Are natural killer cells superior CAR drivers?

T lymphocytes engineered to express a chimeric antigen receptor (CAR) are being celebrated as a major breakthrough of anticancer immunotherapy. Natural killer cells have not received similar attention as CAR effectors, although the use of these relatively short-lived cytotoxic cells is associated with several advantages.

The latter American Society of Hematology (ASH) meeting, which took place last December in New Orleans, was dominated by the enthusiasm on the anti-leukemia effects of T lymphocytes engineered to express a CD19-targeting chimeric antigen receptor (CAR). Fourteen out of 16 pediatric patients with relapsed or advanced acute lymphoblastic leukemia (ALL) entered a remission in response to the adoptive transfer of these cells. A similar outcome was documented for adults with chronic lymphocytic leukemia.

CAR-expressing T cells are usually generated from autologous T cells, but T lymphocytes from allogeneic donors are also being explored in this sense, especially upon relapse after stem cell transplantation. CAR-bearing T cells are usually activated with anti-CD3/CD28 beads and expanded in culture flasks (such as the WaveR system) in the presence of interleukin (IL)-2. CARs against an expanding array of cell surface-exposed tumor-associated antigens (TAAs) have been and continue to be engineered. Since the majority of these TAAs are not tumor specific, CAR-expressing T cells can cross-react with healthy cells, mediating an “on-target/off-tumor” side effect. For example, T cells expressing a CD19-targeting CAR can cause a profound and long-lasting B-cell deficiency as they eliminate normal B cells. T lymphocytes bearing a CAR specific for interleukin 3 receptor, α (ILR3A, also known as CD123) kill not only leukemic cells but also bone marrow cells that express the same receptor, leading to prolonged and profound marrow suppression. In some cases this on-target/off-tumor side effect can be fatal, as it happened in a patient with metastatic colon carcinoma who received T cells engineered to express a HER2 targeting CAR. In this case, the side effects of CAR-expressing T cells on low level HER2 expressing lung epithelium led to fatal pulmonary complications combined with a massive cytokine release.

It has been suggested that the antineoplastic activity of CAR-expressing T cells is related to and dependent on their persistence in the patient circulation and malignant tissue. If this were indeed the case, the on-target/off-tumor effects would also persist. For CD19-redirecetd T cells, this would entail a prolonged depletion of normal B cells and hence long-term defects in humoral immunity.

As recent clinical trials have suggested, antigen loss cancer variants can emerge as a result of the selective pressure imposed by immunotherapeutic interventions, often driving disease relapse. In this setting, TAA-specific T cells would continue to mediate on-target/off tumor effects, such as the suppression of normal B cells or bone marrow precursors. A potential solution to this issue is provided by the transduction of T cells with CAR-coding mRNAs, usually resulting in the loss of expression over a few days. Indeed, most CAR-expressing T cells currently tested in clinical trials are obtained with lentiviral constructs, which integrate into the genome and hence ensure persistent transgene expression.

Natural killer (NK) cells may represent alternative cytotoxic effectors for CAR-driven cytolysis. Allogeneic NK cells are expected to induce an immune response and be rejected after a few days, and even autologous NK cells should disappear relatively rapidly from the circulation, owing to their limited lifespan. NK cells have additional advantages over T cells (Table 1). In particular, while T lymphocytes only kill their targets by a CAR-specific mechanism, NK cells are endowed with spontaneous cytotoxic activity and can trigger the demise of target cells in a TAA-unrestricted manner via specific natural cytotoxicity receptors (NCRs), including NCR3 (also known as NKp30), NCR2 (also known as NKp44), NCR1 (also known as NKp46), and killer cell lectin-like receptor subfamily K, member 1 (KLRK1, best known as NKG2D). NK cells also express the Fc fragment of IgG, low affinity III, receptor (FcγRIII), that binds the Fc fragment of antibodies to elicit antibody-dependent cell-mediated cytotoxicity (ADCC). This specific feature of NK cells would enable the combination of 2 targeted therapies recognizing different (or the same) TAA(s), namely CAR-expressing NK cells and a TAA-specific monoclonal antibody.

Table 1. Comparison of CAR- expressing T, natural killer and NK-92 cells

Parameter	T cells	NK cells	NK-92 cells
Collection	Leukopheresis	Leukopheresis	Continuously growing cell line consisting of “pure” (100%) activated NK cells
Preparation	Activation of cells with anti-CD3/CD28 beadsAllogeneic donor: MHC match required	NK cells represent only 10% of all lymphocytes.Autologous: Enrichment needed (selection for CD56+ cells).Allogeneic donor: MHC-matched donor or depletion of alloreactive T-cells to prevent GvH reactions	No processing necessary prior to CAR engineering
Expansion	Flasks, bags or WaveR expansion system	Requires engineered feeders (example: K562 cells expressing IL-15 and TNFSF9) plus IL-2 (in flasks, bags or bioreactors)	Expansion in serum free-medium without feeders, but IL-2 only (in flasks, bags or bioreactors)
Transduction	Lentiviral systems transduce about 1/3 of T cells	Low transfection efficiency even with viral vectors	Transfection efficiency of about 50%, compatible with sorting
Cytotoxic mechanisms	CAR-restricted killingIn case of antigen loss on tumors, CAR-expressing T cells become ineffective	Multiple receptors can trigger CAR-independent and FcR-dependent cytotoxicity	Multiple receptors can trigger CAR-independent and FcR-dependent cytotoxicity
Adverse events	Can cause “off target” effectsSurvive for prolonged periods in the patient circulationCan induce a storm of pro-inflammatory cytokines	Limited life span in patientsNo concern about persisting CAR-associated side effects	Limited life span in patientsNo concern about persisting CAR-associated side effects
Miscellaneous	Suicide genes are required to control life span in vivo	No need for suicide gene	No need for suicide gene
Clinical results	Phase I studies have shown clinical benefit	Proof of clinical benefit pending	Proof of clinical benefit pending
Off-the-shelf CAR-specific cellular product?	Autologous cells, required preparation on a per patient basis	Possible to have donor NK cells cryopreserved, but recovery is poor after upon thawing	Possible to have NK-92 cryopreserved and expanded upon thawing (before infusion)

Abbreviations: CAR, chimeric antigen receptor; FcR, Fc receptor; GvH, graft-vs.-host; IL, interleukin; NK, natural killer; TNFSF9, tumor necrosis factor superfamily, member 9.

Additional features of NK cells could make them better and potentially safer CAR drivers than T cells. For instance, NK cells produce a host of cytokines that are different from those produced by T cells, including interferon γ (IFNγ) and granulocyte macrophage colony-stimulating factor (GM-CSF). The cytokine storm initiated by the infusion of CAR-expressing T cells is indeed largely mediated by their pro-inflammatory cytokines such as tumor necrosis factor α (TNFα), IL-1, and IL-6. It is also known that NK cells are “serial killers.” Thus, time-lapse videomicroscopy studies have shown that NK as well as NK-92 cells (a continuously-growing, highly-active, NK cell-derived cell line) diligently move from one target to the next one, killing on as many as 7–10 cells. Evidence for such a serial killing by T cells is lacking at this point.

Nonetheless, there are some obstacles for the use of circulating NK cells for CAR-based immunotherapy. Like T lymphocytes, NK cells are obtained (from patients or donors) by leukopheresis, which can be time-consuming, costly and occasionally requiring central venous access. Since only about 10% of circulating lymphocytes are NK cells, some extent of selection/enrichment is necessary, which is generally performed by positive CD56-based, magnetic immunoselection. In case of allogeneic donors, peripheral blood mononuclear cells also must be depleted of T cells to prevent graft-vs.-host reactions. Moreover, a feeder layer and additional cytokines are required to maximize the expansion of circulating NK cells in vitro. A feeder layer consisting of the leukemia K562 cells engineered to express tumor necrosis factor superfamily, member 9 (TNFSF9, best known as 4–1BBL) and IL-15 seems to be very effective, but a product testing before infusion must ensure that all malignant cells have been completely removed. Finally, the transfection efficiency of circulating NK cells is variable and generally not very high, even when viral vectors are employed.

Although the challenge of introducing CAR-coding genes into sufficient numbers of circulating NK cells may be overcome at some point, NK-92 cells present an open cellular platform for CAR-based immunotherapy. The transfection efficiency of NK-92 cells is about 50%, even with non-viral methods. Besides being technically more simple and under less-constraining regulations, avoiding viral vectors eliminates the risks of oncogene activation and insertional mutagenesis. Upon sorting, CAR-expressing NK-92 cells can be enriched to obtain a population near-to-exclusively composed of NK-92 carrying the CAR of interest.

So far, NK-92 cells have been efficiently transduced with a number of different CAR-coding constructs (Table 2) and pre-clinical studies in xenotransplanted immunodeficient mice have demonstrated the potential therapeutic effects of this approach. Several centers are gearing up to test whether CAR-expressing NK cells can keep up with their T-cell counterparts. Eventually, we might even discover that both these cell types have their place in the multimodal approach that is required to eliminate cancer and control its recurrence.

Table 2. CAR-coding genes transfected/transduced so far into natural killer cells

Target	Indication(s)	Blood NK cells	NK-92 cells	Refs.
CD19	Lymphoid malignancies	X	X	10,12–15
CD20	Lymphoid malignancies	X	X	16–18
CD38	Multiple myeloma		X	19
ERBB2	Breast carcinomaHead and neck cancerOvarian carcinomaGlioblastoma	X	X	11,20
GD2	Neuroblastoma		X	21
EPCAM	Breast carcinomaPancreatic cancer		X	22
EBNA3C	EBV infections		X	23
CS1	Multiple myeloma		X	24
LMAN1	MelanomaNeuroblastoma		X	25

Abbreviations: EBNA3C, Epstein-Barr nuclear antigen 3C; ERBB2, v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2; EPCAM, epithelial cell adhesion molecule; GD2, ganglioside GD2; LMAN1, lectin, mannose-binding, 1