The fully humanized anti-human BCMA CAR was generated and used as described (36). The human CXCR4 coding sequence was cloned downstream to the anti-human BCMA CAR sequence linked via a self-cleaving peptide P2A. The entire CAR and CXCR4 sequences were synthesized by GeneArt, and cloned into the retroviral vector MP71. For construction of a CXCR4-only expressing vector, a PCR reaction with primers to introduce a NotI and a EcoRI restriction site was run on the human CXCR4 coding sequence downstream of the aforementioned P2A site contained in a MP71 retroviral vector. The MP71 plasmid was digested with NotI and EcoRI, followed by ligation of the CXCR4 PCR sequence into the vector backbone.

The mutant form CXCR4^R334X was generated by Gibson assembly. Briefly, the MP71 plasmid BCMA CAR-CXCR4 was digested with BsteII and EcoRI. An overhang extension PCR was used to create a CXCR4-fragment with a stop codon at amino acid position 334. For construction of all other CXCR4 mutant forms, CXCR4 was cut with BsteII and EcoRI and replaced with mutant CXCR4-fragment R134N (CXCR4^{G protein-def.}) sequences. All sequences were verified by restriction digest and Sanger sequencing.

NK-92 and YTS NK cell lines cells stably transduced with various CAR constructs were seeded in 96-well U-bottom plates and mixed with a fixed number of 5x10⁴ Raji^BCMA target cells. SP6-CAR NK-92 cells were used as a negative control, and tumor cells only were used for calculation of NK-92 cell-specific killing. Co-cultures were kept for 17 hours and stopped by inclusion of ice-cold FACS buffer. Cells were resuspended in blocking buffer supplemented with Fx True stain. To distinguish effector and target cells, anti-CD56-PE-Cy7 and anti-CD19-FITC antibodies in combination with 7-AAD for dead cell exclusion were used. Samples were acquired on a FACS Canto II flow cytometer (BD Biosciences), or a MACSQuant X analyzer (Miltenyi Biotec) and further analyzed with FlowJo v. 10.0.8 software.

Bone marrow-derived cells from three MM patients were obtained freshly and further purified using a Pancoll gradient before freezing. The study was conducted according to the Declaration of Helsinki and in accordance with local ethical guidelines; written informed consent of all patients was obtained. The diagnosis of MM was made by expert Charité-University Medicine Berlin hematologists and pathologists, and it integrates cell morphology, immunohistology, marker expression in flow cytometry, and pathophysiological behavior. Thawed target cells were used in co-cultures similar to MM.1S-GFP⁺ cell line. NK-92 effector cells in co-culture were labeled with eFluor670 or with anti-CD56, and after 6 hours target MM cells were defined by CD38^high and eFluor670 ^- . A 7-AAD staining was used for live-dead cell discrimination.

1x10⁶ CAR-transduced NK-92 cells were pelleted by centrifugation at 400x g for 5 min. Cells were resuspended with BCMA protein-coupled magnetic beads (6 μg/mL; ACRO Biosystems) in PBS. A biotinylated BCMA fragment covers the N-terminal aa 1–54 that includes the cognate epitope for the scFv BCMA. BCMA-beads were added to NK-92 cells at 37°C for 5 min. Following the incubation period, the cells were placed on ice and lysed with ice-cold 2X RIPA buffer (300 mM NaCl, 100 mM Tris-HCl, 2% NP-40, 1% C₂₄H₃₉NaO₄, 0.2% SDS) containing protease and phosphatase inhibitors. The lysate was incubated on ice for 30 min and then centrifuged at 13,000 g for 20 min. Supernatant was processed for immunoblot analysis.

To quantitate the kinetics of BCMA-CAR surface downregulation, CAR-transduced NK-92 cells were seeded in 1 ml of RPMI/0.5% BSA medium and kept at 37°C in a waterbath. 100 μl BCMA protein-coupled magnetic beads (Acro Biosystems) were washed, resuspended in medium and added to the cell suspension at 37°C. Cells were incubated for 10 min, centrifuged and resuspended in medium at 37°C. Aliquots were taken and added to an equal volume of ice-cold FACS buffer (PBS, 0.5% BSA, 0.05% NaN₃). Cell suspensions were immediately centrifuged at 4°C and resuspended in FACS buffer. For detection of CARs, a PE-coupled goat anti-human IgG antibody (Southern Biotech) was used. Cells were then centrifuged and resuspended in fixation buffer (BioLegend), followed by analysis on a FACS Canto II instrument.

The recruitment of voluntary blood and bone marrow donors was conducted according to the declaration of Helsinki and in accordance with local ethical guidelines. The study was approved by the responsible ethic committee at Charité-University Medicine Berlin (EA2/142/20).

To gauge the migration of NK cells towards tissue-specific chemoattractants, we measured expression levels of chemokine receptors CXCR3, CXCR4 and CCR6 and adhesion molecule CD62L on the clinically relevant cell line NK-92. NK-92 cells were devoid of CCR6 and CD62L, possessed negligible levels of CXCR4 and ubiquitously expressed high levels of CXCR3 ( Supplementary Figures S1A, B ).

To assess if ectopically expressed CXCR4 coordinates chemotaxis of BCMA CAR-expressing NK cells towards CXCL12, we developed a BCMA CAR retroviral vector (CAR) (36) equipped with either a wildtype (WT), CXCR4 (CAR-CXCR4) or mutant CXCR4 (CAR-CXCR4^R334X) chemokine receptor sequence ( Figure 1A ). The CXCR4^R334X sequence is a C-terminally-truncated, gain-of-function mutant with increased signal transduction, reduced desensitization, and impaired receptor internalization (37–41). Retroviral transduction of NK-92 cells with either the CAR, CAR-CXCR4 or CAR-CXCR4^R334X construct induced comparable levels of BCMA CAR expression ( Figures 1B, C ). Substantially higher levels of CXCR4 were detected on NK-92 cells transduced with the bicistronic CAR-CXCR4 or CAR-CXCR4^R334X constructs as compared to CAR NK-92 cells ( Figure 1D ). Furthermore, CXCR4 expression was greater on CAR-CXCR4^R334X NK-92 cells than CAR-CXCR4 NK-92 cells ( Figures 1D, E ).

Next, we generated CAR and CAR-CXCR4 primary NK cells (pNK; CD3^- CD56⁺) from peripheral blood ( Figure 1F ). These cells were transduced and expanded in the presence of IL-2 and IL-15 using ( Supplementary Figure S1C ) anti-CD2/CD355 MACS activator beads. On Day 9 (2-days post-transduction), gating on the CD56⁺CD16^high population, we observed substantially higher levels of CXCR4 expression on CAR-CXCR4 and CAR-CXCR4 R334X transduced pNK cells ( Figure 1F ). CXCR4 expression also increased modestly in CAR pNK cells, indicating endogenous CXCR4 upregulation was not induced by viral infection (12, 23, 41). Co-modified CAR-CXCR4 pNK cells expressed CXCR4 at two- to threefold higher levels (gMFI) when compared to pNK cells transduced with the conventional CAR ( Figures 1F, G ).

We functionally characterized the activity of ectopically-expressed CXCR4 by quantitating phosphorylation of downstream signaling modules in response to a 5 min CXCL12 exposure. CXCL12 addition triggered phosphorylation of p44/42 MAPK (ERK1,2) ( Figure 1H ; Supplementary Figure S1D ) in CAR-CXCR4 and CXCR4^R334X NK-92 cells to a larger extent than in CAR NK-92 cells. We further confirmed AKT phosphorylation in CAR-CXCR4 NK-92 cells ( Supplementary Figures S1E, F ) after 5 min CXCL12 co-incubations (42).

CAR-CXCR4 co-expression led to fourfold improved chemotaxis towards CXCL12 (25 ng/ml) compared to NK-92 cells expressing the conventional CAR ( Figure 1I ). Along these lines, both CAR-CXCR4 and CAR-CXCR4^R334X pNK cells featured a higher migratory ability towards CXCL12 than CAR pNK cells ( Figure 1J ). For context, serum concentrations in MM patients are reportedly in the range of 453 ± 124 pg/mL (43). Despite a modest upregulation of endogenous CXCR4 in pNK cells ( Figure 1F ), they retained an inefficient ability to migrate towards CXCL12 at the concentrations indicated. Collectively, incorporation of either a WT or mutant CXCR4 sequence into a BCMA-CAR-encoding retroviral vector leads to the expression of a fully-functioning protein capable of instigating enhanced migration towards CXCL12 in vitro.

Previous reports in T cells have shown that recruitment of chemokine receptors within the IS enhances TCR-induced signaling and IS longevity (33, 35). Canonically, these co-stimulatory effects have been attributed to a chemokine stimulus acting on the cognate chemokine receptor. However, the conditions for non-conventional CAR-dependent signaling processes in NK cells are unknown. Hence, we first investigated whether enforced CXCR4 expression boosts the antigen-dependent cytolytic efficacy of BCMA CAR-NK cells. CAR, CAR-CXCR4, CAR-CXCR4^R334X and SP6-control (irrelevant CAR binder) NK-92 cells were co-cultured with an engineered BCMA-expressing Raji target cell line (Raji^BCMA) ( Supplementary Figure S2 ). The proportion of viable Raji^BCMA cells remaining in culture after 17 hours was quantified by flow cytometry and showed that CAR-CXCR4 and CAR-CXCR4^R334X NK-92 cells possessed a 15–25% higher antigen-dependent killing capacity compared to CAR NK-92 cells at lower effector-to-target ratios (E:T: 0.125:1) ( Figures 2A, B ). To exclude effector cell type-specific peculiarities, we conducted analogous experiments with the NK cell line YTS (44) ( Supplementary Figure S3A ). Indeed, CXCR4-equipped CAR YTS cells exhibited a higher killing capacity against Raji^BCMA cells compared to conventional CAR YTS cells ( Supplementary Figures S3B, C ).

We next performed a CD107a (LAMP-1) secretory lysosome degranulation assay to scrutinize whether CXCR4 overexpression endows BCMA CAR-NK-92 cells with enhanced antigen-dependent cytolytic efficacy (45, 46). Following 30 min co-cultures with Raji^BCMA cells, we detected substantially higher levels (approximately 20%) of LAMP-1 expression on the surfaces of CAR-CXCR4 and CAR-CXCR4^R334X NK-92 cells ( Figures 2C, D ). This difference was diminished after 60 min, indicating that CXCR4 supports a rapid mobilization of lytic granules following CAR activation ( Figure 2D ).

Whereas previous reports have postulated a synergistic, mutually-reciprocal state of TCR- and chemokine-induced receptor activation (47, 48), others have alluded to a state whereby CXCL12-induced signals do not override those emanating from TCR activation (49, 50). Consistent with the latter, we found that CXCL12 supplementation did not enhance antigen-dependent degranulation among CAR-CXCR4 or CAR-CXCR4^R334X NK-92 cells ( Figures 2E–G ). This was, in part, because CAR NK cells lacking CXCR4 overexpression exhibited similar low-level gain-of-function ( Figures 2F, G ). We also generated a CXCR4 mutant with a defect in G protein coupling (R134N) in the conserved DRY motif (CAR-CXCR4^{G protein-def.}) ( Figures 2E–G ). This mutation does not affect downstream PI3K-activation, tyrosine phosphorylation of focal adhesion proteins and chemotaxis (51). When linked to the BCMA CAR in our bicistronic constructs, we confirmed that CAR-CXCR4, CAR-CXCR4^R334X and CAR-CXCR4^{G protein-def.} NK-92 cells maintained a superior antigen-dependent degranulation capacity ( Figures 2F, G ). Our observations imply a hierarchy in which the enhanced killing capacity originates from CAR-driven activation of CXCR4 signaling nodes.

To rule out an artifact of enhanced BCMA expression in the transfected Raji^BCMA cell line, we compared the antitumor cytolytic activity of CAR and CAR-CXCR4 NK-92 in co-cultures with the MM cell line MM.1S and with four patient-derived primary MM specimen. In agreement with the improved cytolytic activity against Raji^BCMA cells, CXR4 co-expression endowed CAR NK-92 cells with an enhanced cytolytic capacity against both cell types exhibiting endogenous BCMA expression ( Figures 3A–D ). CAR-CXCR4 NK-92 cells gained a 25% higher antigen-dependent killing capacity for MM.1S cells compared to CAR NK-92 cells, visible at lower effector-to-target ratios (E:T: 0.125:1) ( Figure 3B ). The FACS-based killing assay was corroborated by a degranulation assay, using the MM.1S cell line as a stimulus ( Supplementary Figures S4A, B ). Similar to the MM.1S cell line, CXCR4 co-expression endowed CAR NK-92 cells with much stronger cytolytic capacity against primary MM cells ( Supplementary Figure S4C ) at low effector to target ratios (E:T: 0.25:1), which even exceeded the gain-of-function compared to Raji^BCMA and MM.1S cells ( Figures 3C, D ). Notably, the frequencies of MM cells within patient-derived primary bone marrow aspirates are much lower than homogenous target cell populations as for Raji^BCMA and MM.1S cell lines.

To exclude a CAR-independent response of chemokine receptor overexpression, we also generated NK-92 cells expressing CXCR4 only. We studied the innate killing capacity of CXCR4 NK-92 cells in a co-culture with MM.1S cells. Isolated CXCR4 overexpression was not sufficient to endow NK-92 cells with enhanced degranulation capacity. Their innate reactivity towards MM.1S tumor cells was similar to NK-92 cells only (mean LAMP-1 upregulation: 2.3%; CXCR4 NK-92: mean 3.4%), and substantially weaker than CAR NK-92 cells. This result indicated that the gain of cytotoxic capacity in NK-92 cells depends on a co-stimulatory function of the chemokine receptor in the process of BCMA CAR activation ( Supplementary Figures S4D–F ).

To dissect the mechanisms underlying the increased cytolytic activity of CAR-CXCR4 NK-92 cells, we focused on CAR stabilization and the initiation of proximal or distal signaling. Because reorganization and accumulation of chemokine receptors within the IS of conventional T cells with antigen-presenting cells (APC) has been shown to stabilize the IS (33, 35), we explored whether ectopic CXCR4 overexpression affects BCMA-CAR internalization during antigen-engagement. CAR and CAR-CXCR4 NK-92 cells were exposed to cognate BCMA peptide-coated magnetic beads (BCMA-beads) and examined for the surface expression of the BCMA-CAR over-time. BCMA-beads offer the operational advantage of selectively binding to the BCMA-CAR cognate receptor. Although both CAR and CAR-CXCR4 NK-92 cells progressively lost BCMA-CAR surface expression after BCMA-bead exposure, CAR-CXCR4 NK-92 cells prolonged BCMA-CAR surface expression to a greater extent than CAR NK-92 cells ( Figures 4A, B ). BCMA-bead stimulation did not affect CXCR4 surface expression, but on the contrary, CXCL12 induced internalization of its cognate receptor ( Figures 4C, D ). These observations suggest that the integrity of the BCMA CAR-driven IS is maintained in the presence of abundant CXCR4. Exposing CAR- and CAR-CXCR4 NK-92 cells to BCMA-beads produced no striking difference in IFN-γ secretion ( Supplementary Figure S5 ).

Recently, trogocytosis induced by CAR-activation on NK cells was associated with higher levels of degranulation towards target cells (52). To assess the degree of trogocytosis induced by CAR engagement, we quantified the ratio between CD56⁺/CD19⁺ double-positive (TROG⁺) and CD56⁺/CD19^- (TROG^-) BCMA CAR-expressing NK-92 cells in response to CD19⁺ Raji^BCMA target cell exposure. CAR-CXCR4 (mean TROG⁺/TROG^-=1.04 ± 0.59) and CAR-CXCR4^R334X (1.17 ± 0.53) NK-92 cells triggered trogocytosis to a substantially higher degree than CAR NK92 cells (0.79 ± 0.47) ( Figures 4E, F ). We then asked whether trogocytosis modulated effector function of CAR-CXCR4 NK-92 cells by assessing the level of degranulation. Whereas minimal levels of LAMP-1 expression were detected on CAR-CXCR4^TROG- NK92 cells (mean % ΔLAMP-1+ cells= 3.39 ± 2.14), a larger proportion of CAR-CXCR4^TROG+ NK92 cells had undergone degranulation (64.02 ± 8.74) ( Figures 4G, H ). Prolonged expression of the BCMA CAR at the cell surface and higher levels of trogocytosis following target cell challenge suggest that CAR-CXCR4 co-modification equips NK92 cells with a substantially enhanced effector function.

CAR engagement results in the formation of a nonclassical, disorganized IS characterized by the rapid initiation of proximal signaling, and quicker recruitment of cytotoxic granules (53–55). It was recently determined that the sensitivity of a CAR-T cell towards antigen is approximately 1000-times reduced compared to that of conventional antiviral T cells, at least in part due to inefficient recruitment of ZAP-70 to the engaged CAR (56). Considering that chemokine receptor signal transduction relies on ZAP-70 (48, 57) recruitment and furthermore, promotes chemokine-enrichment in T cell synapses, we asked whether enforced CXCR4 during CAR engagement would augment ZAP-70 recruitment. We visualized the presence of phosphorylated ZAP-70 (P-ZAP-70) within the synapse of engaged CAR, CAR-CXCR4 and CAR-CXCR4^R334X NK-92 cells following a short co-culture with Raji^BCMA cells ( Figures 5A, B ). Thereafter, we quantified the expression (MFI) of P-ZAP-70 within IS. These data showed that P-ZAP-70 expression was significantly higher within the synapse of CAR-CXCR4 and CAR-CXCR4^R334X NK-92 cells compared to CAR-NK-92 cells, which was most pronounced after 5 min ( Figure 5B ). After 20 min, synaptic P-ZAP-70 MFI values in CAR NK-92 cells were comparable to those obtained in CAR-CXCR4 NK-92 cells, further suggesting that CXCR4 supports the initial events triggered by CAR engagement. Of note, CXCR4 expression in Raji^BCMA target cells interfered with the quantitation of synaptic CXCR4 recruitment in APC-conjugated CAR NK cells.

Immunoblot analysis of protein lysates derived from short-term (5 min) BCMA-bead stimulated NK-92 cells revealed a significant enhancement of pZAP-70 expression in CAR-CXCR4 and CAR-CXCR4^R334X modified NK-92 cells ( Figures 5C, D ). We further inspected the signaling kinetics of antigen-engagement over-time by comparing induction of ERK1,2, which occurs downstream of ZAP-70 activation. After only one min exposure to BCMA-beads, ERK1,2 phosphorylation was significantly increased in CAR-CXCR4 NK-92 cells compared to CAR NK-92 cells. We conclude that BCMA-CAR activation is more potent when NK-92 cells are equipped with CXCR4 ( Figures 5E, F ).

CXCR4 expression could influence the formation of mature IS by enabling increased integrin inside-out (33) signaling. Using a conjugate formation assay we quantitated stable IS formation between CAR NK effector cells and Raji^BCMA target cells. All NK-92 CAR constructs were devoid of integrin LFA-1 (open conformation) and adhesion molecule CD62L, but ubiquitously expressed CD11b and CD44 ( Figure 6A ). Conjugate formation after 30 min was indistinguishable between those cell types, addition of CXCL12 had no stimulatory impact in this process ( Figures 6B, C ). Potentially, conjugate formation could occur earlier than 30 min, which would be consistent with an earlier activation of the signaling cascade when CXCR4 is overexpressed. This was not further investigated herein.

Inadequate CAR reactivity towards low-density cognate antigen is a major source of treatment failure (58). Insufficient recruitment of ZAP-70 to the engaged CAR has emerged as an important cause of reduced CAR sensitivity (56). Because enforced CXCR4 expression enhances the presence of activated ZAP-70 within the IS ( Figures 5A–D ), we examined whether the presence of CXCR4 equips BCMA-CAR NK cells with an improved ability to recognize target cells with lower antigen amounts. We exposed CAR and CAR-CXCR4 NK-92 cells to lowering amounts of BCMA-beads and assessed the induction of downstream ERK1,2 signaling. CAR-CXCR4 NK-92 cells recognized and transmitted signals when antigen availability was lower compared to its conventional CAR counterpart ( Figures 6D, E ). More specifically, at a BCMA-bead concentration of 5 and 1.67 µL, CAR-CXCR4 NK-92 cells were still able to trigger ERK1,2 phosphorylation ( Figure 6D ). These data suggest that enhanced ZAP-70 recruitment at the IS, afforded by the presence of synthetic CXCR4, would enable CAR-CXCR4 NK-92 cells to eliminate lower-antigen-expressing tumor cells.