CD56 as a marker of an ILC1-like population with NK cell properties that is functionally impaired in AML.

(1)

Article

Reference

CD56 as a marker of an ILC1-like population with NK cell properties that is functionally impaired in AML.

SALOMÉ, Bérengère, et al.

Abstract

An understanding of natural killer (NK) cell physiology in acute myeloid leukemia (AML) has led to the use of NK cell transfer in patients, demonstrating promising clinical results.

However, AML is still characterized by a high relapse rate and poor overall survival. In addition to conventional NKs that can be considered the innate counterparts of CD8 T cells, another family of innate lymphocytes has been recently described with phenotypes and functions mirroring those of helper CD4 T cells. Here, in blood and tissues, we identified a CD56+ innate cell population harboring mixed transcriptional and phenotypic attributes of conventional helper innate lymphoid cells (ILCs) and lytic NK cells. These CD56+ ILC1-like cells possess strong cytotoxic capacities that are impaired in AML patients at diagnosis but are restored upon remission. Their cytotoxicity is KIR independent and relies on the expression of TRAIL, NKp30, NKp80, and NKG2A. However, the presence of leukemic blasts, HLA-E-positive cells, and/or transforming growth factor-β1 (TGF-β1) strongly affect their cytotoxic potential, at least partially by reducing the [...]

SALOMÉ, Bérengère, et al . CD56 as a marker of an ILC1-like population with NK cell properties that is functionally impaired in AML. Blood Advances , 2019, vol. 3, no. 22, p. 3674-3687

PMID : 31765481

DOI : 10.1182/bloodadvances.2018030478

Available at:

http://archive-ouverte.unige.ch/unige:132570

Disclaimer: layout of this document may differ from the published version.

(2)

REGULAR ARTICLE

CD56 as a marker of an ILC1-like population with NK cell properties that is functionally impaired in AML

B éreng ère Salom é,^1,2Alejandra Gomez-Cadena,^1,3,* Romain Loyon,^1,* Madeleine Suffiotti,^1,3,4Valentina Salvestrini,⁵Tania Wyss,^1,3,4 Giulia Vanoni,^1,3Dan Fu Ruan,²Marianna Rossi,⁶Alessandra Tozzo,⁷Paolo Tentorio,⁸Elena Bruni,^8,9Carsten Riether,¹⁰

Eva-Maria Jacobsen,¹¹Peter Jandus,¹²Curdin Conrad,^13,14Manfred Hoenig,¹¹Ansgar Schulz,¹¹Katarzyna Michaud,¹⁵

Matteo Giovanni Della Porta,^6,16Silvia Salvatore,⁷Ping-Chih Ho,^1,3David Gfeller,^1,3,4Adrian Ochsenbein,¹⁰Domenico Mavilio,^8,9 Antonio Curti,¹⁷Emanuela Marcenaro,¹⁸Alexander Steinle,¹⁹Amir Horowitz,²Pedro Romero,¹Sara Trabanelli,^1,3,^†and Camilla Jandus^1,3,^†

1Department of Oncology UNIL CHUV, University of Lausanne, Lausanne, Switzerland;²Department of Oncological Sciences, Precision Immunology Institute, Icahn School of Medicine at Mount Sinai, New York, NY;³Ludwig Institute for Cancer Research, University of Lausanne, Lausanne, Switzerland;⁴Swiss Institute of Bioinformatics, Lausanne, Switzerland;⁵Department of Specialistic, Diagnostic and Experimental Medicine, Institute of Hematology“Ser `agnoli,”University Bologna, Bologna, Italy;⁶Cancer Center IRCCS Humanitas Research Hospital, Milan, Italy;⁷Pediatric Department, University of Insubria, Varese, Italy;⁸Unit of Clinical and Experimental Immunology, Humanitas Clinical and Research Center, Rozzano-Milan, Italy;⁹Department of Medical Biotechnologies and Translational Medicine, University of Milan, Milan, Italy;¹⁰Department of Medical Oncology, Inselspital, Bern University Hospital, University of Bern, Bern, Switzerland;¹¹Department of Pediatrics, University Medical Center Ulm, Ulm, Germany;¹²Division of Immunology and Allergology, Department of Medicine, University Hospital and Medical Faculty, Geneva, Switzerland;¹³Department of Dermatology, University Hospital CHUV, Lausanne, Switzerland;¹⁴Faculty of Biology and Medicine, University of Lausanne, Lausanne, Switzerland; ¹⁵University Center of Legal Medicine Lausanne-Geneva, Lausanne University Hospital, University of Lausanne, Lausanne, Switzerland;¹⁶Department of Biomedical Sciences, Humanitas University, Milan, Italy;¹⁷Institute of Hematology L. e A. Ser `agnoli, Department of Hematology and Oncology, Azienda Ospedaliero-Universitaria S. Orsola-Malpighi, Bologna, Italy;¹⁸Department of Experimental Medicine and Centre of Excellence for Biomedical Research, University of Genoa, Genoa, Italy; and¹⁹Institute for Molecular Medicine, Goethe University Frankfurt, Frankfurt, Germany

Key Points

•Human ILC1-like cells kill tumors in a KIR- independent manner.

•The cytotoxicity of human ILC1-like cells is impaired in AML at diagnosis but is restored in remission.

An understanding of natural killer (NK) cell physiology in acute myeloid leukemia (AML) has led to the use of NK cell transfer in patients, demonstrating promising clinical results. However, AML is still characterized by a high relapse rate and poor overall survival. In addition to conventional NKs that can be considered the innate counterparts of CD8 T cells, another family of innate lymphocytes has been recently described with phenotypes and functions mirroring those of helper CD4 T cells. Here, in blood and tissues, we identiﬁed a CD56¹innate cell population harboring mixed transcriptional and phenotypic attributes of conventional helper innate lymphoid cells (ILCs) and lytic NK cells. These CD56¹ILC1-like cells possess strong cytotoxic capacities that are impaired in AML patients at diagnosis but are restored upon remission. Their cytotoxicity is KIR independent and relies on the expression of TRAIL, NKp30, NKp80, and NKG2A. However, the presence of leukemic blasts, HLA-E–positive cells, and/or transforming growth factor-b1 (TGF-b1) strongly affect their cytotoxic potential, at least partially by reducing the expression of cytotoxic-related molecules. Notably, CD56¹ILC1-like cells are also present in the NK cell preparations used in NK transfer–based clinical trials. Overall, we identiﬁed an NK cell–related CD56¹ILC population involved in tumor immunosurveillance in humans, and we propose that restoring their functions with anti-NKG2A antibodies and/or small molecules inhibiting TGF-b1 might represent a novel strategy for improving current immunotherapies.

Introduction

Acute myeloid leukemia (AML) is the most common acute leukemia in adults, with a 3.7/100 000 incidence per year. AML has a high relapse rate, which decreases patients’5-year overall survival to

Submitted 23 December 2018; accepted 10 October 2019. DOI 10.1182/

bloodadvances.2018030478.

*A.G.-C. and R.L. contributed equally to this study as joint second authors.

†S.T. and C.J. contributed equally to this study as joint last authors.

The data for this study have been deposited in the European Nucleotide Archive at EMBL-EBI under accession number PRJEB34980 (https://www.ebi.ac.uk/ena/data/

view/PRJEB34980).

The full-text version of this article contains a data supplement.

Downloaded from https://ashpublications.org/bloodadvances/article-pdf/3/22/3674/1543973/advances030478.pdf by BIBLIOTHEQUE UNIVERSITAIRE DE MEDECINE user on 11 March 2020

(3)

19%.¹The conventional treatments consist of chemotherapy or allogeneic hematopoietic stem cell transplantation.² Moreover, natural killer (NK) cell transfer therapy has been developed and provides good outcome improvement if the donor and recipient are KIR mismatched.^3-6

In addition to conventional NKs (cNKs), another lymphocytic innate cell family has recently been identified and named innate lymphoid cells (ILCs). ILCs constitutively express the interleukin-7 (IL-7) receptorachain (CD127) and are deprived of somatically rearranged antigen-specific receptors and common lineage markers. Whereas cNKs functionally mirror adaptive CD8 T cells, conventional ILCs are considered the innate counterpart of helper CD4 T cells⁷; ILCs secrete pro- or anti-inflammatory cytokines upon sensing the microenvironment and“help”effector cells.^7-11

Despite the clear-cut ILC subset delineation, unexpected phenotypic and functional heterogeneity within NK and ILC subsets has recently been reported,^12-15opening novel opportunities for innate cell-based immunotherapies.

Here, we describe an unconventional human ILC1-like cell population with cytotoxic properties that expresses the ILC marker CD127 and CD56^16,17but lacks CD16 and c-Kit (CD117) expression. These CD56¹ILC1-like cells are related to the stage 4b (S4b) NK cells.

Their cytolytic mechanism is KIR independent but requires NKp80, NKp30, and TRAIL engagement to lyse both major histocompatibility complex class I (MHCI) positive and negative targets. Similar to previous reports of conventional ILCs^18,19 and NKs,²⁰ the frequency and functions of CD56¹ILC1-like cells are impaired in AML patients. At diagnosis, CD56¹ ILC1-like cells are significantly reduced, and their killing capacity is defective due to the persistence of NKG2A expression, the inability to release cytotoxic mediators, and the downregulation of NKp80, NKp30, and TRAIL, which is at least partially mediated by transforming growth factor-b (TGF-b). Notably, during remission, the cytotoxic machinery and the receptors’ expression on CD56¹ ILC1-like are completely restored.

Overall, we propose that this CD56¹ ILC1-like cell population represents an attractive target for immunomodulatory drugs, such as anti-NKG2A antibodies and TGF-bRI inhibitors, in AML patients.

Given the presence of these cells in NK-cell preparations used for adoptive transfer, exploiting their properties might provide a powerful approach for maximizing the efficacy of KIR-mismatch independent immunotherapy.

Methods

All the methods used in this article are described as supplemental Information.

Results

CD56¹CD16²ILC1-like cells have NK properties and are impaired in AML patients at diagnosis

We recently reported that the ILC1 compartment is numerically and functionally impaired in AML patients at diagnosis.¹⁹Here, we identify a CD16²CD127¹c-Kit²CRTH2²CD56¹cell population, which falls in the ILC1 gate (Figure 1A-B).²¹t-t-SNE analysis based on CD127, CD56, CD16, CRTH2, and c-Kit expression is sufficient to clearly discriminate this cell population from conventional ILC (ILC1, ILC2, and “ILC progenitors” [ILCP]²²) and NK subsets

(Figure 1C). In particular, their CD56^dim CD16² CD127¹ c-Kit² phenotype distinguishes them from the conventional CD56^brightand CD56^dimNK subsets (Figure 1D). Of note, CD56¹ILC1-like cells do not express CD49a (supplemental Figure 1).

In AML patients at diagnosis (n 5 60, [35 to 97 years old];

supplemental Table 1), the proportions of this population among lymphocytes are strongly reduced (Figure 1E-F), independently of the patients’age (data not shown). The CD56^dimNK compartment is also impaired, whereas the CD56^brightNK cell proportions are comparable between the patients and healthy donors (HDs) (Figure 1E-F).

In the HDs (n547, median age 48, interquartile range 31 to 64), CD56¹ILC1-like cells represent 38.5% of all Lineage²CD127¹ cells in the peripheral blood (range 9.4% to 69.0%; Figure 1G).

Similar to the cNK frequencies, which are known to be influenced by age,^23-25compared with the other“helper”ILCs, we observed a significant CD56¹ ILC1-like increase in the elderly donors (Figure 1H). The analysis of ILC frequencies within lymphoid and nonlymphoid organs from HDs and cadavers shows that CD56¹ ILC1-like cells are also present in those tissues (Figure 1I).

We then evaluated the CD56¹ ILC1-like cytokine profile in response to IL-12, IL-15, and IL-18 stimulation (n54). When stimulated, these cells secrete interferon-gand IL-8, the latter distinguishing them from the CD56^dim NKs (supplemental Figure 2A-B).

RNA sequencing reveals a CD56¹ILC1-like cell transcriptomic signature

To investigate specific CD56¹ILC1-like cell features, we performed an RNA-sequencing analysis of highly pure ILC and cNK subsets from the peripheral blood of HDs (n53) (supplemental Figure 3A-B).

In a principal component analysis of the ILC and NK subsets for their expression of cNK markers, CD56¹ILC1-like cells appear to be more closely related to CD56^brightNKs and ILCP than ILC1, ILC2, and CD56^dimNKs (Figure 2A; supplemental Figure 3C-D).

We then analyzed their transcription factor profile at RNA level.

These cells express similar levels of TBX21 (T-bet), EOMES, and RUNX3 transcripts as cNKs. Their expression of ZBTB16 (promyelocytic leukemia zinc finger protein [PLZF]) and aryl hydrocarbon receptor (AHR) transcripts is intermediate compared with CD56^bright and CD56^dim NKs. However, CD56^dim NKs express higher levels of NFIL3 transcripts than CD56¹ ILC1-like cells (Figure 2B). Furthermore, a principal component analysis reveals that the expression of PLZF, GATA3, Eomes, RORgt, and T-bet at protein level is sufficient to discriminate CD56¹ ILC1-like cells from the other ILC and NK subsets (supplemental Figure 3E-H).

Then, we analyzed the 100 genes with the lowest or highest log fold change between the CD56¹ ILC1-like cells and conventional ILC/NK subsets (Figure 2C). The CD56¹ILC1-like cells display higher levels of NK-associated transcripts than ILC1, ILC2, and ILCP (eg, EOMES, KLRD1, KLRC2, and KLRF1). CD56¹ILC1-like cell transcriptomic signature includes genes encoding cytokine receptors and chemokine receptors. By associating Gene Ontology (GO) pathways to each of the differentially expressed genes, we observed that they are mostly involved in lymphocyte activation (GO

(4)

B

Lineage with CD16

CD127

CD56+

ILC1-like

ILC1

ILCs c-Kit

CRTH2

CD56

FS-H ILC2

ILCP

A

CD56

CD16

CD56bright NK CD56dim NK

FS-H

Lineage without CD16

D

CD56bright NK ^{CD56dim NK}

CD56+ ILC1-like ILC1

c-Kit

CD127

CD56dim NK CD56bright NK

CD56+ ILC1-like ILC1

CD56

CD16

C

50

–25

–50

–40 –20 0 20

ILC1 ILC2 ILCP

CD56bright NK CD56dim NK CD56+ ILC1-like

t-SNE 1

t-SNE 2

25

0

E

CD56

FS-H CD56+

ILC1-like

ILC1 CD56

CD16 CD56bright NK

CD56dim NK

F

% of lymphocytes

1.0 ****

0.8 0.6 0.4 0.2 0.0

HD AML

% of lymphocytes

ns ****

HD AML HD AML 20

15 10 5 3 2 1 0

HD AML CD56bright NK CD56dim NK CD56+ ILC1-like

H

60

40

20

0

0 20 40 60 80

Age

% of ILCs

ILC1

ILC2 ILCP

CD56+ ILC1-like p=0.0015

ns

p=0.0441 ns

G

100 80 60 40 20 0

ILC1 ILC2 ILCP CD56+

ILC1-like

% of ILCs

2.0 HD

1.5

1.0

0.5

0.0

Total ILCs

% of lymphocytes

HD

Figure 1.Identification of a CD56¹ILC1-like population with NK properties that is impaired in AML patients at diagnosis.(A-B) Representative density plots of the extracellular flow cytometry panel used to identify the cNK cell subsets (A) and the ILC subsets (B) in peripheral blood (PB) mononuclear cells (PBMCs; ILC1 as CRTH2² c-Kit²CD56², ILC2 as CRTH2¹c-Kit^1/2CD56^1/2, ILCP as CRTH2²c-Kit¹CD56^1/2, and cNKs as CD16²CD56^brightand CD16¹CD56^dim).^21,22Lineage markers used

(5)

pathway GO0046649), immune effector process (GO0002252), leukocyte migration (GO0050900), and metabolism (GO0008152).

To confirm that CD56¹ILC1-like cells display a specific metabolism compared with the conventional ILC and NK subsets, we evaluated their nutrient uptake and mitochondrial activity (supplemental Figure 4). The CD56¹ ILC1-like cells display a lower glucose analog (2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deox- yglucose) uptake than the ILC2 and cNKs and a higher fatty acid uptake than the CD56^dim NKs. To assess their specific mitochondrial activity, we evaluated the ratio of the uptake of MitoTracker deep red to that of MitoTracker green.²⁶The ratio in CD56¹ILC1-like cells is higher than in the ILC1 and tend to be higher compared with cNKs. Hence, the distinct metabolic transcriptional pattern of CD56¹ILC1-like cells is reflected by their specific nutrient uptake and mitochondrial activity.

CD56¹ILC1-like cells are cytotoxic effectors regulated by the NKp30, NKp80, TRAIL, and HLA-E pathways

Based on the CD56¹ILC1-like cell-specific transcriptomic signature involving genes participating in immune effector processes and lymphocyte activation, we directly evaluated the CD56¹ILC1-like cell cytotoxic activity (Figure 3). CD56¹ILC1-like cells express NKp30, NKp80, CD94/NKG2A, and DNAM-1 at a high level, whereas low levels of NKp44, NKp46, and NKG2C were observed (Figure 3A).

TRAIL expression is comparable between the CD56¹ILC1-like cells and CD56^brightNKs but varies depending on the donor’s age (data not shown). Subsequently, we aimed to ascertain whether CD56¹ ILC1-like cells contain death-inducing mediators. We observed a significant CD56¹ILC1-like cell production of granzymes A, B, K, and M, perforin and granulysin, with the highest levels measured for granzyme A and M (Figure 3B; supplemental Figure 5A). Next, we cocultured CD56¹ ILC1-like cells with

the standard NK-sensitive MHCI² K562 cell line (Figure 3C).

After 4 hours, CD56¹ ILC1-like cells upregulate CD107a, demonstrating their ability to degranulate. To confirm their cytotoxicity, we analyzed CD56¹ILC1-like cell-mediated target lysis using ⁵¹Cr-release assays. CD56¹ ILC1-like cells lyse K562 to an extent comparable to that of cNKs, whereas the helper ILCs fail to induce any target cell lysis (Figure 3D). In accordance with the absence of KIR receptors, CD56¹ILC1- like cells are also able to lyse the MHCI¹BJAB and U937 cell lines (Figure 3E; supplemental Figure 5B). To investigate whether DNAM-1, NKp30, NKp80, and TRAIL, which are expressed at the highest levels by CD56¹ILC1-like cells (Figure 3A), are involved in cytotoxicity, we cocultured CD56¹ILC1-like cells with targets in the presence of specific blocking reagents (Figure 3F). The addition of anti–DNAM-1 blocking antibodies does not affect the CD56¹ ILC1-like cell killing potential. However, the presence of anti-NKp30 blocking antibodies strongly reduces their killing activity. We also confirmed that TRAIL is involved in their cytotoxicity as the CD56¹ ILC1-like cell-mediated killing of the TRAIL-sensitive BJAB line^27-29was decreased in the presence of a TRAIL decoy receptor. Subsequently, we cocultured the CD56¹ILC1-like cells with the U937 line, which expresses the ligand for NKp80 (ie, Activation-Induced C-type Lectin (supplemental Figure 5C). The addition of NKp80 blocking antibodies decreases the CD56¹ ILC1-like cell-killing capacity. Finally, based on the high expression of CD94/NKG2A by CD56¹ ILC1-like cells (Figure 3A), we compared the CD56¹ ILC1-like cell-mediated lysis of wild-type (HLA-E negative) and HLA-E–transfected (HLA-E¹) 721.221 tumor cells (Figure 3G;

supplemental Figure 4D). CD56¹ ILC1-like cell cytotoxicity is impaired when the targets express HLA-E. Overall, we show that CD56¹ILC1-like cells display high cytotoxicity triggered by the NKp30, NKp80, and TRAIL pathways and inhibited upon HLA-E binding.

I

60

40

20

0

% of ILCs

% of CD56+ ILC1-like in PB

thymus tonsils spleen BM LN colon lung bladder kidney skin

Figure 1.(Continued).for the“helper”ILC staining include CD3, CD4, CD8, CD14, CD15, CD16, CD19, CD20, CD33, CD34, CD203c, and FceRIa; same lineage markers, except for CD16, were used for the cNK staining. (C) CD127, CD56, CD16, CRTH2, c-Kit fluorescence intensity on ILC and NK subsets were concatenated from 5 HDs and analyzed witht-distributed stochastic neighbor embedding (t-SNE). c-Kit, CD127, CD56, CD16 expression levels on ILC1, CD56¹ ILC1-like cells and NKs are represented in panel D. Representative gating (E) and quantification of CD56¹ILC1-like cell and cNK proportions among lymphocytes in blood from HDs and AML patients at diagnosis (F) (CD56¹ILC1-like cells: HD, n547; AML patients: n560; cNKs: HD: n512, AML: n518). (G)Summary of the results of the total ILC proportions and ILC1, ILC2, ILCP, and CD56¹ILC1-like cell subset frequencies among the total ILCs in HD peripheral blood (N547, age median 48, interquartile range 31 to 64). (H)Correlation between ILC subsets’relative frequencies in blood and age (cord blood: n59, children: n56, [3 to 12] years old, adults: n547, age mean 48). (I) CD56¹ILC1-like cell relative frequencies among the total ILCs in tissues from healthy adults (n53-18). Spearman

correlations were used in panel H. One dot51 donor. Mann-Whitney unpairedUtests were used in panel F. ****P,.0001. BM, bone marrow; LN, lymph node.

ns, not significant.

(6)

Metabolism Activation Migration Effector CD56+ ILC1-like ILC1 ITGAX XCL2 SH2D1B KLRC1 NCAM1 KIR2DL4 PPP1R9A ZMAT4 BMP2CABLES1 RASSF8 SERPINE1 COL24A1 SH3BP4 TIE1 ZNF521 CATSPER1 DOK7 KCNK17 ITGAD ROR1 MMRN1 CD33 LINC01451 RP4–738P11.4 HOXA10 HOXA9 RASSF8–AS1 DACH1 RP4–738P11.3 AC002383.2 CYTL1 SLITRK5 MAMDC2 SERP2 SHROOM1 YWHAEP7 AC079779.7 FOXC1 RP11–629G13.1 SCN9A HOXA3 F2RL3 PTX3 AC012454.4 KLRF2 SLC2A10 CNTNAP2 COL15A1 CA3 IL6REPPK1 PLCL1 ZC3H12D TRAV13–1 TRBV7–9 ADTAP FBLN7 CHNITRBV4–1 KANKI TRAV29DV5 TRAV9–2 TRAV21 TRBV10–3 TRBV12–4 TRAV20 PTGDR2 AIRE ANKRD36BP2 PTPN13 ALS2CL LINC00565 TRAV30 GP5 EVC2COL6A3 SEMA5A TRAV22 F2RL1 TRAV35 C14orf132 ST8SIA1 LRRN3 MTUS1 LINC00511 SLC22A17 ADAM23 ISM1 TRAV26–1 KIF19 SEPT10 MYOI6 NETO2 EDAR ST6GALNAC1 TRBV5-5 ANKRD55 TRBV2 ANK1

Metabolism Activation Migration Effector CD56+ ILC1-like ILC2 GNLY KLRDI XCL1 KLRF1 XCL2KLRC1 NCAM1 EOMES lGFBP4 IL12RB2 KIR2DL4 PTGDR RAMP1 SPTSSB SPRY2 B3GNT7 PPP1R9A ZMAT4 ATP8B4 PACSIN1 KLRC2 CDHR1 MYO7A CABLES1 STYK1 RASSF8 COL24A1 TIE1 ZNF521 RAB38 DEPDC7 MIR9–3HG RP5–1028K7.2 KCNK17 ITGAD LINC01451 IFNG–AS1 SGSM1 RP4–738P11.4 DACH1 ENPP5 RP4–738P11.3 KLRC3 AP11–330A16.1 AC002383.2 CYTL1 SLITRK5 MAMDC2 NLRP7 KRT81 C10orf128 MBOAT2 CHDH RBM47 PLCL1 CLIC4 NEFLKRT1 NRIP3 DAPK1 TNFRSF19 IL2 ADGRB2 HLF GRIP1 PKIBSLC35G1 ZP1 PKP2 MAMLD1 PTPN13 SCN2A KCNMA1 ZG16B LGALS12 DSG2 ALS2CL PTGDR2 FREM3 GJA1 LRP6 MCOLN3 MAFTRR IL13HAS2 ASAP3 ADAM23 lGSF9B PPARG GATA3–AS1 EVA1A MAL2HTR1F CRLF2 DNER SEMA5A LINC01229 FSTL4 GAP43 HPGDS

Metabolism Activation Migration Effector CD56+ ILC1-like ILCP KIR2DL4 SLAMF7 FCRL3 ENC1 SCD MR9–3HG TlGIT ENPP5 DTHD1 WNT10B NCS1 SLITRK5 KIR3DL2 MAMDC2 TLR5 TRO KIR2DL1 MCC GEMIN6 EP8SL1 IL1RL1 RIMKLA AC079779.7 RP11–629G13.1 PLCG1–AS1 NTNG2 CASC8 EPB41L5 SEPT8 PM20D1 SEC1P KlAA1161 TLCD2 NHS RWDD2A ANKRD18EP KCNK12 ALS2CR12 EFNA5 DAB1 HEATR9 KB–1460A1.5 AF131215.4 INSC RP11–3D4.3 RP11–543C4.1 LINC00943 MPP2 TDRD9 IGLV2–18 MATN2 CADM1 CLIC4 KRT1 DAPK1 IL1BTHAP9 ADGRD1 RAPH1 CTD–2527I21.9 RBMS3 RP11–146I2.1 GNG4 ABCG2 EPHB3 PTGDR2 MYB CRISPLD2 LINGO4 ZG16B HPN ALS2CL NETO2 ZNF385C SMARCA1 SOST KANK2 FAM64A MYBL2 ABCA9 LINC01229 B3GALT5 RP5–1086K13.1 CDH4 MALRD1 RP11–108K14.12 CYP26C1 SAPCD2 FAM189A2 FLT3 JADE3 TSPAN13 TFPI TTLL7 DNER ANPEP SLCO2A1 TRH APOL4 TGM2

Metabolism Activation Migration Effector CD56+ ILC1-like CD56dim NK IL7RIGFBP4 CCR7 HBB INPP4B NELL2 SPINK2 ITM2C SIRPG DPP4 PACSIN1 PI16 BMP2 MYO7A GNAI1 MSRB3 WDR86-AS1 RASSF8 TSPAN18 COL24A1 SH3BP4 MS4A1 CAMK4 ZNF521 IGHM MALKCNK17 ROR1 CDCP1 IL1R1 AXIN2 RP11–678G14.3 IFNG–AS1 CCR2RORC RASSF8–AS1 TRBV4–2 DACH1 RP11–330A16.1 FAAH2 AC002383.2 KlAA0125 IL23R LRP5 SLITRK5 ENPP1 MAMDC2 SIT1 CCR8 IGHD PTGDS CX3CR1 CTNNA1 DAB2 ARVCF PALLD LGR6 LAIR2 FBXO2 CMKLR1 DRAXIN NME8 KIR2DP1 CAPN5 GNAL BOK COL13A1 LRRC43 PDGFRB RRM2 C20orf197 FBLN2 PCDH1 PODN CXCR1 SLC1A7 DGKK BNC2 VIT GTSE1 LINC00565 SMPDL3A GLI3 GRM2FAM182B C2orf48 PCDHGB6 SLC47A1 OPHN1 RBPMS2 HEY2 ARHGEF28 MYBL2 B3GAT1 SORBS2 CHSY3 TBC1D30 KIF19 MGAM CXCR2

Min Max

logCPM

Metabolism ActivationMigration Effector CD56+ ILC1-like CD56bright NK HBB S100A9 S100A8 CD6HBA2 IGHM CD8B CD28 PPBP TRBV4–2 LAG3 S100A12 PF4 VCAN TUBB1 SIT1 MNDA MYLK4 IL1RL1 CLEC7A RGS18 SLAMF1 GNG11 HBA1 THEMIS FCER1A APOBEC3A QPCT CR1GSTM3 WNT7A CSTA PROK2 IGLC2 MARCH1 AQP9 PTX3FCRL2 ZNF80 HK3 GP9 MMACHC ELOVL4 CLC NTN4SEPT5 CD1C RP11–379K17.4 SOAT2 RP11–52J3.2 SDR42E1 FSD1L MKI67 DENND5B THAP9 BTBD6P1 SNX7 ZNF334 MOCS1 PIGZ CTD–2260A17.1 RP1–45N11.1 USP6 STOX1 LILRP2 SPAG6 LRRC20 PRDM16 PARD3 RGS6 ZG16B MARK2P8 CARD6 MYB PLSCR4 RP5–1126H10.2 FAM182B FAM187B2P MNS1GLIS3 RP11–573D15.8 AC100830.3 ZFP2 GHSR GTSE1 PHLDA3 LDHDDIRAS3 APOL4 RP11–34E5.4 PFKFB1 PLPP7 TMPPE JADE3 RP5–1148A21.3 PDE1C MGAM ZNF404 TRIP13 TSPAN13

C

TBX21 ILC1 ILC2 ILCP CD56+ ILC1-like CD56bright NK CD56dim NK

EOMES GATA3 RORC ZBTB16 ZNF683

z-score Max

Min TOX

NFIL3 TCF7 ETS1 RUNX3 GFI1 AHR GATA2 BCL11B –3

–10 –5

PC1 (60.2%)

PC2 (15.1%)

PC3 (6.2%)

0 5 10

–4 –3 –2

–1 0 2 1

3 4

–2 –1 0 1 2 3

ILC1 ILC2 ILCP

CD56+ ILC1-like CD56bright NK CD56dim NK

A B

Figure 2.Transcriptomic signature of ex vivo CD56¹ILC1-like cells in peripheral blood from HDs. (A) Principal component analysis (PCA) of ex vivo fluorescence-activated cell–sorted ILC and NK subsets from HDs peripheral blood (n53). (B) Heat map ofz scores of the expression levels of genes encoding ILC/NK transcription factors (n53). (C) Heat map of log counts per million (CPM) of the 100 most differentially expressed genes between CD56¹ILC1-like cells and ILC1, ILC2, ILCP, CD56^brightNKs or CD56^dimNKs. The GO pathway to which each gene belongs is represented at the left of each heat map: Metabolic process GO0008152 (“Metabolism"); Lympho- cyte activation GO0046649 (“Activation”); Leu- kocyte migration GO0050900 (“Migration”);

Immune effector process GO0002252 (“Effec- tor”). Max, maximum; Min, minimum.

(7)

0 20 40 60 80 100

Positive cells (%)

NKp3 0

NKp4 4

NKp4 6

NKp8 0

NKG2A NKG2C

DNAM-1 KIR2D

L1

KIR2D S1

KIR2D L2

KIR2D L3

KIR3D L1

KIR2D S4 TRAI

L CD

94

CD56dim NK CD56bright NK CD56+ ILC1-like ILCP

A

0

ILCP CD56+ ILC1-like CD56bright NK CD56dim NK 20

40 60 80

100 ***

granzyme B

*

0

40 60 80 100

****

granzyme K

****

Positive cells (%)Positive cells (%)

0

40 60 80 100 ****

granzyme A

*

B

granulysin

0

40 60 80 100

* granzyme M

0

40 60 80 100

**

perforin

0

40 60 80

100 ****

**

100 80 60 40 20 0

medium a-NKp30

K562

p=0.12 100

80 60 40 20 0

medium TRAI

L - D R p=0.015

BJAB

100 80 60 40 20 0

medium a-NKp80 p=0.06

U937

100

Specific lysis (%)

80 60 40 20 0

medium a-D

NAM-1

K562

F

*** *** ***

100

Specific lysis (%)

80 60 40 20 0

30.1 10.1 3.1 WT 721.221 HLA-E+721.221

1.1 0.3.1 0.1.1

G

50 40 30 20 10 0

medium K562

CD107a+ (%)

****

C

100 80 60 40 20 0

30.1 10.1 3.1

Effector:target ratio

Specific lysis (%)

1.1 0.3.1 0.1.1 U937 K562 BJAB

E

0.1.1

D

100

Effector:target ratio

80 60 40 20 0

30.1 10.1 3.1 1.1 0.3.1 CD56+ ILC1-like NK helper ILCs

Specific lysis (%)

Figure 3.CD56¹ILC1-like cells are cytotoxic effectors regulated by the NKp30, NKp80, TRAIL, and HLA-E pathways.(A) Extracellular flow cytometry analysis of NK receptor expression in ILCP, CD56¹ILC1-like cells, and cNKs (n54-18). (B) Intracellular flow cytometry was performed using HD PBMCs to assess CD56¹ILC1-like cell production of granzyme A (n515), granzyme B (n515), granzyme K (n512), granzyme M (n512), perforin (n512), and granulysin (n512). (C)Extracellular flow cytometry was performed after a 4-hour coculture of ILC/NK-enriched PBMCs with K562 (ratio E:T 1:1), anti-CD107a, and Golgistop to assess CD56¹ILC1-like cell degranulation (n516). (D) Specific lysis of the K562 tumor cell line by CD56¹ILC1-like cells, cNKs, and helper ILCs (results in duplicate). (E) Specific lysis of the U937, K562, and BJAB tumor cell lines by CD56¹ILC1-like cells (results in duplicate). (F) Specific lysis by CD56¹ILC1-like cells of K562 (ratio E:T 20:1), BJAB (ratio E:T 20:1), and U937

(8)

CD56¹ILC1-like cells possess common features with NK cell developmental intermediates

Freud and colleagues recently showed that ILC3 and NKs share a common progenitor that is distinct from ILC2 progenitors.³⁰ They also demonstrated that S4a NK developmental intermediates possess both NK- and ILC3-related characteristics.³¹We thus compared the CD56¹ ILC1-like cells with the NK developmental intermediates in terms of phenotype, based on their previously published data³¹(Table 1). NKp80 is acquired in NKs upon maturation from the S4b stage, alongside the production of cytotoxic mediators. Based on their expression of NKp80 and cytolytic activity, CD56¹ILC1-like cells are more mature than the S3 and S4a NK cells. However, being CD16-negative, these cells are clearly distinct from S5 NKs. CD56¹ ILC1-like cells share similar features with S4b NKs, because they express CD94/NKG2A heterodimers, NKG2D, CXCR3, and CD122, while being devoid of KIR receptors in blood. However, S4b NKs are reportedly CD127^low/²and c-Kit¹^/². S4b mainly comprises CD56^brightNKs, which have a distinct transcriptomic profile compared with CD56¹ILC1-like cells (Figure 2C). In order to formally evaluate the capacity of CD56¹ILC1-like cells to differentiate into other ILC subsets or NK developmental stages, we sorted these cells and cultured them on OP9 stromal cells in the presence of IL-7 with or without the addition of IL-15 (supplemental Figure 6A-B). Their phenotype was analyzed after 10 days. The CD16² CD94^high c-Kit² cell population remains the major one upon culture, in particular, in the presence of IL-15. Two minor cell populations emerge by the upregulation of CD16 or c-Kit. Next, to determine whether CD56¹ ILC1-like cells represent an effector population generated in the periphery, we monitored their presence in human fetal tissues and during immune reconstitution in humanized mice.

CD56¹ILC1-like cells were detectable in all human fetal tissues analyzed (supplemental Figure 6C) and could be identified as early as 8 weeks postreconstitution in humanized mice (supplemental Figure 6D-E). To further dissect their developmental requirement, we monitored the CD56¹ ILC1-like cell frequencies in patients with severe combined immunodeficiency (supplemental Figure 6F-G).

As conventional ILCs and NKs, CD56¹ ILC1-like cells require JAK3 and ADA genes for their development, whereas they are able to develop in the absence of ARTEMIS and CD3D genes.

We identified CD56¹ILC1-like cells and cNKs, yet at reduced levels, in an IL2RG-deficient patient, suggesting a hypomorphic mutation with reduced penetrance.^32,33 These cells, as cNKs and helper ILCs, are also present in a RAG1-deficient patient, whereas they are decreased in a RAG2-deficient patient, in contrast to cNK and helper ILCs. RAG1 is expressed in common lymphoid progenitors that are committed to the NK lineage³⁴and might thus be required for the differentiation of CD56¹ILC1-like cells. The absence of CD56¹ ILC1-like cells in the RAG1- deficient patient might also be due to the ILC and NK genomic instability observed in mice upon RAG deficiency.³⁵ Overall, these results suggest that CD56¹ILC1-like cells are related to S4b NK cells and therefore might share similarities with cNKs in terms of development.

CD56¹ILC1-like cytotoxicity is impaired in AML patients at diagnosis

In order to evaluate the role of CD56¹ ILC1-like cells in AML disease, we investigated their cytotoxic profile at diagnosis. The CD56¹ILC1-like cell expression of TRAIL, NKp30, and NKp80 is strongly reduced in the patients compared with HDs (Figure 4A).

Notably, TRAIL and NKp80 are specifically decreased in the CD56¹ILC1-like cells but not in the cNKs. Then, we analyzed the CD56¹ILC1-like cell release of cytotoxic mediators. The patients’

CD56¹ILC1-like cells produce amounts of granzymes A, B, and K and perforin that are similar to those in the HDs, but no granulysin is produced. In contrast, the granulysin levels in the cNKs are not impaired in the patients, and the granzyme A level is even increased in CD56^brightNKs in AML (Figure 4B). The degranulation of CD56¹ ILC1-like cells in AML is significantly impaired as assessed in cocultures with the K562 line or autologous blasts (Figure 4C-D).

This impairment could be explained by the regulation of CD56¹ ILC1-like cell cytotoxicity through NKp30, TRAIL, NKp80, or HLA-E in the case of primary AML blasts. Indeed, we observed HLA-E expression in primary AML blasts and NKG2A expression in patients’

CD56¹ ILC1-like cells, suggesting that the HLA-E/CD94-NKG2A pathway negatively regulates CD56¹ILC1-like cells (Figure 4E-F).

In line with this, in the patients, the CD94/NKG2A expression is preserved on CD56¹ILC1-like cells. Overall, we show that the CD56¹ ILC1-like cell cytotoxic machinery is impaired in AML patients at diagnosis, suggesting that these cells are unable to limit AML oncogenesis.

Table 1.Extracellular phenotype of CD56¹ILC1-like cells compared to previously published NK development stage phenotypes

Marker S3 S4a S4b S5 CD56¹ILC1-like

CD16 2 2 2 1 2

CD56 1/2 1 1 1 1

CD127 1 1/2 Low/2 2 1

c-KIT 1 1/2 1/2 1/2 2

CD94 2 1 1 1/2 1

NKG2A 2 1 1 1 1

NKG2C 2 2 1/2 1/2 1/2

NKG2D 2 1 1 1 1

KIR2D 2 2 2 1 2

NKp80 2 2 1 1 1

CXCR3 2 1/2 1 1 1

CD122 2 1 1 1 1

Perforin 2 2 1 1 1

Granzyme A 2 2 1/2 1 1

Granzyme B 2 2 1/2 1/2 1

Granzyme K 2 2 1 1 1

+, expression of the marker;2, no expression of the marker; Low, low expression of the marker.

Figure 3.(continued)(ratio E:T 10:1) tumor cell lines in the presence of anti-DNAM-1, anti-NKp30, and anti-NKp80 blocking antibodies or TRAIL decoy receptor.

(G) Specific lysis of wild-type (WT) or HLA-E–transfected 721.221 tumor cell lines by CD56¹ILC1-like cells (results in triplicate). One dot51 donor. Statistical tests used for analyses: panel B: Mann-Whitney unpairedUtest; panel C: Wilcoxon pairedttest, panel G: multiple Holm-Sidakttests. *P,.05, **P,.01, ***P,.001, ****P,.0001.

(9)

100 80 60 40 20 0

NKp80

% of positive cells

****

CD56+

ILC1-like

CD56bright NK

CD56dim NK

****

100 80 60 40 20 0

% of positive cells

TRAIL

A

CD56+

ILC1-like

CD56bright NK

CD56dim NK

100 80 60 40 20 0

NKp30

% of positive cells

**** *** ***

CD56+

ILC1-like

CD56bright NK

CD56dim NK

100 80 60 40 20 0

% of positive cells

Granzyme K

**

CD56+

ILC1-like

CD56bright NK

CD56dim NK

B

100 80 60 40 20 0

% of positive cells

Granzyme A

**

CD56+

ILC1-like

CD56bright NK

CD56dim NK

100 80 60 40 20 0

% of positive cells

Granzyme B

CD56+

ILC1-like

CD56bright NK

CD56dim NK

HD

AML at diagnosis CD56bright NK CD56dim NK CD56+ ILC1-like

100 80 60 40 20 0

% of positive cells

Perforin

CD56+

ILC1-like

CD56bright NK

CD56dim NK

100 80 60 40 20 0

% of positive cells

Granulysin

***

CD56+

ILC1-like

CD56bright NK

CD56dim NK

HD

AML at diagnosis CD56bright NK CD56dim NK CD56+ ILC1-like

FS-H

CD107a

HD - medium HD - K562

AML - medium AML - K562 AML - blasts

C

50 40 30 20 10 0

CD107a+ (%)

**** ***

medium K562 blasts

HD CD56+ ILC1-like AML (diagnosis) CD56+ ILC1-like

ns ns

HD AML

D

100 80 60 40 20 0

F

Positive cells (%)

HLA-E

E

FS-H

HLA-E

100 80 60 40 20 0

Positive cells (%)

CD94

G

_ns

100 80 60 40 20 0

NKG2A

Positive cells (%)

ns

100 80 60 40 20 0

Positive cells (%)

NKG2C ns

HD AML at diagnosis

Figure 4.CD56¹ILC1-like cell cytotoxicity is impaired in AML patients at diagnosis.(A) TRAIL, NKp30, and NKp80 expression in CD56¹ILC1-like cells and cNKs in HDs (CD56¹ILC1-like: n512-18, cNKs: n511-15) and AML patients at diagnosis (CD56¹ILC1-like: n511-20, cNKs: n58-10) by extracellular flow cytometry. (B) Intracellular flow cytometry was performed using PBMCs from AML patients at diagnosis to assess granzyme A (n54), granzyme B (n54), granzyme K (n54), perforin