Metastatic Single Tumor Cells Evade NK Cell-mediated Killing by Thrombin-mediated Loss of the Activating Ligand CD155/PVR/Necl-5

Natural killer (NK) cells lyse invading tumor cells to limit metastatic growth in the lung, but how some cancers evade this host protective mechanism to establish a growing lesion is not known. Here we have combined ultra-sensitive bioluminescence whole body imaging with intravital two-photon microscopy involving genetically-encoded biosensors to examine this question. NK cells eliminated disseminated tumor cells from the lung within 24 hrs of arrival, but not thereafter. Intravital dynamic imaging revealed that a disseminated tumor cell in a pulmonary capillary interacts with an NK cell every 2 hrs on average. In the first 4 hrs after tumor cell arrival, 50% of such encounters lead to tumor cell death but after 24 hrs of arrival, nearly 100% of the interactions result in the survival of the tumor cell. This evasion of NK cell surveillance is mediated by thrombin-dependent loss of cell surface CD155/PVR/Necl-5, a ligand for the NK cell activating receptor DNAM-1. This loss prevents the NK cell signaling needed for effective killing of tumor targets. By quantitatively visualizing the evasion of NK cell surveillance, we have uncovered a molecular mechanism for cancer evasion and provided an explanation for the anti-metastatic effect of anticoagulants. SUMMARY Intravital functional two-photon microscopy reveals that metastatic tumor cells lodged in pulmonary capillaries acquire resistance to patrolling NK cells. Protease-mediated loss of the activating ligand CD155/PVR/Necl-5 on tumor cells prevents NK cells from ERK activation and tumor cell killing.

control mice is caused primarily by NK cells. 126 To explore the NK cell-mediated elimination of tumor cells after the early rapid phase decline (1 127 hr-8 days), we administered luciferin immediately before each round of imaging ( Figure 1B). The 128 bioluminescence signals were normalized to that at 1 hr after B16-Akaluc injection in each mouse. In 129 the control mice, the bioluminescence signal of melanoma cells reached a nadir 24 hrs after tumor 130 cell injection and increased thereafter, indicating proliferation of melanoma cells. On the other hand, 131 in the αAGM1-treated mice, the bioluminescence signal decreased very little after 4 hrs and started 132 increasing after 12 hrs. Importantly, after 24 hrs, we did not observe any significant difference in the 133 relative increase of the bioluminescence signal between the control and αAGM1-treated mice, 134 implying that NK cells eliminate disseminated melanoma cells primarily in the acute phase (< 24 135 hrs) of lung metastasis. In another model system, B16F10 melanoma cells were inoculated into the 136 foot pad two weeks prior to the intravenous injection of the B16-Akaluc cells ( Figure 1C). αAGM1 137 treatment hampered the rapid decrease of the bioluminescence signal, suggesting that NK 138 cell-mediated elimination of metastatic melanoma cells operates in the melanoma-burdened mice as 139 well. 140 We extended this approach to other syngeneic mouse tumor cell lines: Braf V600E mutated  The effect of T cell immunity on tumor elimination was examined by using BALB/c nu/nu mice. 151 The Akaluc luciferase was transduced in BALB/c-derived 4T1 breast cancer cells to generate   Table). The 205 parameters on flowing NK cells were deduced from the NK cell count in blood and the flow rate in 206 pulmonary capillaries. In short, a tumor cell lodged in a pulmonary capillary will be contacted by the 207 flowing and crawling NK cells roughly every 10 min and every 2 hrs, respectively.  other than NK cells, these additional data are consistent with our imaging data and the idea that 267 ERK-dependent NK cell activation contributes to the elimination of disseminated tumor cells.

268
With these data in hand, we next proceeded to visualize ERK activation in vivo.  Figure 5C, upper panel; Figure 5D, magenta), followed by Ca 2+ influx in the melanoma cells at 5.5 274 min and cell death at 10 min ( Figure 5C, lower panel; Figure 5D, green). ERK activation, defined by 275 a more than 30% increase in the FRET/CFP ratio, was observed within 3 min in 60 NK cells during 276 88 contact events ( Figure S8A; Figure 5E). Ca 2+ influx was observed at a median of 4 min in 43 of 277 the 60 tumor cells that came into contact with the NK cell having ERK activation ( Figure S8B; 278 Figure 5F). We did not observe Ca 2+ influx in 28 tumor cells that were touched by the NK cells that

Staining of intravascular NK cells of the bone marrow and lung 449
Mice were subjected to intravenous injection of the 3 μg αCD45 antibody 3 min before sacrifice.

596
The mice were anesthetized with 1.0% isoflurane supplied through a tracheostomy tube Surflo Bellows Falls, VT) for tdTomato fluorescence. All movies were median-filtered for noise reduction.

703
After taking images at 24 hrs, moving to the other field of view without B16-Necl-5-ScNeo cells and 704 1.0×10 6 B16-Necl-5-ScNeo cells were newly injected from tail vein. Images of newly injected 705 B16-Necl-5-ScNeo cells were acquired with the same condition with the image acquisition of 24 hrs.

706
The intensity of mScarlet and mNeonGreen in the plasma membrane was isolated by using 707 MetaMorph software and the mScarlet/mNeonGreen ratio was calculated.

709
In vitro protease digestion 710 1.0×10 5 B16F10 cells or 293T cells were resuspended in serum free RPMI and incubated for 3 hrs at 711 37℃ with 100 ug/ml thrombin. The expression level of murine Necl-5 was analyzed by FACS Aria IIu cell sorter. Data analysis was performed using FlowJo software.

714
Quantification and statistical analysis 715 The statistical differences between the two experimental groups were assessed by Welch's t-test.

716
Kaplan-Meier survival analyses were performed using MATLAB, and the log rank test was used to 717 determine significance.