**6. The impact of cytomegalovirus infection on natural killer cell-mediated graft-versus-leukemia effect: reconstitution promotion and memory induction**

#### **6.1. The impact of cytomegalovirus infection on natural killer cell reconstitution**

Although in healthy individuals, CMV infection is potentially controlled by T cells and NK cells, it can cause life-threatening complication in immunodeficient transplant recipients. Lethal CMV-caused illness is now uncommon thanks to the use of drugs such as ganciclovir and foscarnet. In view of the strong immune response caused by CMV, it is attractive to investigate whether CMV reactivation can reduce the risk of relapse in patients after HSCT. As early as 30 years ago, Lonnqvist et al. reported that CMV infection was associated with lower relapse in BMT recipients with various hematological malignancies [130]. Subsequently, a number of reports confirmed that CMV reactivation following HSCT resulted in reduced relapse in AML patients. Among these reports, Elmaagacli et al. found CMV reactivation occurred within the first 100 days after HSCT significantly reduce the rate of AML relapse from 42 to 9% [131]. Remarkably, this CMV reactivation-mediated relapse prevention effect was independent of aGvHD. Furthermore, in another study of large cohort of 2566 patients with variable hematological malignancies, Green et al. comprehensively analyzed the results among all kinds of disease with adjustment for potential variables and reported that CMV reactivation was significantly associated with a decreased risk of relapse [132].

Upon the association between CMV reactivation and decreased rate of relapse after HSCT was confirmed, several studies next tried to test whether CMV-specific donor T cells accounted for this beneficial effect. However, CMV-specific donor T cells appeared not to impact relapse rates after HSCT [133]. On the other hand, NK cells have been shown to play an important role in CMV reactivation-mediated relapse prevention. Foley et al. demonstrated that the beneficial effect of CMV reactivation was specifically associated with the clonal-like expansion of NK cells characterized by the NKG2C<sup>+</sup> NKG2A− self-KIR<sup>+</sup> CD57<sup>+</sup> CD56dim signature [80]. These NK cells have been shown to exhibit enhanced cytolytic activity and cytokine production. During CMV reactivation, both NKG2C<sup>+</sup> NK cell percentages and numbers are significantly increased with enhanced capability of producing IFN-γ upon the stimulation of myeloid leukemia cell line K562. Nevertheless, the mechanisms underlie the increased expansion and functions of NKG2C<sup>+</sup> NK cells are not entirely clear. Guma et al. reported that after *ex vivo* stimulation with CMV-infected fibroblasts, NKG2C<sup>+</sup> NK cells from seropositive healthy donors rapidly proliferated [134]. These results indicated that NKG2C may play a unique role in NKG2C<sup>+</sup> NK cell expansion. However, the exact ligands of NKG2C on CMV infected cells are not completely confirmed. Although HLA-E has been shown to be a ligand for NKG2C [135], other undefined ligands may also take part in NKG2C-mediated effect. Leukemic cells can retain HLE expression while downregulate the level of classical class I HLA molecule [136]. Therefore, it is now believed that the change in receptor repertoire from inhibitory receptor NKG2A to activating receptor NKG2C may play a crucial role in this NK cell subset-mediated GvL effect [137].

It has been shown that NK cells from UCB transplants will undergo slower reconstitution compared with NK cells from adult grafts. In contrast, in CMV-infected UCB transplantation patients, NK cells are found to reconstitute rapidly with decreased NKG2A expression and increased KIR expression [138]. Their functional reconstitution is also promoted as they can produce high level of IFN-γ in response to K562 cells. Intriguingly, during CMV-infected UCB transplantation, Della Chiesa et al. detected a unique subset of hyporesponsive NK cells which shared similar surface markers with the expanded NKG2C<sup>+</sup> NK cells except for CD56 expression [139]. Briefly, this NK cell subset does not express CD56. IL-2 stimulation can reverse the hyporesponsive state of CD56− NK cells thereby this NK cells subset may represent a stage of NK cell differentiation in the case of cytokine deficiencies.

The effect of CMV reactivation was also investigated in recipients received grafts either from peripheral blood or bone marrow [140]. Strikingly, the expansion of NKG2C<sup>+</sup> NK cells seemed to be dependent on the serostatus of recipient. The percentage of NKG2C<sup>+</sup> NK cells was significantly increased when CMV seropositive recipients received grafts from CMV seronegative donors. Oppositely, when grafts from CMV seropositive donors were infused into CMV seronegative recipients, NKG2C<sup>+</sup> NK cells failed to expand, and their percentage was declined to the levels similar to CMV seronegative donor/recipient pairs.

Besides NKG2C, activating KIR may also contribute to CMV-induced NK cell expansion. Della Chiesa et al. proved that when donor grafts were lack of NKG2C, NKG2C− NK cells can rapidly expanded expressing activating KIR in the recipients following CMV infection [141]. These data indicated that activating KIR can recognize CMV and promote NK cell reconstitution independent of NKG2C. Correspondingly, the increased presence of donor activating KIR has been shown to result in reduced risk of CMV infection [142, 143]. However, the ligands expressed on the CMV infected cells specific for activating KIR are still elusive, which become an obstacle to explore the actual role of activating KIR in CMV-infected HSCT.

Additionally, CMV seropositivity has been shown to be associated with presence of a population of NK cells which does not express FcγRIγ [144]. FcγRIγ is an adapter molecule that is used by CD16, NKP30, and NKp46 for signal transmitting. CD3ζ is also required for CD16 associated signal transduction. Lack of both FcγRIγ and CD3ζ in HIV-infected patients was proved to diminish CD16 signaling [145], while single down-regulation of FcγRIγ in CMV seropositive healthy donors resulted in enhanced CD16 signaling [146]. The deficiency of FcγRIγ induced NK cells hyporesponsive to CMV infected fibroblasts, whereas CMV specific antibody can reverse NK cells to degranulate and produce IFN-γ and TNF-α against CMV infected fibroblasts.

#### **6.2. Cytomegalovirus infection induces natural killer cell memory**

Upon the association between CMV reactivation and decreased rate of relapse after HSCT was confirmed, several studies next tried to test whether CMV-specific donor T cells accounted for this beneficial effect. However, CMV-specific donor T cells appeared not to impact relapse rates after HSCT [133]. On the other hand, NK cells have been shown to play an important role in CMV reactivation-mediated relapse prevention. Foley et al. demonstrated that the beneficial effect of CMV reactivation was specifically associated with the

CD56dim signature [80]. These NK cells have been shown to exhibit enhanced cytolytic

ages and numbers are significantly increased with enhanced capability of producing IFN-γ upon the stimulation of myeloid leukemia cell line K562. Nevertheless, the mechanisms

Guma et al. reported that after *ex vivo* stimulation with CMV-infected fibroblasts, NKG2C<sup>+</sup> NK cells from seropositive healthy donors rapidly proliferated [134]. These results indi-

exact ligands of NKG2C on CMV infected cells are not completely confirmed. Although HLA-E has been shown to be a ligand for NKG2C [135], other undefined ligands may also take part in NKG2C-mediated effect. Leukemic cells can retain HLE expression while downregulate the level of classical class I HLA molecule [136]. Therefore, it is now believed that the change in receptor repertoire from inhibitory receptor NKG2A to activating receptor NKG2C may play a crucial role in this NK cell subset-mediated GvL effect [137].

It has been shown that NK cells from UCB transplants will undergo slower reconstitution compared with NK cells from adult grafts. In contrast, in CMV-infected UCB transplantation patients, NK cells are found to reconstitute rapidly with decreased NKG2A expression and increased KIR expression [138]. Their functional reconstitution is also promoted as they can produce high level of IFN-γ in response to K562 cells. Intriguingly, during CMV-infected UCB transplantation, Della Chiesa et al. detected a unique subset of hyporesponsive NK cells

expression [139]. Briefly, this NK cell subset does not express CD56. IL-2 stimulation can

The effect of CMV reactivation was also investigated in recipients received grafts either from

nificantly increased when CMV seropositive recipients received grafts from CMV seronegative donors. Oppositely, when grafts from CMV seropositive donors were infused into CMV

Besides NKG2C, activating KIR may also contribute to CMV-induced NK cell expansion. Della

expanded expressing activating KIR in the recipients following CMV infection [141]. These data indicated that activating KIR can recognize CMV and promote NK cell reconstitution

NKG2A−

self-KIR<sup>+</sup>

NK cells are not entirely clear.

NK cells except for CD56

NK cells seemed

NK cells was sig-

NK cells can rapidly

NK cells thereby this NK cells subset may repre-

NK cells failed to expand, and their percentage was declined

NK cell expansion. However, the

NK cell percent-

CD57<sup>+</sup>

clonal-like expansion of NK cells characterized by the NKG2C<sup>+</sup>

138 Natural Killer Cells

underlie the increased expansion and functions of NKG2C<sup>+</sup>

which shared similar surface markers with the expanded NKG2C<sup>+</sup>

sent a stage of NK cell differentiation in the case of cytokine deficiencies.

peripheral blood or bone marrow [140]. Strikingly, the expansion of NKG2C<sup>+</sup>

to be dependent on the serostatus of recipient. The percentage of NKG2C<sup>+</sup>

Chiesa et al. proved that when donor grafts were lack of NKG2C, NKG2C−

to the levels similar to CMV seronegative donor/recipient pairs.

reverse the hyporesponsive state of CD56−

seronegative recipients, NKG2C<sup>+</sup>

cated that NKG2C may play a unique role in NKG2C<sup>+</sup>

activity and cytokine production. During CMV reactivation, both NKG2C<sup>+</sup>

Although traditionally viewed as a member of innate immune cells, NK cells have been demonstrated to mount memory response to hapten and mediate hapten-induced contact hypersensitivity responses [21]. In MCMV model, NK cells were also observed to develop memory-like properties including viral ligand m157 specific expansion, persisting over time and enhanced effector function after rechallenge with MCMV [22, 23]. This MCMV-induced memory-like population of NK cells is characterized by the expression Ly49H. Intriguingly, during acute human CMV, the expanded NKG2C<sup>+</sup> NKG2A− self-KIR<sup>+</sup> CD57<sup>+</sup> CD56dim NK cells which we have described above was proved to be human analog of Ly49H<sup>+</sup> memory-like NK cells [139, 147]. After HSCT, these CMV-specific memory-like NK cells can be detected in peripheral blood of CMV-infected HSCT recipients. As we have mentioned, when CMVseropositive recipients received grafts from CMV-seropositive donors, these NKG2C<sup>+</sup> NK cells underwent expansion even in the absence of detectable viremia, which meant a nonclassical recall response was developed [140]. However, the mechanisms underlined the generation of memory NK cells have not been exactly defined. Recently, two studies demonstrated that CMV infection resulted in modified effector function through driving epigenetic alterations in these "adaptive" NK cells [148]. These memory-like NK cells were proved to lack expression of signaling proteins such as Fc epsilon receptor I (FcεRIγ), spleen tyrosine kinase (Syk), ewing's sarcoma-associated transcript 2 (EAT2) and exhibit increased capacity to mediate ADCC as well as produce IFN-γ in response to tumor targets. In contrast, memory-like NK cells produced extremely less level of IFN-γ in response to cytokines such as IL-12 and IL-18 mainly due to the downregulation of cytokine receptors.

In summary, NKG2C<sup>+</sup> NK cells expanded after CMV-infected HSCT display memory-like properties and associated with reduced risk of relapse. Nevertheless, it is still unclear whether the GvL effect observed in CMV-infected HSCT is mainly due to NKG2<sup>+</sup> C NK cells cytotoxicity toward leukemic blasts or primarily due to the cytolytic effect of CMV. This potent effect of CMV reactivation in shaping immune response after HSCT may provide new choice for therapeutic strategies, such as utilizing CMV vaccine to mimic this beneficial effect, isolating NKG2C<sup>+</sup> NK cells for adoptive transfer setting and optimizing the criterion for donor selection.
