Gemcitabine Activates Natural Killer Cells to Attenuate Pancreatic Cancer Recurrence

With an overall 5-year survival rate of only 7%, pancreatic ductal adenocarcinoma (PDA) has now eclipsed breast cancer as the third leading cause of cancer death in the United States¹ and it is predicted to be second by 2030.² Despite improvements in surgical and clinical management, the majority of patients with localized PDA who undergo successfully R0 surgical resection (ie, resection followed by absence of cancerous cells seen microscopically at the margins) succumb to recurrent local or distant metastatic disease. Clinical trials demonstrate single-agent 5-fluorouracil or gemcitabine adjuvant chemotherapy provides overall and disease-free survival benefit, and further adjuvant trials are addressing the efficacy of multi-agent chemotherapy regimens shown to be beneficial in advanced, inoperable PDA.^3,4 The number and types of adjuvant trials has lagged far behind those for advanced stage PDA, in part owing to the limited number of patients who undergo surgery. Assessment of adjuvant systemic therapies are complicated by further variables introduced by the surgical procedure, inconsistencies in pathologic evaluation (ie, margin status), and extended length of study.⁴ Although it may be appropriate in some instances to extrapolate clinical trial findings for advanced PDA to the adjuvant setting, differences in tumor burden along with the physiologic differences after surgical resection of primary tumor are biological confounders and important caveats to such cross-comparisons.

One approach for accelerating the development and validation of most promising forms of adjuvant therapy would be initial assessment in preclinical models. Genetically engineered mouse models of PDA have been key factors in advancing our understanding of PDA biology, providing unique insights into the role of the tumor microenvironment and inflammatory response, while also serving as important platforms for assessing therapeutic agents.⁵ However, most PDA genetically engineered mouse models best recapitulate locally advanced or metastatic disease. They are also further complicated by variable time to onset, tumor multifocality, and other histologic abnormalities such as acinar to ductal metaplasia and dysplasia that diffusely involves the pancreas. PDA mouse models amenable to surgical resection and adjuvant treatment while also replicating the tumor microenvironment and intact immune response in which PDA naturally arise have been lacking.

Gürlevik et al⁶ now present a rapid and flexible system for reproducibly modeling a localized form PDA amenable to surgical resection and suitable for assessing the efficacy adjuvant systemic therapies in relation to tumor recurrence and overall survival. In this model, a plasmid carrying a Cre recombinase was electroporated using tweezers-type electrode in the pancreas of LSL-Kras-G12Dxp53fl/fl mice. These animals developed single tumor nodules at the site of electroporation. Histologically, the tumors showed moderately differentiated PDA at day 21 postelectroporation without any observable metastatic spread. To closely resemble the human PDA, the authors introduced to the model a transposon encoding constitutively active form of Akt2, a kinase known to promotes invasive growth and metastasis.^7,8 Tumors carrying activated Kras-G12D, Akt2 and p53 deletion showed accelerated primary tumor growth, local invasion and distant metastasis and significantly reducing survival, while preserving the overall histologic features of PDA, including its stromal microenvironment.

In their model, mice receiving adjuvant gemcitabine after R0 surgical resection had an improvement in overall survival attributable to a reduction in local, but not metastatic, tumor recurrence—a finding that parallels that observed for patients who received adjuvant gemcitabine in the large phase III CONKO-001 clinical trial that established the efficacy of gemcitabine in the adjuvant setting.⁹ Beyond offering important proof-of-principle for the model itself, this provocative observation raises important questions about the therapeutic mechanism of action for gemcitabine, given its effects were limited to the control of residual/recurrent local disease without apparent influence on metastatic disease.

Accumulating data in PDA and other cancer types indicate cytotoxic chemotherapy facilitates protumorigenic and antitumorigenic effects linked to the stromal microenvironment and immune response. The immunomodulatory effects of chemotherapy range from suppressive to stimulatory depending upon tumor type, specific cytotoxic agent, and its dosing regimen.^10,11 For instance, gemcitabine given at concentrations similar to or below those used for patients can attenuate immune suppression through reductions in either total numbers or phenotypic subsets of myeloid-derived suppressor cells (MDSCs).¹⁰ Low-dose gemcitabine reduces regulatory T-cell (Treg) numbers to modestly improve survival in a murine orthotopic PDA model.¹² Low-dose (metronomic) cytotoxic chemotherapy can selectively ablate Tregs relative to other T-cell populations,¹¹ offering rationale for the use of low-dose cyclophosphamide in combination with GVAX and CRS-207 vaccines in recent promising clinical trial for metastatic PDA.¹³ Furthermore, higher dose gemcitabine selectively eliminates splenic MDSCs and increase the anti-tumor activity of natural killer (NK) and CD8⁺ T cells in mice bearing large tumors,¹⁴ whereas data from PDA patients receiving gemcitabine points to reduction in the number of Tregs¹⁵ and increased CD14⁺ monocytes and myeloid dendritic cells in peripheral blood (PB).¹⁶

Gürlevik et al⁶ extend these observations to the adjuvant treatment setting to identify a novel immunomodulatory activity for gemcitabine localized to the remnant pancreas after R0 surgical resection. Adjuvant gemcitabine selectively reduced the Cd11b⁺ Gr1^lowF4/80⁺ subset of MDSCs and increased the number of NK cells without altering Tregs or macrophage polarization (Figure 1). These changes were specific to remnant pancreas and not observed in PB. Depletion of NK cells, but not CD8⁺ T cells, reversed survival advantage and inhibition of local tumor recurrence seen with gemcitabine, mechanistically linking its therapeutic action to an enhanced NK cell response after tumor resection.⁶

Figure 1.

Single nodule tumors (A) after R0 surgical resection received adjuvant gemcitabine (C) or vehicle treatment (B). The gemcitabine treated tumors showed lower numbers of myeloid-derived suppressor cells (MDSCs) and increase numbers of natural killer (NK) cells. These findings explain in part the lower local recurrence seen with gemcitabine in clinical trials. PDA, pancreatic ductal adenocarcinoma.

NK cells bridge innate and adaptive immunity and mediate antitumor activity through constitutive cytotoxic activity and secretion of immunomodulatory cytokines. NK cell number and functional activities are subject to modulation by cellular and soluble factors and their engagement of activating or inhibiting receptors on NK cells.¹⁷ Although increased NK cells observed in their adjuvant model would seem to be tied to alterations in MDSCs and their immunosuppressive activity, it is possible that gemcitabine may have also acted directly on residual tumor cells to modulate their susceptibility to NK immune surveillance. For example, gemcitabine can up-regulate the expression of NK group 2 member D (NKG2D) ligand on lung cancer cells through a DNA stress/ATM-dependent mechanism, leading to augmentation in NK tumor cell recognition and cytolysis.¹⁸

Gemcitabine prosurvival effects on local disease recurrence in this adjuvant model bring focus on the implication of primary tumor resection not only in terms of eradicating immunomodulatory activity mediated by the primary tumor, but also in relation to local and systemic inflammatory responses induced by the surgical procedure itself. A clinical study of NK cell function in PB after pancreaticoduodenectomy for PDA finds a transient reduction in NK cytotoxicity within the first week after surgery followed by a subsequent normalization or augmentation of NK activity at 30 days, the magnitude of which was correlated with improved survival.¹⁹ With patient tissue evaluation not possible in the postoperative setting, the adjuvant mouse model presented here by Gürlevik et al highlights the further likelihood of a clinically important, tissue-localized NK responses that may differ from that seen in PB and which could be augmented to further therapeutic advantage through the use of cytotoxic agents or other immune modulators.

As mentioned, both the timing and dose of gemcitabine or other cytotoxic agent may be further critical variables in unmasking an innate and/or adaptive immune response able to inhibit local disease recurrence or subsequent metastasis. Of potential importance, the initiation of gemcitabine immediately after surgery in this adjuvant model contrasts the clinical scenario in which patients only start adjuvant chemotherapy after a period of recovery from surgery. It is unclear whether a unique window of opportunity exists in the period immediately before, during, or after surgery to most effectively intervene with a therapy that augments local immune surveillance by NK cells. Beyond this issue, there are known differences in functional NK cell subpopulations present in PB or secondary lymphoid tissues, as well as differences in NK cell functional activity observed between PB, the primary tumor, the premetastatic niche, and established metastatic tumor sites.¹⁷ These differences are likely to have important implications for NK cell-targeted agents in patients with either early or later stage PDA and could be explored further in this adjuvant model by instead introducing genetic alterations that mediate slower disease kinetics to allow their evaluation.

Indeed, the flexibility and apparent ease by which different combinations of molecular alterations can be introduced readily into this electroporation-based model are one of its greatest strengths, offering the opportunity to mechanistically assess the importance that individual or combination alterations play in regulating various aspects of PDA biology, including modulation of an immunostimulatory or immunohibitory tumor microenvironment. Such mechanistic insights could eventually be leveraged to identify prognostic or predictive biomarkers useful in individually tailoring PDA immunotherapy regimens based on a combination of tumor profiling and immune cell analysis of tumor and/or blood samples. Finally, although not addressed in their paper,⁶ this model also offers a suitable platform for exploring and validating the use of neoadjuvant therapeutic approaches before surgery, an area of intense interest and debate in the field.