR-T cells. In vivo, Raji tumor-bearing
NSG MHC I/II DKO mice were adoptively transferred with 4 million PBMCs
plus either 4 million untransduced allogeneic NK cells (Group 1 mice in
Fig. [190]6) or 4 million allogeneic NK cells transduced to express
shRNA #1, CD19 CAR, and PD-L1 (Group 3 mice in Fig. [191]6). The same
panel of 48 human cytokines was measured in the sera 7 days after
transfer. Although fewer human cytokines were detected in mice,
transduced NK cells did not affect the levels of many of the same
cytokines as in the coculture, but significantly reduced the level of 9
cytokines, including complete inhibition of IL-6, MIP-1β and RANTES
(Fig. [192]7b). Thus, knockdown of HLA-ABC and expression of PD-L1 not
only enable allogenic CAR NK cells to evade rejection by host T cells
and enhance tumor control in a CAR-dependent manner but also suppress
production of cytokines involved in cytokine release syndrome.
Fig. 7. HLA-ABC reduction and PD-L1 overexpression by allogenic CAR NK cells
suppress cytokine production.
[193]Fig. 7
[194]Open in a new tab
a Comparison of cytokine profiles following coculture of PMBC with
unmodified allogeneic NK cells or the same donor NK cells transduced
with lentivector expressing #1 + 19CAR + PD-L1. The coculture was
carried out for 48 h, and supernatants were collected to measure the
level of a panel of 48 human cytokines. Data are mean ± s.d. from two
independent experiments with 4 independent healthy donors in two
technical replicates. b Comparison of cytokine profiles following
infusion of PMBC with unmodified allogeneic NK cells or the same donor
NK cells transduced with lentivector expressing #1 + 19CAR + PD-L1. NSG
MHC I/II DKO mice were engrafted with Raji tumor cells and 2 days later
infused with 4 million PBMCs plus either 4 million untransduced or
transduced allogeneic NK cells. Sera were collected 7 days after cell
transfer for measuring the levels of a panel of 48 human cytokines.
Data are mean ± s.d. with 5 mice in each group. A two-tailed and paired
sample t-test was performed to determine the statistical difference
between the two groups for each cytokine. * indicates statistical
significance with p < 0.05.
Discussion
Development of allogeneic products has become a major effort in
cellular therapy. We here describe a one-step approach to produce
allogeneic CAR-NK cells that can evade rejection by the host immune
system and simultaneously exhibit enhanced anti-tumor activity and
safety profiles. In our studies, we systematically evaluated the role
and degree of TCR recognition of allogeneic pMHC and immune checkpoints
PD-L1, HLA-E, and CD47 in rejection of allogeneic NK cells. Our results
show that rejection of allogeneic human NK cells is mainly mediated by
host CD8^+ T cells. First, blocking TCR recognition of allogeneic pMHC
by suppressing HLA-ABC expression on allogenic NK cells is the most
effective method to prevent CD8^+ T cell rejection (Fig. [195]1). Among
T cells, CD8^+, but not CD4^+, T cells, were activated to express CD69
in xenograft mouse models, and the degree of CD8^+ T cell activation
correlated with rejection of allogeneic NK cells. Engagement of PD-1 on
activated T cells by overexpressing PD-L1 on allogenic NK cells only
moderately inhibits allo-rejection (Fig. [196]2), whereas
overexpression of HLA-E is even less effective (Fig. [197]4a),
consistent with the fact that only a fraction of T cells expresses
NKG2A^[198]35,[199]41. Second, human NK cells are inhibited from
killing allogeneic NK cells through both HLA-E-NKG2A and CD47-SIRPα
mechanisms. Consistent with a previous report^[200]42, human NK cells
express high levels of CD47, and most of them also express SIRPα
(Supplementary Fig. [201]2) Even when allogeneic human NK cells are
knocked down of both HLA-ABC and HLA-E, they are not killed by host NK
cells, unless CD47-SIRPα interaction is abrogated by addition of
anti-CD47 and Fc receptor blocker (Figs. [202]3g–j and [203]4b).
Compared to HLA-E, which potently inhibits NK cell activation,
secretion of IFNγ, and cytotoxicity (Fig. [204]3), PD-L1 does not
inhibit NK cell activity probably due to minimal expression of PD-1 by
human NK cells^[205]43,[206]44 (Fig. [207]4a, b). Demonstration that
rejection of allogeneic human NK cells is primarily mediated by CD8^+ T
cells suggests that the efforts to produce off-the-shelf allogeneic
CAR-NK cell therapeutics should be focused towards inhibiting CD8^+ T
cell rejection.
Most efforts targeting HLA-ABC expression have involved abrogating
either B2M^[208]45–[209]47 or TAP^[210]48, which suppresses expression
of all the MHC class I molecules, including HLA-E, making the
allogeneic cells susceptible to NK cell-mediated rejection^[211]49. In
addition, two steps of genetic manipulations are used to first suppress
HLA-ABC expression and then express CAR in previous studies.
Interestingly, a recent study shows that the knockdown of CD54/CD58 by
CRISPR-Cas9 can abrogate immune synapse formation, leading to the
inhibition of allo-rejection by host NK cells and T cells^[212]50.
However, this approach also requires multiple steps when combined with
CAR expression. We chose to develop shRNA to knockdown HLA-ABC with the
goal to easily integrate the shRNA into the same CAR lentivector for
one-step construction of allogeneic CAR-NK cells from the peripheral
blood NK cells. We identified a specific shRNA that is effective in
knocking down surface expression of the 119 most prevalent HLA-ABC
alleles, including HLA-A*02:01, without affecting HLA-E expression due
to two nucleotide mismatches between shRNA and HLA-E heavy chain
(Fig. [213]1a and Supplementary Fig. [214]4). To simplify HLA-E
expression and identify the most potent inhibitory peptide, we tested
three SCE variants and different peptides derived from the leader
sequences of HLA-ABC and HLA-G (Fig. [215]3a–f). Our results show that
SCE Y84C presenting the HLA-G leader peptide is most highly expressed
and the most potent in inhibiting NK cell reactivity.
By combining the novel shRNA, CD19 or MSLN CAR, PD-L1 or SCE into the
same lentivector, we were able to produce allogeneic CAR-NK cells from
peripheral blood NK cells by one-step lentiviral transduction and show
that these allogeneic CAR-NK cells evade rejection by host immune
system and can kill target tumor cells in a CAR-dependent manner in
vivo. While allogeneic NK cells were rapidly rejected by host T cells
in NSG recipient mice as indicated by simultaneous disappearance of
allogeneic NK cells and activation (CD69) and expansion of CD8^+ T
cells at 7 days post transfer (Figs. [216]4c–k and [217]6), allogeneic
CAR-NK cells with HLA-ABC knockdown and expression of PD-L1 or SCE were
not rejected. In tumor-bearing mice, significantly higher proportions
of allogeneic CAR-NK were activated as indicated by CD107a expression,
leading to better control of tumor burden and survival (Fig. [218]6).
Correlating with more potent inhibition of T cell-mediated
allo-response in vitro and in vivo, expression of PD-L1 in combination
with HLA-ABC knockdown and CAR expression is more effective than
expression of SCE for tumor control and survival in vivo. Thus, HLA-ABC
knockdown and PD-L1 expression is the most potent combination for
inhibiting CD8^+ T cell rejection of allogenic human NK cells.
Overexpression of PD-L1 has also been used to avoid the rejection of
allografts, particularly the transplanted islets, to treat type I
diabetes^[219]28,[220]51. We overexpressed PD-L1 on allogeneic NK cells
and found a moderate inhibition of host T cell responses. These
findings suggest that PD-L1 alone on immune and likely non-immune cells
is insufficient to inhibit allogeneic immune responses completely, and
other engineering, such as HLA-ABC knockdown, should be combined to
achieve better survival of engineered cells in allogeneic hosts.
Notably, allogeneic CAR NK cells with HLA-ABC knockdown and exogenous
PD-L1 expression also potently inhibited cytokine production by human
PBMCs both in coculture and in a xenograft mouse model, including IL-6,
IFNγ, and TNF that mediate cytokine release syndrome of CAR-T cells.
Thus, the combination likely endows CAR NK cells with a better safety
profile, probably by inhibiting alloreaction of CD8^+ T cells.
Unexpectedly, the exogenous expression of PD-L1 or SCE significantly
enhances NK cell killing of K562 cells in the absence of CAR
(Supplementary Fig. [221]11) and Raji and OVCAR8 tumor cells in a
CAR-dependent manner (Fig. [222]5). Our findings are consistent with
previous report showing that human NK cells upregulate PD-L1 expression
after encountering cancer cells or upon cytokine activation, and these
NK cells are more effective in killing target cells^[223]33. PD-L1
expression varies in NK cells derived from UCB, peripheral blood, and
iPSC. In our study, we exclusively used NK cells from the peripheral
blood. Comparison of transcriptional profiles of unmodified NK cells
(no CAR) and CAR-NK cells with or without coculture with target OVCAR8
cells revealed the upregulation of genes involved in NK cell
cytotoxicity following PD-L1 or SCE expression. In particular,
following coculture with OVCAR8 target cells, MSLN CAR-NK cells
upregulated genes involved in stress responses, which were inhibited by
exogenous PD-L1 or SCE expression. NK cells likely become stressed and
exhausted during the process of killing target cells (degranulation).
Expression of PD-L1 or SCE overcomes this stress and enables NK cells
to survive and respond better to stress for achieving enhanced
cytotoxicity. In this respect, expression of PD-L1 by tumor cells may
also enable them to respond better to stress^[224]33. Our finding that
exogenous expression of PD-L1 is sufficient to enhance NK cell activity
suggests that PD-L1 can reprogram NK cell physiology, consistent with
observations that PD-L1 possesses signal transduction capacities
without needing to engage PD-1 and acts through mTOR-AKT pathways to
positively regulate cell proliferation, autophagy, and resistance to
pro-apoptotic cytokines^[225]52.
Our finding that overexpression of SCE in NK cells enhances their
cytolytic activity is not only unexpected but also paradoxical,
considering the key role of HLA-E/NKG2A interaction in regulating NK
cell reactivity. We show that on one hand, SCE-overexpressing NK cells
kill K562, Raji, and OVCAR8 target cells more effectively than their
counterparts without SCE expression, and on the other hand, the same
SCE-overexpressing allogeneic NK cells also potently inhibit their own
killing by host NK cells (Figs. [226]3 and [227]4 and Supplementary
Fig. [228]10). The latter would suggest that SCE can engage NKG2A on
host NK cells to inhibit their cytolytic activity. Since SCE
overexpression in NK cells did not affect their expression of NKG2A,
NKG2C, CD94, and KIRs (Supplementary Fig. [229]16), SCE-overexpressing
NK cells also express NKG2A. Then, it is puzzling why their cytolytic
activity is not inhibited by intracellular and/or intercellular
SCE/NKG2A engagement. Furthermore, considering that HLA-E is thought
not to have signaling domains, additional studies are required to
elucidate the mechanisms underlying the paradoxical functions of NK
cells following SCE overexpression.
In summary, we have developed a one-step approach to construct
allogeneic CAR NK cells with reduced HLA-ABC expression but PD-L1 or
SCE overexpression. The resulting allogeneic CAR NK cells can evade
host immune cell rejection while mediating enhanced anti-tumor
responses and exhibiting a better safety profile, thus representing a
significant advancement in enabling “off-the-shelf” allogeneic cellular
immunotherapies.
Methods
Animals, cell lines, and antibodies
The NSG MHC I/II DKO mice (strain #: 025216) combined the features of
severe combined immune deficiency (scid), IL-2 receptor gamma chain
deficiency, and MHC I and MHC II deficiency, were purchased from the
Jackson Laboratory and co-housed under specific pathogen-free
conditions at Massachusetts Institute of Technology (MIT). All the
mouse studies and procedures were approved by MIT Committee for Animal
Care (Protocol ID: 0322-021-25). The HEK293T and OVCAR8 cells were
cultured in Dulbecco’s modified Eagle’s medium supplemented with fetal
bovine serum (FBS) (10% v/v, Gibco), penicillin/streptomycin (1% v/v,
Gibco), and 2 mM L-glutamine. Suspension-adapted HEK293T cells were
cultured in FreeStyle 293 medium (ThemoFisher Scientific #K900001)
supplemented with penicillin/ streptomycin (1% v/v, Gibco). Jurkat T
cells and Raji cells were cultured in RPMI-1640 medium supplemented
with 10% FBS, 1% penicillin/streptomycin, and 2 mM L-glutamine.
Antibodies, their commercial sources and dilutions are provided in
Supplementary Table [230]1. In this study, experiments were not blinded
due to the complexity of the study. Sample sizes were chosen based on
convention, cost consideration, and previous studies that were
sufficient for statistical analysis.
Design of shRNA
For shRNAs targeting B2M, TAP1, and TAP2 mRNA sequences, the mRNAs were
first obtained from the National Center for Biotechnology Information.
shRNAs were predicted using the Genetic Perturbation Platform developed
at Broad Institute
([231]https://portals.broadinstitute.org/gpp/public/seq/search) with
the default parameters. We randomly took at least three top-ranked
shRNAs for each target gene for further investigation. For designing
shRNAs targeting HLA-ABC heavy chain specific regions that are common
to prevalent HLA-ABC alleles but have at least two nucleotide
mismatches with HLA-E*01:01 and HLA-E*01:03, we first obtained the
prevalent HLA-ABC alleles in global populations from the Allele
Frequency Net Database ([232]https://www.allelefrequencies.net/). The
chosen HLA-ABC alleles are listed in Supplementary Fig. [233]3a. The
coding sequences of the chosen HLA class I alleles were obtained from
the IDP-IMGT/HLA database ([234]https://www.ebi.ac.uk/ipd/imgt/hla/)
and were aligned using Geneious Prime. The regions that are common to
HLA-ABC alleles but have mismatches to HLA-E were selected to predict
shRNAs specifically knocking down HLA-ABC. Those potential regions were
analyzed using the Genetic Perturbation Platform with default
parameters to predict efficient shRNAs, and the top-ranked shRNAs were
selected for further investigation.
NK cell isolation, CIML stimulation, expansion, and resting
Whole blood was collected from anonymous healthy donors under the blood
collection protocol approved by the Institutional Review Board of
Dana-Farber Cancer Institute (Study ID: T0197). Peripheral blood
mononuclear cells (PBMC) were first isolated using Ficoll-Paque density
gradient media (Cytiva #17144002) with SepMate PBMC isolation tubes
(STEMCELL #85450) according to the manufacturer’s instructions. The
isolated PBMCs were either used for NK cell or T cell isolation
immediately or frozen down in liquid nitrogen for later use. All the
primary NK cells used in the present study are cytokine-induced
memory-like (CIML) NK cells as described previously^[235]53. NK cells
were isolated using the human NK cell isolation kit (Miltenyi
#130-092-657) according to the manufacturer’s instructions. NK cells
were subjected to CIML stimulation immediately. Briefly, NK cells were
adjusted to 3–4 million per ml in NK MACS medium (Miltenyi
#130-114-429) supplemented with 10% human AB serum and 1%
penicillin/streptomycin (NK complete media), and 3 ml of NK cells were
seeded into one well of a 6-well plate. For CIML stimulation,
recombinant human IL-12 (Miltenyi #130-096-705), IL-15 (Miltenyi
#130-095-760), and IL-18 (Miltenyi #170-076-183) were added to NK
complete media to concentrations of 10 ng/ml, 50 ng/ml, and 50 ng/ml,
respectively. The NK cells were incubated at 37 °C, 5% CO[2] for
18–24 h. Successfully CIML-stimulated NK cells formed clusters and
attained sticky/branched morphology. When NK cells reached this state,
the CIML stimulation media were completely replaced with the expansion
medium (fresh NK complete medium supplemented with 500 IU/ml of IL-2
(Miltenyi #130-097-748) and 5 ng/ml of IL-15). NK cells were expanded
in the expansion medium for about two weeks, with viability greater
than 90%. After the NK cell expansion was completed, the expansion
medium was replaced with fresh NK complete medium supplemented with
1 ng/ml of IL-15 for two to three days before any functional assays.
Production of lentiviral vectors and transduction of NK cells
We pseudotyped lentiviruses with baboon endogenous retroviral envelope
(BaEV)^[236]54 and titrated them on Jurkat T cells. The titrated
lentiviral vectors were used to transduce NK cells from healthy donors
at a multiplicity of infection (MOI) of 10. Briefly, suspension adapted
293 T cells were adjusted to 4 M/ml and transfected with the transfer
plasmid encoding the gene of interest, the envelop plasmid encoding the
BaEV (derived from pMD2.G by replacing the VSVG with BaEV encoding
sequence), the packaging plasmid (pCMV-dR8.91), and the pAdVAntage
plasmid, complexed with polyethyleneimine (Polysciences #49553-93-7) at
an optimal mass ratio. The transfected 293 T cells were pelleted and
resuspended in fresh media at 15 h post-transfection. The supernatant 1
containing the lentiviral vectors was collected by pelleting the cells
24 h later. The cells were resuspended in fresh media and cultured for
another 24 h. The supernatant 2 was collected by spinning down the
cells. Supernatant 1 and supernatant 2 were combined and centrifuged at
3000 rpm, 4 °C, for 10 min to remove the cell debris. The clear
supernatant was then filtered through the 0.45 µm filters, and the
lentiviral vector was concentrated by centrifugation at 10,000 × g,
4 °C, for 20 h. The virus pellet was resuspended in PBS, aliquoted, and
frozen down at −80 °C for later use. The lentiviral vectors were
titrated on Jurkat T cells. 0.5 million Jurkat T cells were transduced
with a serial amount of virus with 8 µg/ml polybrene (Sigma-Aldrich
#TR-1003). The well that yielded around 10–20% positive cells was used
for calculating the viral titer with the formula: titer
(TU/ml) = (500000 × % positive cells)/virus volume. The CIML-stimulated
NK cells were transduced with the lentiviral vector in the presence of
10 µg/ml of Vectofusion-1 (Miltenyi #130-111-163) in the NK expansion
media at an MOI of 10. After adding the virus, the cells were
centrifuged at 1000 × g, 37 °C for 1.5 h. The cells were then put back
in the incubator and cultured for 3 days prior to the examination of
the expression of the protein of interest.
In vitro NK cell functional and killing assays
CIML-stimulated and expanded NK cells were rested for at least 3 days
prior to any functional and killing assays. For the killing of
suspension target cells, the cells were pre-labeled with CTV and seeded
at 2 × 104 in the v-bottom 96-well plate, and the NK cells were added
according to the E:T ratios. For the killing of adherent target cells
(OVCAR8), the cells were pre-labeled with CTV and seeded at 2 × 104 in
the flat-bottom 96-well plate and incubated overnight to allow cell
attachment to the plate. The NK cells were added according to the E:T
ratios and co-cultured with target cells overnight. After co-culture,
the supernatant was transferred to a v-bottom plate, and the attached
cells in the flat-bottom 96-well plate were trypsinized and added to
the corresponding wells in the v-bottom plate. All the NK cell killing
assay was conducted in the presence of 1 ng/ml of IL-15. CD107a and
IFNγ expression in NK cells was measured by staining the cells with
anit-CD56, anti-CD107a, and anti-IFNγ antibodies and determined by flow
cytometry. For the lysis of target cells, the cells were stained with
Zombie NIR (BioLegend #423106), and the precision counting beads
(BioLegend #424902) were added before flow cytometry. The CTV positive
and Zombie NIR negative cells (live target cells) were counted using
the formula: Absolute cell number =
[MATH:
CellscountBeadscountX :MATH]
(added beads volume × beads concentration). The percentage of target
cell lysis was calculated by (total cell number—live cell number)/total
cell number.
In vitro T cell killing assay
We first prepared feeder allogeneic NK cells by culturing them in
RPMI-1640 containing 10 µg/ml mitomycin C (STEMCELL #73274) at 37 °C
with 5% CO2 for 3 h. The cells were then centrifuged and washed with
PBS three times. After that, freshly isolated CD8^+ or CD4^+ T cells
were primed with mitomycin C-pretreated (10 µg/ml) allogeneic NK cells
in the presence of 500 IU/ml IL-2 for 7 days. Primed CD8^+ or CD4^+ T
cells were sorted by negative selection kits (Miltenyi #130-096-533 and
#130-096-495) for the subsequent killing assay. The primed CD8^+ or
CD4^+ T cells were cocultured with CTV prelabeled NK cells (from the
same donor as the priming NK cells) with either normal or reduced
HLA-ABC expression (due to knockdown by shRNA #1 or #13), or with or
without PD-L1 and SCE overexpression, at indicated E:T ratios in the
v-bottom 96-well plate overnight. After coculture, the cells were
centrifuged and stained with Zombie NIR, and the precision counting
beads were added before flow cytometry. The CTV-positive and Zombie
NIR-negative cells were the live NK cells. The lysis of target cells by
CD8^+ or CD4^+ T cells was calculated as described above.
Mixed lymphocyte reaction
Freshly isolated CD3^+ T cells were cocultured with autologous or
HLA-mismatched allogeneic NK cells with or without PD-L1 overexpression
in a v-bottom 96-well plate at a ratio of 1:1. The cell mixture was
mixed by pipetting and centrifuged at 1000 rpm for 1 min. Twelve hours
later, the cells were centrifuged and stained with anti-CD3, anit-CD25,
and anti-CD69 antibodies and analyzed by flow cytometry to determine
the CD3^+CD25^+ and CD3^+CD69^+ T cells. In a separate experiment, CTV
prelabeled CD3^+ T cells were cocultured with autologous or
HLA-mismatched allogeneic NK cells with or without PD-L1 overexpression
in a v-bottom 96-well plate at a ratio of 1:1 for 5 days, and the
proliferation of CD3^+ T cells was determined by dilution of CTV using
flow cytometry.
Mouse studies
All mice were sex-matched and age-matched (8–10-week-old) into groups
and irradiated with 100 cGy two days before infusion of human cells.
Both sexes of mice were used. Before proceeding to any mouse
experiment, the successful NK cell transduction was first confirmed.
For instance, NK cell transduced with a lentivector expressing shRNA
#1 + PD-L1 showed reduced surface HLA-ABC and high level of surface
PD-L1, and NK cell transduced with a lentivector expressing shRNA
#1 + CD19CAR + PD-L1 showed reduced surface HLA-ABC and expression of
CD19 CAR and HLA-E. All transduced NK cells were used within two weeks
at the time of infusion. Tumor burden was monitored by BLI. Mice were
weighed every two days. Mice were considered moribund and euthanized
with 20% weight loss.
For assaying in vivo allogeneic responses, 10 million NK cells with
HLA-ABC knockdown but overexpression of PD-L1 or SCE were mixed with 10
million PBMCs from an HLA class I mismatched donor, infused into NSG
MHC I/II DKO mice through intravenous (i.v.) injections. Before
infusion, the percentages of NK cells (donor 1 transduced NK cells,
donor 1 untransduced NK cells, and donor 2 NK cells) and donor 2 T
cells were characterized in detail by staining the cells with
anti-hCD45-FITC, anti-CD3-BUV395, anti-CD56-PE-Cy7, anti-HLA-ABC-PE,
anti-HLA-A2-PerCP-Cy5.5, anti-PD-L1-APC (or anti-HLA-E-APC when SCE
overexpressed cells were infused), anti-CD69-BV711,
anti-anti-CD107a-APC-Cy7 and evaluated by flow cytometry. The antibody
panel was designed using FluoroFinder and evaluated (Supplementary
Figs. [237]12 and [238]19). Peripheral blood from mice was obtained on
day 7, day 14, and day 21, by facial bleeding into an EDTA-coated tube.
Red blood cells were lysed with 4 ml (10× the blood volume) ACK buffer
(ThermoFisher Scientific #A1049201) for 10 min on the ice. Cells were
then washed twice with FACS buffer (PBS + 1% BSA) and then stained with
the above antibody panel. Mice were euthanized by carbon dioxide
(CO[2]) inhalation followed by cervical dislocation on day 21, lung,
liver, spleen, and bone marrow were collected, and leucocytes were
isolated from the organs as described previously^[239]55. In brief, the
tissues (lung, liver, and spleen) were diced using surgical scissors on
a 10-cm sterile cell culture plate and resuspended in 4 ml RPMI-1640.
The diced tissues were further homogenized with gentleMACS Tissue
Dissociator. Leucocytes were isolated using the Percoll density
separation method. The bone marrow cells were aspirated using a 28 G
insulin needle, lysed with ACK buffer, and filtered through a 40 µm
cell strainer to acquire a single-cell suspension. After separation,
the cells from the tissues were stained with the above antibodies. All
the samples were analyzed on a FACS Symphony A3 instrument.
For tumor killing experiment, xenograft models were established by
intravenous injection of 0.5 × 106 Raji-luciferase cells into NSG MHC
I/II DKO mice. Two days later, 4 million allogeneic CAR NK cells and 4
million PBMCs from a different donor were infused into mice through
intravenous injection. Tumor burden was assessed on day 7, day 14, and
day 21, using the IVIS Lumina Series III (Perkin Elmer). BLI intensity
was analyzed by Aura imaging analysis software. Blood was collected on
day 14, and blood, lung, liver, spleen, and bone marrow were collected
when the mice met the prespecified endpoints according to approved
protocols. The blood and tissues were processed and stained with
antibodies, as described above.
Flow cytometer
Flow cytometry was performed on FACS Symphony A3 cytometer or Sony
ID7000 spectral cytometer and analyzed using FlowJo software. Cell
sorting was performed using a FACS Aria 3 sorter.
RNA-seq, mapping, and analysis
NK cells isolated from 4 unrelated donors were transduced with
lentivectors expressing shRNA #1 + MSLN CAR, shRNA #1 + MSLN
CAR + PD-L1, and shRNA #1 + MSLN CAR + SCE, respectively. Untransduced
NK cells from the same donors were used as controls. NK cells were
labeled with CTV cultured in media alone or cocultured with OVCAR8
cells for 18 h in triplicates. The cocultured NK cells were purified by
FACS sorting the CTV^+ cells, and NK cells without coculture were
obtained by pipetting from the wells. For RNA-seq, around 0.5 × 106
live NK cells for each sample were processed. RNA extraction and cDNA
libraries were performed using the NEBNext Ultra II RNA Library Prep
Kit (NEB #E7770L) per the manufacturer’s instructions. Libraries were
sequenced using the Illumina NovaSeq SP 100-nt kit (illumina
#20028402). RNA-seq reads were aligned to the human genome with Salmon
(v0.14.1) using Ensembl Homo sapiens GRCh38 (release 112) transcript
annotations. The resulting counts were analyzed in R (v4.3.2) using
tximport (v1.30.0) and DESeq2 (v 1.42.1) for differential expression
analysis. The Wald test and Bonferroni post-hoc correction were used to
identify differentially expressed genes. Log fold change (LFC)
shrinkage was performed on differentially expressed genes with apeglm
LFC shrinkage algorithm^[240]56. Gene set enrichment analysis (GSEA)
was performed using fgsea (v 1.29.1), and the p-values were calculated
with weighted Kolmogorov–Smirnov statistic tests and adjusted using the
Benjamini–Hochberg correction. Gene sets were curated from the msigDB
(v 7.5.1) database. Data visualization was done with ggplot2 (v3.5.0).
Statistical analysis
The data were analyzed statistically using Prism Version 8 or SPSS
Statistics Package 25 (SPSS Inc., Chicago, USA). One-way or two-way
ANOVA with Tukey’s multiple comparisons was performed between the
indicated groups to determine the statistical difference. For the
survival of mice, the data were analyzed using Kaplan–Meier method
incorporated in Prism.
Reporting summary
Further information on research design is available in the [241]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[242]Supplementary Information^ (4.6MB, pdf)
[243]Transparent Peer Review file^ (484KB, pdf)
[244]Reporting Summary^ (120.3KB, pdf)
Source data
[245]Source Data^ (62.3KB, xlsx)
Acknowledgements