ILT3.Fc–CD166 Interaction Induces Inactivation of p70 S6 Kinase and Inhibits Tumor Cell Growth

A deregulated cell division cycle is a central trait of the cancer cell. Chemotherapeutic agents selectively target dividing cells, yet produce side effects caused by their cytotoxic activity on normal growing cells from tissues that contain a significant fraction of dividing cells. Recently, to maximize the efficiency and minimize the toxicity of conventional therapy, combinations of chemo- and immunotherapy were explored.

The blockade of immune checkpoints by anti-receptor and/or anti-ligand mAb is one of the most recent and promising approaches to cancer immunotherapy (1, 2). Agents that target the inhibitory receptors CTLA4 and PD1 have yielded impressive results in the treatment of various advanced malignancies. Multiple costimulatory and inhibitory interactions that regulate T cell responses may inhibit autoimmune reactions or enhance immune responses against malignancies (1). Several clinical studies combining VEGF-A or VEGFR inhibitors with checkpoint therapy have reported enhancement in tumor immune responses with associated clinical benefit (3).

Ig-like transcript 3 (ILT3), also known as LILRB4, LIR5, or CD85k (4), is an inhibitory receptor expressed by normal and malignant human cells of myelomonocytic origin (5–7). We previously showed that upregulation of membrane ILT3 on dendritic cells renders them tolerogenic, because they induce anergy in CD4 Th cells and elicit the in vitro differentiation of CD8⁺CD28⁻ T suppressor cells (Ts)/regulatory T cells (Treg) (5). A similar tolerogenic effect is displayed in vitro and in vivo by soluble recombinant ILT3.Fc protein, which binds to a ligand expressed by T cells only upon activation (8, 9). ILT3 has a cytoplasmic domain that contains ITIM that inhibit cell activation by recruiting the tyrosine phosphatase Src homology region 2 domain–containing phosphatases (4, 8). The extracellular Ig-like domains of ILT3, used for engineering the ILT3.Fc protein, bind to activated T cells and induce their differentiation into Ts/Treg (8, 9). There is no ortholog of the ILT3 gene in mice.

Although the ligands of numerous members of the Ig-like transcript family have been identified, the ligand of ILT3 has remained elusive. The identification of this ligand is of clinical interest for the potential use of ILT3 and/or its cognate receptor in checkpoint therapy of cancer, autoimmunity, and transplant rejection.

We now demonstrate that a CD166/activated leukocyte cell adhesion molecule (ALCAM) is the ligand of ILT3. ILT3 and CD166 belong to the Ig superfamily. Like ILT3, which only binds to activated T cells, CD166/ALCAM is expressed on activated, but not resting, T cells. However, it is also expressed by hematopoietic and mesenchymal stem cells, myeloid progenitors, and many types of tumors. CD166 comprises five extracellular Ig domains (V1, V2, C1, C2, and C3) plus a transmembrane domain and a short cytoplasmic tail (10, 11), which regulates adhesion through a link with the cytoskeleton through a supramolecular complex, including ezrin and syntenin (12). The N-terminal domains (V1, V2) of CD166 mediate low-affinity homophilic interaction, which may be involved in the growth of numerous types of malignancies and much stronger heterophilic interaction with CD6, a surface receptor expressed by T cells, thymocytes, and a subset of B cells (10, 13). Our studies suggest that recombinant ILT3.Fc protein inhibits tumor cell growth, possibly disrupting CD166–CD166 homophilic interactions between malignant cells.

Jurkat and CCRF-CEM acute T cell leukemia (ATL) and HH and H9 cutaneous T cell lymphoma (CTCL) were obtained from the American Type Culture Collection. Acute T cell lymphoblastic leukemia (ALL)-SIL, HPB-ALL, and DND-41 (ALL) were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen. All leukemia/lymphoma tumor cell lines (TCL) were cultured in complete RPMI 1640 medium supplemented with 10% FBS, 2 mM l-glutamine, and 50 mg/l gentamicin sulfate (Mediatech). All cell lines were received in 2015–2016, cryopreserved, and propagated in our laboratory for two to eight passages and used within 1 mo after thawing. The cell lines were authenticated by flow cytometry, morphology, and biologic behavior. Cells were grown in a humidity-controlled incubator with 5% CO₂ at 37°C and routinely tested for mycoplasma contamination before experiments using a PCR-based commercial kit (ABM).

Buffy coats of blood obtained from healthy blood donors were purchased from the New York Blood Center. PBMC were separated by density gradient centrifugation. Purified (>95%) CD3⁺ T cells were obtained by negative selection using CD3 isolation kits (Miltenyi Biotec). T cells were stimulated with PWM (Sigma-Aldrich, 0.05 μg/ml) and with murine anti-human CD3/CD28 (BD Pharmingen, 5 and 1 μg/ml). All cell cultures were performed in complete RPMI 1640 medium. After 3 d of incubation, cells were stained with anti-CD166 PE or ILT3.Fc-FITC or double stained with both reagents. MLC were performed in triplicate cultures using 5 × 10⁴ responding T cells from one individual and an equal number of irradiated allogeneic T cell–depleted PBMC as stimulating cells. ILT3.Fc and hybridoma supernatants were added to the medium at the initiation of the cultures. After 5 d of incubation, cultures were labeled with [³H]TdR for 18 h and then harvested, and incorporation was counted in a PerkinElmer MicroBeta² counter. Hybridomas whose supernatants modulated the inhibitory activity of ILT3.Fc (12.5 μg/ml) were cloned and selected for further studies.

Primary CD3⁺ T cells (pT) were transfected with CD166-specific or mock control small interfering RNA (siRNA; Invitrogen) using a Human T Cell Nucleofector Kit (Lonza). After 4 h of incubation, cells were stimulated with PWM or anti-CD3/CD28 mAb in the presence or absence of ILT3.Fc (12.5 μg/ml), as described above. Forty-eight hours later, proliferation was measured by [³H]TdR incorporation. CD166-knockdown efficiency was determined 48 h after stimulation by flow cytometry analysis of CD166 expression on siRNA-transfected T cells.

For Ts assays, purified T cells (1 × 10⁷ cells) transfected with CD166-specific or control siRNA were incubated with irradiated (3000 rad) allogeneic CD3-depleted PBMC (5 × 10⁶ cells) for 7 d in cultures containing ILT3.Fc (50 μg/ml) or control human IgG (50 μg/ml). Immunomagnetic sorting was used to obtain CD8⁺CD28⁻ T cells from these cultures. These putative Ts (5 × 10⁴ cells) were added to MLC containing CFSE-labeled autologous responding cells (5 × 10⁴ cells) and APC (5 × 10⁴ cells) from the original stimulator. Proliferation of CFSE-labeled T cells was determined by flow cytometry, as previously described (14).

Eukaryotic expression plasmids pRK7-HA-S6K1-WT, encoding wild-type (WT), kinase-dead pRK7-HA-S6K1-F5A (F5A), or rapamycin-resistant mutant pRK7-HA-S6K1-F5A-E389-R3A (R3A) (15, 16) were transfected into H9 and Jurkat cells using nucleofection (Lonza). Tumor cells were suspended in fresh complete RPMI 1640 medium and seeded into a six-well plate 1 d before transfection. The following day, 1 × 10⁶ cells per reaction were nucleofected with 2000 ng of specified plasmids using a Cell Line Nucleofector Kit V, according to the manufacturer’s protocol.

Murine mAb specific for the putative ligand of ILT3 were generated by immunizing BALB/c mice with cells from the Jurkat TCL. Hybridoma supernatants were screened by flow cytometry for binding to the immunogen, but not to resting normal T cells, and by MLC-inhibition studies to determine whether they modulate the inhibitory effect of ILT3.Fc. mAb 2D9 was selected because of its specific binding to Jurkat ATL and its capacity to enhance T cell proliferation in cultures containing ILT3.Fc.

For T cell stimulation, anti-CD3 and CD28 Ab were purchased from BD Pharmingen. For flow cytometry staining, the following murine anti-human CD166 mAb were used: anti-CD166 mAb (clone 3A6; BD Pharmingen), anti-CD166 mAb (clone AZN-L50; specific for the C-terminal domain; Hycult Biotech), and anti-CD166 mAb (clone J4-81; specific for the N-terminal domain; ANTIGENIX AMERICA). Mouse IgG (BD Pharmingen) was used as an isotype control. The following Ab were used for Western blotting experiments: phospho-p70S6 kinase (T389) Rabbit polyclonal (catalog number 9205S), p70S6 kinase (49D7) Rabbit mAb (catalog number 2708), phospho–NF-κB/p65 (S536) (clone 93H1) Rabbit mAb (catalog number 3033S), NF-κB p65 (clone D14E12) Rabbit mAb (catalog number 8242P), phospho–PLC-γ1 (Tyr⁷⁸³) Rabbit polyclonal (catalog number 2821S), PLC-γ1 Ab Rabbit polyclonal (catalog number 2822), phospho-p42/44 MAPK (Thr²⁰²/Tyr²⁰⁴) Rabbit polyclonal (catalog number 9101), P44/42 MAPK (Erk1/2) (clone 137F5) Rabbit mAb (catalog number 4695), phospho-Ezrin (Y353) Rabbit polyclonal (catalog number 3144S), Ezrin Rabbit polyclonal (catalog number 3145S), Anti-Rabbit IgG HRP-Linked Ab (catalog number 7074), anti-Mouse IgG HRP-linked Ab (catalog number 7076S), and β-actin (8H10D10) Mouse mAb (catalog number 3700) (all from Cell Signaling Technology).

Flow cytometry studies were performed on a BD FACSCalibur instrument. For each Ab, a corresponding isotype-matched Ab conjugated with the same fluorescent dye was used as a negative control. Cells were incubated with saturating concentrations of the indicated mAb for 10 min at room temperature (RT) in PBS containing 2% FBS and 0.01% sodium azide, washed twice, and analyzed by FACS using CellQuest software. ILT3.Fc was labeled with FITC dye using a FluoReporter FITC Protein Labeling Kit (Thermo Fisher Scientific). The final concentration of ILT3.Fc-FITC was 2 mg/ml. FITC-conjugated Human IgG (Sigma-Aldrich) was used as a control for ILT3.Fc-FITC staining. For ILT3.Fc-FITC staining, 2 μl of ILT3.Fc-FITC or an equal amount of IgG-FITC was added to 100 μl of cell suspension (5 × 10⁴ cells). After incubation for 25 min at RT (25°C), cells were washed twice and resuspended in PBS containing 0.1% paraformaldehyde before analyzing. Double staining was performed by incubating cells with ILT3.Fc-FITC for 15 min and then together with anti-CD166–PE for an additional 10 min at RT.

For the study of Ts, magnetically sorted CD3⁺ responder T cells were labeled by incubation with 3 μM CFSE (Invitrogen) for 10 min in PBS/0.1% BSA at RT. The reaction was stopped by addition of ice-cold culture medium containing 10% FCS for 5 min, followed by extensive washing. In test cultures, CFSE-labeled T cell responders were incubated with autologous CD8⁺CD28⁻ T cells that were immunomagnetically sorted from the CD166 siRNA–transfected population that was allostimulated in the presence of ILT3.Fc or human IgG. In control cultures, CFSE-labeled CD3 T cell responders were coincubated with autologous CD8⁺CD28⁻ T cells from the mock-transfected population allostimulated in the presence of ILT3.Fc or human IgG. T cell proliferation in response to APC from the same donor was assessed on day 3 by analyzing dye dilution using a FACSCalibur instrument and CellQuest software (BD Biosciences).

Protein–protein interactions were measured with a Biacore 3000 surface plasmon resonance instrument (Biacore, Uppsala, Sweden). To prepare an immobilized CD166 sensing surface, CD166-His (Sino Biological) was charged with 500 mM NiCl₂ at 5 μl/min for 1 min and activated in 200 mM N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide/50 mM N-hydroxysuccinimide for 6 min. Activated CD166-His (10 μg/ml) in HBS-P buffer containing 50 μM EDTA and 2 mM MgCl₂ was injected onto a HisCap chip (Pall ForteBio) at 5 μl/min to achieve 250 or 2000 response units (RU). One RU is equivalent to 1 pg of protein per square millimeter on the sensor surface. After immobilization, Ni was removed from the system by injection of 0.5 M EDTA for 3 min and equilibrated in a running buffer containing 10 mM HEPES, 150 mM NaCl, 200 μM EDTA, 2 mM MgCl₂ (pH 7.4 0.05%), and Tween-20 prior to use. This buffer allows binding in the Ni-dependent channel but no binding in the Ni-independent (nonspecific) channel when 2 μM proteins were injected. Nonspecific binding, but not specific binding, was completely abolished when 1 mg/ml BSA was included in the running buffer system. Therefore, we used this buffer throughout the assay.

For Biacore analysis, three recombinant proteins, CD6.Fc (Sino Biological), ILT3.Fc, and IgG1.Fc (R&D Systems), were resuspended in the running buffer. CD6.Fc and IgG1.Fc were used as the positive and the negative controls, respectively. All data were fitted globally to a single-site 1:1 (Langmuir) model for simultaneous determination of k_on (on rate) and k_off (off rate). K_d was determined by a plot of steady-state signal versus ILT3.Fc concentration. Raw data were presented together with the fitted data. ILT3.Fc protein at different concentrations (0.1–2 μM) was captured by the CD166-immobilized sensor chips by injection at a flow rate of 10 μl/min for 3 min. The concentrations of CD6.Fc, Fc from IgG1, and ILT3.Fc were 70 nM, 20 μM, and 2 μM, respectively.

A double-target guide RNA lentiviral system was used for knockout (KO) of the CD166 gene in TCL (17, 18). Exon 3 encoding CD166 peptides, which are present in all known CD166 mRNA variants, was targeted by gene-specific guide RNA. Two upstream dsDNA primers (guides 1 and 2) and two downstream dsDNA primers (guides 3 and 4) were designed using the online tool provided by Ran et al. (17). The annealed guide dsDNA oligonucleotides were cloned into the BsmB1 site of LentiCRSPRv2, which encodes the U6 RNA promoter, SpCas9 RNase, and a puromycin resistance gene, to generate the four recombinant guide plasmids 1–4. These plasmids were cotransfected with lentiviral packaging plasmids (GAG and VSV) into 293T cells. Seventy-two hours after transfection, the four individual recombinant lentiviral particles (guide 1–4 LentiCRSPRv2) were concentrated by PEG8000 precipitation of supernatants. To KO the CD166 gene, Jurkat and H9 cells were spin infected for 2 h at RT with lentiviral particles that targeted two genomic regions of exon 3. These cells were incubated for an additional 72 h at 37°C and cultured for 1 wk in medium containing puromycin (1–4 μg/ml). CD166⁻ surviving tumor cells were sorted in 96-well trays by flow cytometry and cloned. Genomic editing of the CD166 gene was screened by PCR. Clones with the expected size of PCR amplicon for CD166-KO cells had 700 or 820 bp, depending on the RNA guides used. CD166-KO clones were further confirmed by DNA sequencing from both ends. The sequences of the primers used in this study are listed in Table I.

Cells were washed twice with cold PBS and fixed with 80% ethanol at −20°C overnight. The following day, cells were washed once with PBS before staining with PI/RNase Staining Buffer (BD Pharmingen), according to the manufacturer’s protocol. Cell doublets were gated out based on pulse width versus pulse area.

The number of viable cells was monitored by trypan blue exclusion (0.4% solution) and counted on a hemocytometer under a microscope.

A total of 1 × 10⁶ cells was washed twice with PBS, resuspended in 1× binding buffer, and stained with an FITC Annexin V Apoptosis Detection Kit I (BD Biosciences), according to the manufacturer’s instruction.

Jurkat cells were seeded at 1 × 10⁶ cells per milliliter in a six-well plate and grown in complete RPMI 1640 with 0, 12.5, or 25 μg/ml ILT3.Fc for 6, 12, or 24 h in duplicate. RNA was extracted from each of these 18 samples using an RNeasy Mini Kit (QIAGEN). Five hundred nanograms of RNA from each sample were used for sequencing. Libraries were prepared from total RNA samples (200 ng to 1 μg per sample, RNA integrity number > 8 required) with Illumina’s TruSeq RNA prep kit, using poly-A pull-down to enrich for mRNAs. Samples were sequenced on an Illumina HiSeq 2000, using lane multiplexing, yielding 30 million single-end 100-bp reads for each sample. The raw RNA sequencing (RNA-Seq) data were submitted to the Gene Expression Omnibus database under accession number GSE106147 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106147). Reads were mapped to the reference human genome (NCBI/build 37.2) using TopHat (19) (version 2.1.0) with four mismatches and 10 maximum multiple hits.

The mechanism of action of ILT3.Fc-induced perturbation was investigated using the Detecting Mechanism of Action by Network Dysregulation (DeMAND) algorithm (20), leveraging a T cell gene-regulatory network (referred to as a T cell interactome) constructed by applying the ARACNe algorithm (21, 22) to 223 gene-expression profiles (GEP) from ALL (23). GEP were normalized with GCRMA prior to processing (24). Gene interactions were inferred from three separate ARACNe runs, involving three types of “hub” genes: transcription factors (1664 genes), cotranscription factors (617 genes), and signaling proteins (2944 genes). The final T cell interactome included 374,053 predicted interactions (120,306 involving transcription factors; 134,192 involving cotranscription factors; and 119,555 involving signaling proteins).

Cells were washed twice with ice-cold PBS, and the pellet was resuspended in Pierce IP Lysis Buffer and incubated on ice for 10 min with periodic mixing. Cell debris in lysates was removed by centrifugation at 13,000 × g for 15 min at 4°C, before being subjected to SDS-PAGE and transferred to nitrocellulose membranes. After blocking for 30 min with 5% nonfat milk at RT, the membrane was immunoblotted with primary Ab overnight at 4°C. The following day, the membrane was incubated with HRP-conjugated secondary Ab for 30 min at RT and developed via ECL. A Bradford assay (Bio-Rad) was used to determine protein content.

Seven- to ten-week-old NOD.Cg-Prkdc Il-2rg/SzJ (NSG) female mice (Taconic, Hudson, NY) were maintained under pathogen-free conditions. Experiments were performed in compliance with the Institutional Animal Care and Use Committee of Columbia University. A total of 5 × 10⁵ H9 CTCL or 1 × 10⁶ Jurkat luciferase-tagged tumor cells (25) was injected s.c. into the right flank of the mouse. Mice were randomized to receive treatment with ILT3.Fc or an equal amount of human IgG. Tumor cell growth was monitored by bioluminescence imaging, and s.c. tumor volume was measured by the two largest perpendicular axes (length and width). The tumor volume was calculated using the formula V = (4/3) r³, where r = (length + width)/4. The fold increase in tumor burden was determined by the ratio of the bioluminescence value before and after treatment for each mouse.

Treatment of mice was initiated when an increase in the bioluminescence signal indicated the onset of tumor xenograft growth. Mice transplanted with the H9 TCL were treated with daily i.p. doses of 250 μg of ILT3.Fc or human IgG for two intermittent cycles (each of 5 d on and 2 d off). Jurkat-xenografted mice received five daily i.p. injections, followed by a course of 10 alternative-day injections (500 μg) of ILT3.Fc or an equal amount of human IgG. Complete necropsies were performed, and s.c. tumors were measured using a caliper. Whole blood, bone marrow, liver, and spleen were harvested at the time of sacrifice and analyzed for human cell surface CD45 and CD166 by flow cytometry.

Tissues from the s.c. injection site (skin), liver, spleen, kidney, and lung were harvested, fixed in 10% buffered formalin, embedded in paraffin, and sectioned at 3 μm for light microscopic evaluation (H&E) and immunostaining for human CD45.

Data are shown as mean ± SD or ± SEM. Statistical analysis was assessed using GraphPad Prism 7.0 (*p < 0.05, **p < 0.01, ***p < 0.001). The significance of the difference between two groups was assessed using a two-tailed Student t test in Microsoft Excel. For multiple-group comparisons, a one-way ANOVA program was used.

To identify the ILT3 ligand, we generated mAb by immunizing mice with Jurkat cells. Hybridoma supernatants that bound to Jurkat cells, but not resting human T cells, were tested for their activity on MLC containing ILT3.Fc (0–12.5 μg/ml). After screening, mAb 2D9 (IgG1, κ), derived from a hybridoma that abrogated the inhibitory activity of ILT3.Fc in MLC, was cloned for further studies. Mass spectrometry analysis identified the Ag immunoprecipitated by 2D9 as CD166/ALCAM.

2D9, together with other anti-CD166 Ab, including N-terminal–specific mAb (clone J4-81), C-terminal–specific mAb (clone AZN-L50), and clone 3A6 were screened together with ILT3.Fc-FITC for their binding to a panel of TCL, including five ATL and two CTCL (Table II). 2D9 showed the same binding pattern as that displayed by ILT3.Fc-FITC and by mAb J4-81 and 3A6 on these TCL. They can bind to all of the TCL in the panel, with the exception of ALL-SIL. This latter TCL was stained only by AZN-L50, which recognizes the C-terminal domains of CD166 (8). Hence, the N-terminal V1 domain of CD166, which is critical for the CD166–CD166 homophilic and CD166–CD6 heterophilic interactions, also contains the ILT3.Fc binding site.

To study the affinity and binding kinetics between ILT3.Fc and CD166, we performed surface plasmon resonance. To this end, we used CD6.Fc, the classical ligand of CD166, and Fc of human IgG1, as positive and negative controls, respectively. We first estimated binding affinities of all of these proteins at a single concentration (2 μM) on an immobilized CD166 sensor chip. As shown in Fig. 1A and Supplemental Fig. 1, a strong capture of CD6.Fc by immobilized CD166-His was demonstrated. K_d was determined to be 14.0 nM, with k_on and k_off of 3.5 × 10⁵ (s⁻¹) and 5.0 × 10⁻³ (M⁻¹s⁻¹), respectively. A strong CD6–CD166 interaction has been documented by other investigators (26). A weaker, but detectable, capture of ILT3.Fc by CD166 was clearly shown (K_d = 1.7 μM, k_on = 2.7 × 10⁵ s⁻¹ and k_off = 4.4 × 10⁻¹ M⁻¹s⁻¹). Examining the ILT3.Fc:CD166 binding kinetics, we found that k_on of ILT3.Fc:CD166 and CD6:CD166 are very similar. In contrast, k_off were quite different: ILT3.Fc:CD166 (k_off = 4.4 × 10⁻¹ M⁻¹s⁻¹) was 100 times higher than that of CD6:CD166 (k_off = 5.0 × 10⁻³ M⁻¹s⁻¹), suggesting that the lower affinity of ILT3.Fc:CD166 is due to the very fast dissociation rate of the complexes. This binding was specific, because it could not be abolished by adding BSA (1 mg/ml) to the running buffer. Human IgG1.Fc showed no binding to the sensor chip, even at a very high concentration (20 μM). This result demonstrates that the interaction of ILT3.Fc with CD166 is specific, although its affinity is weaker than that of CD6 with CD166. To avoid complexities of kinetic analyses of divalent binding, we calculated binding affinities at an equilibrium stage. We injected several concentrations of ILT3.Fc (100 nM–2 μM) into the Biacore instrument under identical conditions. Specific binding of different concentrations of ILT3.Fc to the CD166 sensor chip was measured. Increased RU correlated with the increased concentration of ILT3.Fc (Fig. 1B). The K_d value at the steady-state was determined to be ∼0.9 μM, which is consistent with the K_d estimated from a single concentration (K_d = 1.7 μM) (Fig. 1C).

Taken together, our findings demonstrate that the inhibitory receptor ILT3 interacts specifically with CD166, the ALCAM. Furthermore, although ILT3.Fc’s affinity for CD166 is significantly weaker than that of the heterophilic interaction between CD6 and CD166, it is in line with the known affinity of the CD166 homophilic interaction (26).

CRISPR-Cas9 genomic editing–based KO of CD166 from H9 and Jurkat cells eradicated the binding ability of ILT3.Fc to both of these TCL; double staining of H9 and Jurkat cells with ILT3.Fc and anti-CD166 mAb showed that a large proportion of these TCL populations were double positive. However, after KO of CD166 from Jurkat ATL and H9 CTCL, CD166/ILT3.Fc double-negative TCL (e.g., which did not bind anti-CD166 or ILT3.Fc mAb)were generated (Fig. 2A, 2B).

The overlap between the staining with CD166-PE and ILT3.Fc-FITC of the same population of TCL was further documented using T cells from the blood of healthy individuals. Resting pT showed almost no staining with anti-CD166-PE mAb or ILT3.Fc-FITC. However, T cells stimulated with PWM or anti-CD3/CD28 mAb became double positive (Fig. 2C).

Nucleofection of CD166-specific siRNA in pT resulted in the downregulation of CD166 by 60–70%. Knockdown of CD166 did not change the proliferation induced by PWM or CD3/CD28 Ab stimulation. However, after knockdown of CD166, ILT3.Fc had a significantly lower inhibitory activity (p < 0.05) on the proliferation of responding pT compared with its effect on mock-transfected controls (Fig. 3A, 3B).

To determine whether knockdown of CD166 from pT interferes with the generation of ILT3.Fc-induced Ts, we performed MLC with and without ILT3.Fc using mock-transfected or CD166-specific siRNA-transfected pT as responders and allogeneic APC as stimulators. After 7 d, CD8⁺CD28⁻ T cells were isolated and added to autologous pT, which were then stimulated in MLC with the same allogeneic APC.

CD8⁺CD28⁻ Ts were generated in the presence of ILT3.Fc when WT or mock-transfected pT were used as responders but not when the responders were transfected with CD166-specific siRNA. This indicates that binding of ILT3 to the CD166 ligand expressed by responding cells is required for generation of Ts (Fig. 3C).

Because CD166 is known to be expressed on malignant cells, we next studied the effect of ILT3.Fc on tumor growth and viability in vitro. Repeat experiments showed that ILT3.Fc induced a potent dose-dependent inhibition of tumor cell growth in lymphoid TCL (Fig. 4). [³H]TdR-incorporation studies in six leukemia/lymphoma TCL further confirmed this dose-dependent inhibitory effect of ILT3.Fc on tumor cell proliferation (Fig. 5).

To understand whether the cell cycle was affected by ILT3.Fc in these TCL, DNA staining of the ILT3.Fc-treated H9 and Jurkat cells was performed using flow cytometry. The data indicated that cell cycle arrest occurred at the transition from the G₁ phase to the S phase within 48 h of treatment. ILT3.Fc-induced cell death was also observed, as shown by the increased sub-G₁ population (Fig. 6A–D). Next, annexin V/propidium iodide staining of the same cells showed that ILT3.Fc induced cell death in H9 and Jurkat cells. To determine whether ILT3.Fc’s effect seen above was CD166 dependent, CD166-KO/H9 and CD166-KO/Jurkat cells were treated with ILT3.Fc under the same conditions as their WT counterparts. Viability was not affected by ILT3.Fc in CD166-KO cells (Fig. 6E–H). These data indicate that ILT3.Fc blocks tumor cell proliferation and induces cell death upon its binding to cell surface CD166.

To investigate the in vivo effect of ILT3.Fc on tumor growth, we injected NSG mice (n = 10) s.c. with luciferase-tagged H9 cells. H9 cells grew aggressively in mice treated with human IgG (n = 5) and also progressed, but at a significantly lower rate, in mice treated with ILT3.Fc (n = 5) (Fig. 7A–C). Animals were sacrificed and necropsied on day 28. In the IgG-treated group, metastatic tumor expansion to the liver, bone marrow, and spleen was documented macroscopically and microscopically (Fig. 7D). The incidence of metastasis in ILT3.Fc-treated mice was significantly lower, and fewer human malignant CD45⁺/CD166⁺ tumor cells were detectable by flow cytometry in blood, bone marrow, spleen, and liver cell suspensions (Fig. 7E).

For survival studies, two groups of mice (each group n = 10) were grafted with H9 cells and treated with IgG or ILT3.Fc. Kaplan–Meier survival curves showed that ILT3.Fc treatment significantly prolonged survival (Fig. 7F). The experiment was terminated on day 60 when 3 of 10 ILT3.Fc-treated mice carrying small tumors were also sacrificed.

Similarly, ILT3.Fc inhibited s.c. transplanted Jurkat cell growth, as illustrated by the higher tumor burden and size in IgG-treated mice compared with ILT3.Fc-treated mice and prolonged survival (Fig. 8).

To understand the mechanism of action underlying ILT3.Fc’s effect on TCL, we took advantage of RNA-Seq and bioinformatics. Gene-expression profiling in ILT3.Fc-treated Jurkat cells (three ILT3.Fc concentrations at three time points in duplicate cultures) were detected by RNA-Seq, and data were further analyzed using the DeMAND algorithm to identify signaling and regulatory proteins whose activity was significantly altered as a result of ILT3.Fc-induced perturbation (20).

Finally, DeMAND successfully identified 824 hub genes as significantly enriched in dysregulated interactions (adjusted p value < 0.01) after ILT3.Fc treatment (Supplemental Table I). RPS6KB1 was the top enriched gene on the list (adjusted p value = 1.11 × 10⁻³²). RPS6KB1 encodes a kinase that phosphorylates the S6 ribosomal protein, inducing protein synthesis and controlling cell growth during the G₁ to S transition phase (15). Enrichment analysis using high-scoring DeMAND genes, including RAF1, MAP2K1, MAPK13, PLCG1, PIK3R3, PDK1, and PP2A, identified a signaling pathway consisting of PLC-γ, RAF1, and MAPK/ERK that eventually leads to p70S6K activation. Phosphorylation of ERK1 and ERK2, which belong to the RAF1/MEK/MAPK signaling pathway, and of PDK1 is also known to result in the activation of p70S6K (27–35).

To investigate the effect of ILT3.Fc on signaling transduction upstream of p70S6K, we analyzed the phosphorylation level of PLC-γ1, MAPK/ERK, and p70S6K in H9 and Jurkat cells. These signaling molecules are constitutively phosphorylated in Jurkat TCL (Fig. 9A) and H9 TCL (Fig. 9B), consistent with their important role in tumor growth. However, they were dephosphorylated after 24 h of incubation in medium containing ILT3.Fc. The data indicated that ILT3.Fc may inhibit upstream signaling transduction and eventually induce the inactivation of p70S6K in tumor cells. In contrast with the inactivation of p70S6K in WT H9 cells, the phosphorylation level of p70S6K in CD166-KO H9 tumor cells was not affected by ILT3.Fc treatment (Fig. 9C).

After confirming that p70S6K activation was inhibited by ILT3.Fc, we explored its possible role as a master effector of ILT3.Fc-induced perturbation. For this, we transfected Jurkat and H9 cells with various mutant p70S6K plasmids, such as pRK7-HA-S6K1-WT (encoding WT protein), F5A (encoding the kinase-dead mutant), or R3A (encoding the constitutively active mutant) (16). The viability of different transfectants in 3-d cultures, with or without ILT3.Fc, was monitored by trypan blue exclusion. Comparison of the results showed that the rapamycin-resistant mutant (constitutively active form) of p70S6K partially rescued Jurkat and H9 cells from the ILT3.Fc-induced viability decrease (Fig. 9D, 9E). In contrast, neither the WT nor the kinase-dead mutant (F5A) had any protective effect against ILT3.Fc-induced inhibition of cell growth.

Because CD166 stably interacts with actin, binding to syntenin-1 and Ezrin (12), we next studied the phosphorylation level of Ezrin in Jurkat cells upon ILT3.Fc treatment. ILT3.Fc induced phosphorylation of Ezrin after 12 h. However, Ezrin became dephosphorylated and returned to the baseline level after 24 h (Fig. 9A). This suggests that phosphorylation of Ezrin following ILT3.Fc binding to CD166 may induce outside-in signaling and subsequent cellular responses. We also found that ILT3.Fc inhibited NF-κB/p65 activity, consistent with other studies revealing its important role in supporting tumor cell growth (36) (Fig. 9).

Taken together, our findings indicate that p70S6K is a master effector of ILT3.Fc-induced perturbations. This confirms numerous studies demonstrating that, although activation by various upstream pathways is important for tumor cell growth, inhibition of p70S6K signaling suppresses proliferation.

In this study, we have established for the first time, to our knowledge, that CD166 is the ligand of ILT3. Biacore assays showed that the ILT3.Fc:CD166 interaction has a K_d value ∼ 0.9 μM, which is relatively weak compared with that of CD6:CD166. However, it is stronger than that of CD80:CD28 (4 μM) (37) and within the range of many known immune ligand and receptor interactions. For example, using the same forms (Fc conjugated) of recombinant proteins in a Biacore assay, the K_d of CD80:PD-L1 was determined to be ∼1.5 μM, whereas the K_d of PD-1:PD-L1 and PD-1L:PD-L2 were shown to be ∼0.77 and ∼0.59 μM, respectively (38). The weaker affinity of ILT3.Fc for CD166 compared with CD6:CD166 may be due to the fast binding kinetics of ILT3.Fc:CD166. Similar to the fast binding kinetics of CD80:CD28 or CTLA4:CD28, this may be necessary to accommodate dynamic T cell–APC and/or T cell–T cell contacts optimizing bidirectional interactions that modulate immune responses.

We found that KO of CD166 abrogated the capacity of tumor cells to bind ILT3.Fc. The growth of human TCL was inhibited by ILT3.Fc in vitro and in vivo.

We demonstrated that ILT3.Fc protein inhibits PLC-γ–MAPK/ERK–p70S6K signaling pathways, which led to the inactivation of p70S6K, a key kinase controlling protein synthesis and cell growth. Therefore, the CD166/ILT3 ligand–receptor pair appears to be a potentially important checkpoint that can be targeted for immunotherapy in cancer (39).

Numerous studies revealed that CD166/ALCAM is a differentially expressed gene in cancer cells that serves as a valuable prognostic marker of disease progression and poor survival in patients with liver (40), prostate (41), lung (42), colorectal (43), pancreas (44), mammary (45), or head and neck carcinoma (46), as well as melanoma (47).

However, thus far, only one human recombinant single-chain anti-CD166 Ab has been shown to reduce the growth of human tumors in nude mice (48). No murine anti-human CD166 mAb with such tumor inhibitory effect has been described. Thus, the finding that binding of ILT3.Fc to CD166 inhibits signaling events crucial to tumor cell proliferation holds great promise for immunotherapy. Destabilization of homophilic CD166–CD166 interactions (10) may be greatly beneficial to patients with CD166⁺ malignancies. Cell adhesion molecules are involved in cell–cell interactions and cell–extracellular matrix interactions. It is assumed that ALCAM forms a network in the plasma membrane and that disturbance of this network is important for tumor growth (10, 48). We hypothesize that ILT3.Fc disturbs the ALCAM–ALCAM network and, thereby, reduces in vitro and in vivo tumor growth.

The unfortunate sequela of chemotherapy-induced immune suppression is an increased susceptibility to infections. ILT3.Fc does not bind to CD166⁻ resting T cells, acting only on activated CD166⁺ T cells. Hence, although ILT3.Fc is known for its immunosuppressive activity, it is unlikely to interfere with patients’ natural and acquired immunologic competence. The fact that binding of ILT3.Fc to CD166 inhibits tumor growth, whereas anti-CD166 Ab do not, may be attributable to differences in signaling pathways triggered by ligand binding to distinct epitopes. This hypothesis is supported by our present finding that, although ILT3.Fc inhibited T cell proliferation in MLC, anti-CD166 Ab, such as our 2D9, have no such effect, acting as an antagonist, rather than an agonist, of ILT3.Fc.

Because CD4⁺CD25⁺ natural human regulatory cells are also CD166⁺, binding of ILT3.Fc to these cells may diminish their capacity to survive, a highly desirable event for cancer immunotherapy.

In previous studies, we discovered that CD8⁺CD28⁻ Ts, generated by multiple MLC stimulations, differentiate into Ts that induce the downregulation of CD40, CD80, and CD86 and the upregulation of ILT3 and Ig-like transcript 4 on the membrane of the dendritic cells used as specific MLC stimulators (5). Suppression of CD4 T_h cell proliferation by CD8 Ts required cognate recognition by CD4 T_h cells and CD8 Ts of the HLA class I and II Ags, expressed by the tolerogenic APC bridge. The suppressor effect was cell contact dependent, HLA allorestricted, and abrogated by addition of the corresponding anti-ILT3 Ab or of IL-2 to the cultures. The recovery of CD4 Th reactivity after addition of IL-2 demonstrated that membrane ILT3 induced Th anergy. High expression of ILT3 on APC that were rendered tolerogenic was also achieved by exposure to certain cytokines, such as IL-10, IFN-β, and IFN-α (5). The corollary also hold true, because KO of ILT3 from monocytes resulted in a dramatic increase in the production of inflammatory cytokines and migration factors (49).

In later studies, we determined that the inhibitory activity of ILT3 was not limited to its membrane form; it was also displayed by the extracellular Ig-like domains after deletion of the cytoplasmic ITIM-containing tail. The recombinant ILT3.Fc protein inhibited the production of IL-2, IFN-γ, IL-5, and IL-17 by CD4 Th and of granzyme B by CD8 cytotoxic T lymphocytes while promoting the differentiation of CD8 Ts/Treg (8). Suppressors were shown to differentiate because of lack of CD4 T cell help, rather than being induced by soluble factors. ILT3.Fc-induced CD8 Ts showed a significant increase in the level of expression of genes with immune-regulatory function, such as BCL6, DUSP10, SOCS1, TGFBR2, and CXCR4, all of which are known to downregulate AP1 and T cell activation (14, 49–51). DUSP10 was also increased in Jurkat cells treated with ILT3.Fc (14).

Other investigators showed that overexpression of DUSP10 inhibits cell proliferation in colorectal tumors (52) and that abnormal expression of SOCS1 in human carcinoma is associated with dysregulation of signals from cytokine receptors (53).

In the current study, we demonstrate that KO of CD166 from primary human T cells abolishes their sensitivity to ILT3.Fc, which fails to inhibit their proliferative responses to mitogens and allogeneic-stimulating cells, as well as their differentiation into Ts.

It has been suggested that other Ig-like transcript/LILRB family members (LILRB 1–5) may act as immune checkpoint proteins and tumor-sustaining factors (54).

The obvious question is how can ILT3.Fc, a potent immunosuppressive agent, inhibit tumor cell growth? The most likely explanation resides in its capacity to dephosphorylate and inhibit signaling pathways that are crucial to cell proliferation and survival upon binding to its ligand (CD166) expressed by activated T cells (especially CD4⁺ T_h cells) and a wide range of tumor cells.

The wide expression of CD166 on malignant cells and the inverse relationship between its level of expression and prognosis (40–47) suggest the importance of this cell surface molecule in tumor growth. We hypothesize that ILT3.Fc-induced distortion of the homophilic interaction in cis or in trans may cause impaired communication between the tumor cells, as well as production or access to growth factors. ILT3.Fc-mediated inhibition of S6K activity may lead to this effect.

Thus, ILT3.Fc has tremendous translational potential as an anticancer therapeutic, either alone or as part of a combinatorial approach with IL-2, in nonlymphoid malignancies and/or other immune checkpoint blockers, such as anti–PD-1, –PD-L1, –CTLA-4, or agonistic anti-CD40 Abs.

We thank Dr. Adolfo Ferrando and Dr. Andrea Califano for invaluable help and guidance throughout this study. The S6K mutant plasmids were a generous gift from Dr. John Blenis. Biacore studies were performed with the invaluable help of Dr. Alexander Vinitsky.

The authors have no financial conflicts of interest.