Loss of STAT3 in Lymphoma Relaxes NK Cell-Mediated Tumor Surveillance

The transcription factors and proto-oncogenes STAT3 and STAT5 are highly activated in hematological malignancies and represent promising therapeutic targets. Whereas the importance of STAT5 as tumor promoter is beyond doubt, the role of STAT3 in hematological cancers is less well understood. Both, enforced as well as attenuated expression of STAT3 were reported in hematopoietic malignancies. Recent evidence implicates STAT3 as key player for tumor immune surveillance as it both mediates the production of and response to inflammatory cytokines. Here we investigated the effects of STAT3 deletion in a BCR/ABL-induced lymphoma model, which is tightly controlled by natural killer (NK) cells in vivo. Upon STAT3 deletion tumor growth is significantly enhanced when compared to STAT3-expressing controls. The increased tumor size upon loss of STAT3 was accompanied by reduced NK cell infiltration and decreased levels of the cytokine IFN-γ and the chemokine RANTES. Upon transplantation into NK cell-deficient mice differences in lymphoma size were abolished indicating that STAT3 expression in the tumor cells controls NK cell-dependent tumor surveillance. Our findings indicate that STAT3 inhibition in lymphoma patients will impair NK cell-mediated tumor surveillance, which needs to be taken into account when testing STAT3 inhibitors in preclinical or clinical trials.


Introduction
Tumorigenesis involves both tumor-intrinsic alterations as well as modulations of the tumor environment finally favoring tumor growth and maintenance. Transformation of a cell requires the expression of oncogenes or silencing of tumor suppressor genes [1]. Most malignancies remain dependent on the transforming oncogene, a process that is described as -oncogene addiction‖ and well described for oncoproteins such as BCR/ABL or EGFR [2]. Tumor cells may also acquire additional adaptations or changes in signaling pathways that they may become dependent on-a phenomenon called -non-oncogene addiction‖ [3]. Luo et al. hypothesized that interference with such signaling pathways would result in system failure and tumor cell death. Players of the JAK/STAT pathway-in particular STAT3 and STAT5-have been repeatedly shown to represent such critical factors [4][5][6][7][8].
STAT3 was shown to be persistently activated in a variety of solid and hematological cancers [5,6,9,10] and to promote metastasis [11][12][13]. Constitutive STAT3 activation relieves tumor cells from their dependence on cytokines and growth factors-thereby allowing continuous cell cycle progression and proliferation. Recently STAT3 was shown to support transformation via a metabolic function in mitochondria [14,15]. Concordantly, constitutively active STAT3 (STAT3αC)-expressing mice showed accelerated skin tumor formation upon UVB irradiation [16]. Moreover, crossing of MMTV-neu transgenic mice to STAT3αC-knock-in animals resulted in an earlier onset of mammary tumor formation accompanied by increased invasiveness [13]. Constitutive activation of the JAK/STAT pathway was described in patient-derived BCR/ABL + leukemic cells [17][18][19][20]. The tumor promoting role of STAT3 in leukemia formation has been confirmed in mice: transduction of bone marrow (BM) with constitutively active versions of STAT3 rapidly induced leukemia in mice [21]. Tumor formation occurred spontaneously and in the absence of a driving oncogene: STAT3αC alone sufficed to induce leukemia. These observations prompted current investigations evaluating STAT3 as a prognostic biomarker in patients suffering from acute forms of leukemia (NCT01245231, NCT01138332, and NCT01057290). STAT3 inhibitors and antisense oligonucleotides are presently undergoing biosafety studies in clinical trials (NCT00955812, NCT00696176, NCT01563302, NCT01904123, NCT01423903, NCT01867073 and NCT01066663).
However, the role of STAT3 for tumor formation appears complex. A variety of studies described STAT3 as a tumor suppressor [21][22][23][24]. Ecker et al. observed a proliferative disadvantage of fibroblasts and suppression of myc-induced foci formation in the presence of STAT3αC [21]. In line, the subcutaneous injection of myc + STAT3αC + fibroblasts resulted in a profound reduction of tumor weight. Moreover, Mair et al. described that conditional inactivation of STAT3 in hepatocytes and cholangiocytes aggravated bile acid-induced liver injury and fibrosis [25] thereby implicating a protective role of STAT3.
Tumor development is shaped and sculptured by external influences such as the surrounding immune system [26]. Immune cells continuously screen our body to detect and eradicate degenerated or transformed cells. Tumor surveillance is dominated by the interplay of myeloid cells, cytotoxic T lymphocytes (CTL) and natural killer (NK) cells. NK cells are the main players for the eradication of leukemic cells transformed by the BCR/ABL oncogene [27][28][29]. NK-dependent surveillance requires their recruitment, recognition and subsequent lysis of target cells. Attraction of NK cells is mediated by chemokines, such as CCL5 (or RANTES, binding CCR5), CCL19 and CCL21 (binding CCR7), CXCL10 (binding CXCR3) and CXCL12 (binding CXCR4) that are secreted by other immune cells or the tumor itself [30]. Recognition of tumor cells by NK cells is triggered by low MHC class I levels (-missing-self‖) or enhanced expression of stress signals such as NKG2D ligands [31,32]. Receptor-receptor interactions between NK and target cells induce cytotoxicity that is either delivered through exocytosis of granules packed with lytic enzymes (e.g., perforin and granzymes) or by producing cytokines such as interferon (IFN)-γ stimulating adaptive immune responses.
STAT3 has many facets in cancer and may act as tumor promoter and tumor suppressor. STAT3 induces target gene transcription of pro-survival and proliferative genes such as Bcl2 or cyclin D1 that promote tumor growth and survival. Another layer of complexity comes from the fact that STAT3 is critically involved in the production of immune-modulatory cytokines such as IL-6, IL-10, IL-17 and IFN-γ [33,34] and pro-angiogenic factors like VEGF [35]. Persistent STAT3 activity may provoke synthesis of IL-10, IL-17 and VEGF. Moreover, STAT3 represents a critical player downstream of these cytokines. In summary, STAT3 is part of a feed-forward loop that results in immunosuppression and inflammation [36]. The inhibition or genetic ablation of STAT3 relieves immunosuppression and thus markedly ameliorates anti-tumor responses [36,37].
Given the delicate balance between STAT3's tumor promoting and tumor suppressing actions it is critical to assess the net effect of STAT3 inhibition within the tumor and the impact on tumor surveillance. In this study we used a conditional Stat3 knockout mouse model to investigate the consequences of Stat3 deletion for BCR/ABL p185+ -driven tumor growth. Intriguingly, we found that STAT3-deficient tumors were substantially larger, which was accompanied by reduced NK cell recruitment to the tumor. NK cell cytotoxicity is significantly lower against STAT3-deficient compared to STAT3-expressing tumor cells. In line with this, pro-inflammatory cytokines such as IL-6, IFN-γ, TNF-α and RANTES are markedly reduced in STAT3-deficient tumors.

Generation of STAT3-Deficient BCR/ABL + Pro-B Cell Lines
Single cell suspensions of BM derived from Stat3 fl/fl Mx1-Cre and wild-type mice were infected retrovirally with the BCR/ABL p185 oncogene. After three weeks, stable growth-factor-independent BCR/ABL p185+ cell lines were established. Two independent approaches to delete STAT3 in these BCR/ABL p185+ cells were used. First, we deleted the endogenous Stat3 locus in BCR/ABL p185+ Stat3 fl/fl Mx1-Cre cell lines by IFN-β treatment (further on named Stat3 ∆/∆ ). Ten days thereafter loss of STAT3 was confirmed by western blotting. We also tested whether any other closely related STAT family members would compensate for the absence of STAT3. As depicted in Figure 1A protein levels of STAT1 and STAT5 were unaltered upon STAT3 deletion. In an additional approach STAT3 was reduced using lentiviral knockdown in two independently derived wild-type pro-B cell lines (#1 being transformed by v-ABL p160 , #4 by BCR/ABL p185 ). This resulted in a STAT3 knockdown of 64% and 48% compared to control cells, respectively ( Figure 1B). FACS analysis verified that the expression of B cell-specific markers (CD19 and B220) was unaltered upon loss of STAT3 ( Figure 1C). Stat3 ∆/∆ cell lines expressed the typical pro-B cell markers CD19 and B220. One representative cell line is depicted.

In Vitro Proliferation of BCR/ABL + Cells Expressing or Lacking STAT3
STAT3 is an important mediator of cell growth and survival [33,38,39]. Thus, inhibition of STAT3 in tumor patients has been proposed as a promising therapeutic strategy [40]. Together with STAT3, STAT5 was repeatedly described to be constitutively activated and to mediate tumor cell survival and proliferation [41][42][43][44]. As STAT3 and STAT5 share common anti-apoptotic and cell cycle-regulating target genes, it was speculated that they share redundant functions. However, in BCR/ABL-transformed cells survival depends on STAT5 but not on STAT3: deletion of Stat5 resulted in a cell cycle arrest in G1 followed by apoptosis [8]. Apoptosis was not induced by deletion of Stat3 [8]. We found that Stat3 fl/fl and Stat3 ∆/∆ cell lines showed superimposable proliferation rates ( Figure 2A). Further, we could not detect any differences in proliferation between cell lines harboring a STAT3 KD or a control hairpin ( Figure 2B). Accordingly, cell cycle profiles of STAT3-expressing and STAT3-deficient cells under normal culturing conditions were unaltered ( Figure 2C,D).

Lymphoma Development in Vivo
In vitro proliferation of transformed cell lines only partially reflects in vivo tumor formation. Within an organism tumor cells are exposed to a variety of cytokines that may affect survival and proliferation. Importantly the majority of these cytokines signals via the JAK/STAT cascade. STAT3 mediates target gene expression downstream of pro-inflammatory cytokines such as IL-6, TNF-α or IL-10 [45]. We thus speculated that despite the unaltered tumor cell proliferation in vitro, there may be a difference in tumor growth in vivo. To test this, Stat3 fl/fl and Stat3 ∆/∆ BCR/ABL p185+ cells were injected subcutaneously into the flanks of C57BL/6 wild-type mice. Intriguingly, after 11 days STAT3-deficient lymphoma sizes were significantly increased compared to wild-type tumors. The increased tumor size was evident irrespective whether the STAT3 abrogation was achieved genetically ( Figure 3A) or by lentiviral knockdown ( Figure 3B).
Staining of tumor sections with H&E, Ki67 and CD31 revealed obvious differences. Whereas we failed to detect any alterations in the degree of necrosis ( Figure 3C), Stat3 ∆/∆ tumor boundaries were less definite and the infiltrative character was more pronounced compared to Stat3 fl/fl controls ( Figure 3D). The invasive front of Stat3 ∆/∆ tumors was demarcated by formation of granulation tissue ( Figure 3D, see arrows), which was also evident from the increased formation of nascent blood vessels at the marginal tumor area ( Figure 3E). In contrast, STAT3-expressing tumors hardly harbored granulation tissue ( Figure 3D,E). Within the tumor lymph and blood vessel densities were unaltered (data not shown). Ki67 staining unraveled a substantially enhanced proliferation of STAT3-deficient tumor cells in vivo ( Figure 3F), which stands in clear contrast to our observations in vitro.
Taken together, deletion of STAT3 in BCR/ABL + lymphoma resulted in an aggravation of tumor burden. At a first glance these results may be difficult to reconcile as they oppose studies by others, who defined STAT3 as a proto-oncogene [46,47]. Indeed there is growing evidence that STAT3 acts as a tumor suppressor in certain tissues [21][22][23][24]. Depending on the cellular system and on the oncogene that drives transformation STAT3 may thus be either considered a tumor suppressor or a tumor promoter. In the case of BCR/ABL + lymphoma STAT3 acts as a tumor suppressor.

Cytokine Profile in STAT3-Deficient Lymphomas
The discrepancies between our in vitro and in vivo findings might result from the cytokine milieu that influences tumor growth in vivo. The involvement of STAT3 in cytokine signaling and production may contribute to the altered granulation tissue and blood vessel formation as well as cell proliferation in STAT3-deficient tumors.
We thus compared the expression levels of a number of candidate cytokines of in vitro growing and ex vivo derived tumor cells by real-time PCR concentrating on cytokines that either depend on STAT3 for production or signaling. The expression levels of IL-10, CXCL1, CXCL2, CCL2, CCL7 and VEGF were unaltered (Figure 4 (Figure 4). Previous reports have implicated the involvement of IFN-γ and TNF-α in proliferation and wound healing. Upon injury, IFN-γ- [48] as well as TNF-α-deficient animals [49] develop augmented granulation tissue and vessel formation. Accordingly, the reduced levels of IFN-γ and TNF-α in STAT3-deficient lymphomas are likely to contribute to the occurrence of granulation tissue and enhanced proliferation in these tumors.  To determine the cytokine milieu within the lymphoma in vivo, RNA was prepared from tumors depicted in Figure 3A (n = 4-5 per group), meaning after transplantation of Stat3 fl/fl or Stat3 ∆/∆ cell lines subcutaneously into C57BL/6 wild-type recipients. Numbers indicate arbitrary units, relative to the expression in Stat3 fl/fl BCR/ABL + cells cultivated under control conditions (dmso). Bar graphs depict means ± SEM (unpaired t-test). Asterisks denote statistical significance: * p  0.05; ** p  0.01; *** p  0.001; n.s. not significant.
Further, we found significantly lower expression of CCL5 (RANTES) in STAT3-deficient tumor cell lines in vitro as well as in STAT3-deficient lymphoma samples ex vivo (Figure 4). The chemokine CCL5 recruits and stimulates leukocytes to elicit anti-tumor immunity [50][51][52]. Further, CCL5 was described as a downstream target of TNF-α, which itself acts as a chemo-attractant for myeloid cells [53]. These findings prompted us to investigate the composition and amount of lymphoma-infiltrating leukocytes.

Lymphoma-Infiltrating Leukocytes
The majority of the cytokines we tested is produced rather by the tumor microenvironment than by the tumor cells themselves (Figure 4). This suggested that different numbers or an altered composition of lymphoma-infiltrating leukocytes may be present. As depicted in Figure 5A we indeed detected slight changes in the recruitment of CD4 + and CD8 + T cells in STAT3-deficient tumors that however did not meet the criterion of being statistically significant. Numbers of GR1 + CD11b + myeloid cells were comparable. Intriguingly, we found a statistically significant reduction in NK cells in STAT3-deficient tumors (1.1% ± 0.26% in Stat3 fl/fl vs. 0.52% ± 0.09% in Stat3 ∆/∆ tumors, numbers indicate means ± SEM, n = 14). As NK cells produce IFN-γ and TNF-α we reasoned that the decline of NK cell numbers in Stat3 ∆/∆ tumors accounts for the reduced levels of these cytokines (see Figure 4). NK cells kill tumor cells without prior sensitization, which allows testing their cytotoxic ability in vitro. Indeed, STAT3-deficient tumor cells were killed less efficiently than their wild-type controls ( Figure 5B).
These findings indicate that STAT3 expression in lymphoma cells is required to recruit NK cells as well as to evoke their cytotoxic capacity. NK cell cytotoxicity is regulated by a tight balance of inhibitory and activating signals transmitted from a plethora of different NK cell surface receptors [54][55][56]. When determining the levels of activating NKG2D-(MULT-1 and RAE-1) and NKp46-ligands as well as of inhibitory MHC class I (H-2K b and H-2D b ) on the lymphoma cells we failed to detect any differences in either of these ligands (data not shown). At present we lack definitive knowledge on the underlying mechanism that accounts for impaired NK cell cytotoxicity. We hypothesize that the reduced CCL5 levels in Stat3 ∆/∆ tumors contribute to this phenomenon as previous studies have described CCL5 as a NK cell activating chemokine [52]. Taken together, these data indicate that the enhanced lymphoma development after STAT3-deletion is associated with reduced NK cell infiltration and cytotoxicity.

Lymphoma Development in Immune-Compromised Recipients
As our findings show that Stat3 ∆/∆ tumors recruit and activate NK effector cells less efficiently, we tested lymphoma formation in mice lacking specific lymphocyte subsets.
We first transplanted Stat3 fl/fl and Stat3 ∆/∆ BCR/ABL p185+ tumor cells subcutaneously into Nu/Nu mice lacking T cells, but retaining functional B and NK cells for tumor immune surveillance. Comparable to the outcome of our studies in immune-competent wild-type mice ( Figure 3A), transplantation of STAT3-deficient cells into Nu/Nu mice led to significantly enhanced tumor growth ( Figure 6A). In contrast, upon transplantation into Rag2 −/− γc −/− mice that lack B, T and NK cells, the difference in tumor size was abrogated ( Figure 6B). Comparable results were obtained upon transplantation of v-ABL p160+ cells expressing a hairpin targeting STAT3 (STAT3 KD) or a control hairpin (Ctr KD): No difference in tumor size could be detected after transplantation in Rag2 −/− γc −/− mice ( Figure 6C). It is important to mention that the presence of NK cells is not the only difference in Nu/Nu and Rag2 −/− γc −/− mice. In contrast to Rag2 −/− γc −/− mice, Nu/Nu mice still harbor B cells. Additionally, Rag2 −/− γc −/− may differ in their tumor microenvironment due to the lack of cytokine signaling pathways that rely on the presence of the common γ chain such as IL-2, IL-4, IL-7, IL-9, IL-15 and IL-21 [57]. We cannot rule out that the altered cytokine milieu in Rag2 −/− γc −/− contribute to the observed effects. Nevertheless we and others have previously shown that NK cells dominate BCR/ABL + tumor surveillance [27][28][29]58,59]. We thus conclude that STAT3 acts as a tumor suppressor in subcutaneously transplanted BCR/ABL + lymphoma in the presence of functional NK cells.
This led us to the following concept: in BCR/ABL + lymphoma STAT3 mediates the recruitment and activation of NK cells. Upon loss of STAT3, NK cells eradicate tumor cells less efficiently resulting in enhanced tumor growth. In BCR/ABL + lymphoma STAT3 functions as a tumor suppressor as it was shown for brain and intestinal tumors [23,24]. In contrast, in other tumor types STAT3 was described to support tumor growth via the down-regulation of pro-inflammatory cytokines. Moreover, blockade of STAT3 enhanced leukocyte infiltration into the tumor and activated innate immune cells such as macrophages and neutrophils resulting in an increased anti-tumor cytotoxicity [60,61]. Taken together, the entity of studies on STAT3 indicates the following picture: STAT3 cannot be generally judged as a tumor suppressor or tumor promoter-the part STAT3 plays depends on the tumor type and the responsible immune effector cell(s). In BCR/ABL + lymphoma, which are under the control of NK cells, STAT3 acts as a tumor suppressor. Figure 6. STAT3-dependent differences in tumor growth are lost upon transplantation in immune-deficient mice. (A) 10 5 Stat3 fl/fl or Stat3 ∆/∆ BCR/ABL p185+ cells were injected subcutaneously into T cell-deficient Nu/Nu mice. After 11 days tumors were excised and weighted. STAT3-deficient lymphomas were significantly bigger than control lymphomas. Bar graphs depict means ± SEM (n ≥ 24, p < 0.0001, unpaired t-test); (B) 10 5 Stat3 fl/fl or Stat3 ∆/∆ BCR/ABL p185+ cells were injected subcutaneously into Rag2 −/− γc −/− mice. 11 days after transplantation there was no difference in tumor weight. Bar graphs depict means ± SEM (n = 18, p = 0.15, unpaired t-test); (C) 10 5 v-ABL p160+ cell expressing a hairpin targeting STAT3 (STAT3 KD) or a control hairpin (Ctr KD) were injected subcutaneously into Rag2 −/− γc −/− mice. 11 days after injection, tumors developed equally well from both cell lines. Bar graphs depict means ± SEM (n = 10, p = 0.94, unpaired t-test). Asterisks denote statistical significance: **** p  0.0001; n.s. not significant.
For shRNA-mediated knockdown, hairpins targeting Stat3 were lentivirally transduced into wild-type BCR/ABL p185+ and v-ABL p160+ cells. The encoded hairpin (expressed in a pLKO.1-puro-vector) anneals with the mRNA region transcribed by exon 11 thereby degrading both splice variants of STAT3 (Stat3α and Stat3β). The TRC clone (TRCN0000071456) was obtained from MISSION ® shRNA. A clone containing a validated non-silencing shRNA served as control. Selection of vector-expressing cells was accomplished by puromycin selection (1 µg/mL) for 7 days. Efficiency of shRNA-mediated STAT3 knockdown was verified by western blotting.

Transplantation of Tumor Cells
10 5 BCR/ABL + cells were injected subcutaneously into the flanks of wild-type, Nu/Nu and Rag2 −/− γc −/− mice. After 11 days or when tumors reached a maximum size of 3 cm 3 , mice were sacrificed and tumors excised. For flow cytometric analysis tumors were mashed through a 70 µM filter.

Conclusions
Therapeutic agents blocking STAT3 are currently under development and predicted to block tumor progression in a variety of malignancies. We show here that in BCR/ABL + tumors the absence of STAT3 accelerates tumor growth assigning a tumor suppressing function to STAT3 in this disease. We found that STAT3 is required to attract NK cells to the tumor, which restrict and limit tumor growth. In the absence of STAT3 tumor surveillance is significantly impaired. Caution should be taken when targeting STAT3 in lymphoma as it may provoke adverse effects on immune surveillance in these patients.