Dendritic cell-derived interleukin-15 is crucial for therapeutic cancer vaccine potency

IL-15 supports improved antitumor immunity. How to best incorporate IL-15 into vaccine formulations for superior cancer immunotherapy remains a challenge. DC-derived IL-15 (DCIL-15) notably has the capacity to activate DC, to substitute for CD4+ Th and to potentiate vaccine efficacy making IL-15-based therapies attractive treatment options. We observed in transplantable melanoma, glioma and metastatic breast carcinoma models that DCIL-15-based DNA vaccines in which DC specifically express IL-15 and simultaneously produce tumor Aghsp70 were able to mediate potent therapeutic efficacy that required both host Batf3+ DC and CD8+ T cells. In an inducible BrafV600E/Pten-driven murine melanoma model, DCIL-15 (not rIL-15)-based DNA vaccines elicited durable therapeutic CD8+ T cell-dependent antitumor immunity. DCIL-15 was found to be superior to rIL-15 in “licensing” both mouse and human DC, and for activating CD8+ T cells. Such activation occurred even in the presence of Treg, without a need for CD4+ Th, but was IL-15/IL-15Rα-dependent. A single low-dose of DCIL-15 (not rIL-15)-based DC vaccines induced therapeutic antitumor immunity. CD14+ DC emigrating from human skin explants genetically-immunized by IL-15 and Aghsp70 were more effective than similar DC emigrating from the explants genetically-immunized by Aghsp70 in the presence of rIL-15 in expressing membrane-bound IL-15/IL-15Rα and activating CD8+ T cells. These results support future clinical use of DCIL-15 as a therapeutic agent in battling cancer.


Introduction
Generating robust, durable and effective tumor antigen (Ag)specific CD8 C T cells that are competent to eradicate primary tumors and tumor metastases or to prevent disease recurrence as a consequence of active specific vaccines has proven clinically difficult. 1 This may relate to significant hurdles including the inherently poor immunogenic nature of many tumors and tumor-induced immune suppression mediated by myeloidderived suppressor cells (MDSC) and regulatory T cells (Treg). [1][2][3] Deficiency of CD4 C T helper (CD4 C Th) cell numbers and/or functionality, which are usually needed to optimize CD8 C T cell responses, 4 has further dampened hope for the generalized effectiveness of conventional vaccine strategies in the vast majority of cancer patients.
Dendritic cells (DC), key players in the host handling of injected vaccines (e.g., professionally processing and presenting vaccine Ag and functionally polarizing cognate Ag-specific T cells), are crucial for activating potent Ag-specific T cell responses. 1 In vitro-generated DC or in vivo DC-targeting therapeutic vaccines may be designed in a manner that effectively promotes the induction of clinically-relevant Type-1 antitumor CD8 C T cells in a manner that does not require the participation of CD4 C Th cells that are likely functionally sub-optimal or inappropriately skewed (e.g., induced Treg) in the tumor-bearing host.
Interleukin (IL)-15, a priority agent for cancer therapy, 5 has been explored to improve the efficacy of vaccines, chemotherapies and adoptive T cell transfer approaches due to its ability to support DC, B cell, T cell and NK cell functionality, and to rescue tolerant or dysfunctional CD8 C T cells. [6][7][8][9][10][11][12] Unfortunately, high-doses of IL-15 (necessary for its bioactivity in vivo) via systemic administration of recombinant IL-15 protein (rIL-15) or overexpression of transgene IL-15 have untoward side-effects [e.g., stimulating tumor cell growth, activating negative regulators (e.g., programmed death-1) in CD8 C T cells, exacerbating xenogeneic graft-vs.-host-disease or autoimmunity, and functioning as an "oncogene" resulting in progressive CD8 C T or NK leukemia], [13][14][15][16][17] which have served to limit its benefit-to-risk ratio in the clinic, despite pre-clinical findings supporting the safety of rIL-15 in rhesus macaques. 18   agonists (e.g., IL-15/IL-15Ra-Fc complex and IL-15/IL-15Ra fusion protein) reduce the dose of delivered IL-15 required to reach biologicallymeaningful levels in vivo. [19][20][21][22][23] However, cell (particularly DC) contact-dependent trans-presentation of membrane-bound IL-15/IL-15Ra appears required for optimal IL-15-mediated signaling in vivo. 24,25 In vivo, IL-15 derived from DC (DCIL-15) can "auto"-activate DC and substitute for the functional licensing events normally associated with DC interaction with CD4 C Th during vaccine activation of durable high-avidity CD8 C T cells, even though the mechanisms underlying this biology remain unknown. 10,[26][27][28][29][30] IL-15 is produced by cells (e.g., DC) at very low levels under normal physiologic conditions. The in vivo delivery of transgene IL-15 into DC, which co-express full-length transgenic tumor Ag to allow for simultaneous DC presentation of Ag to T cells, may result in safer and more effective therapeutic vaccines that constitute an urgent, but as yet unmet, clinical need.
We have developed a novel DCIL-15-based cancer vaccine platform in which DC specifically express human IL-15 transgene and simultaneously produce tumor Ag fused to human heat shock protein 70 (Aghsp70) as a specific immunogen, and demonstrated its potent antitumor effects in the prophylactic setting. 10 In the current study, we found that DCIL-15 was superior to rIL-15 in improving both murine and human DC functions and potentiating therapeutic cancer vaccine efficacy in multiple clinically-relevant transplantable murine tumor models, and, notably, in the genetically engineered Braf V600E /Pten-driven melanoma murine model that recapitulates human disease. 31 Results from these pre-clinical studies support the translational potential of this vaccine strategy for the future treatment of patients with cancer.
Human skin-derived DC: Human skins from surgical discard were obtained in accordance with the guidelines and the protocol approved by the Institutional Review Board of the University of Pittsburgh. Human skin epidermal/dermal explants were freshly prepared from skins with skin graft knife and subsequently untreated or immunized by 0.6 mm gold particles (BioRad) conjugated with IL-15/M7 or M7 DNA using a gene gun (GG) in sterile conditions. 40 Clinical-grade rhIL-15 (10 ng in 10 ml AIM V medium) was intradermally (i.d.) injected into »4-5 cm 2 M7-immunized human skin explants or M7-immunized human skin explants were cultured in AIM V medium supplemented with 10 ng/mL rhIL-15 or clinical-grade rhIL-15 (rIL-15/M7). After vaccination, these explants were cultured on sterile steel mesh with the epidermal side up in AIM V medium including 1£antibiotic/antimycotic solution at 37 C in 5%CO 2 . 3 d later, skin emigration cells were harvested from culture medium. Skin DC harvested from untreated or DNA-immunized human skin explants culture medium were stained by anti-HLA-DR-alexa flour 488, -CD14-brilliant violet 570, -IL-15-percep-cy5.5, and IL-15Ra-PE or isotype antibodies (BD Biosciences, eBioscience, Biolegend) and analyzed by flow cytometry on a BD LSRII. CD14 C DC were isolated from harvested skin DC using The EasySep TM Human CD14 Positive Selection Kit (Stem Cell Technologies).
Genetically engineered Braf V600E /Pten-driven melanoma model Braf V600E /Pten-driven melanoma, which was developed by inducing oncogene Braf V600E expression with 4-HT in B6-Tyr-Cre ERT2 Braf CA Pten lox/lox mice with correct genotype, 31 was allowed to grow progressively to a mean tumor size of »3 mm, at which time, melanoma-bearing mice were randomized into cohorts of 3-4 mice with each cohort exhibiting a comparable mean tumor size. Mice were then left untreated or they were vaccinated using a GG with DCIL-15/T7 or T7 DNA on days 0, 7 and 14. 10,43 Melanoma-bearing mice immunized by T7 DNA were i.p. injected with clinical-grade rhIL-15 for 3 d immediately after each immunization as described above. Endogenous CD8 C T cells were depleted by i.p. injection of anti-mouse CD8 mAb 1 day before, on the day of, and 1, 3 d after first vaccination, and then weekly.
In all therapeutic experiments, tumors were measured every 3 d using a digital slide calipers (Fisher Scientific, Pittsburgh, PA) in the two perpendicular diameters. Mice were followed until their unanticipated (natural) death or they were euthanized when tumor reached a mean size of 10 mm. On day 30 after 4T1.2-Neu inoculation, mice were sacrificed and lungs were fixed with Bouin's solution (Sigma) for counting tumor foci. At the same time, tumors in BM were selected in vitro with 4T1.2-Neu culture medium as described above. 35 Statistical analysis Data were statistically analyzed using Student's t-test (immune assays, tumor size and loci) (Graph Pad Prism version 6). Data from animal survival experiments were statistically analyzed using Log rank test (Graph Pad Prism version 6). Animal survival is shown by Kaplan-Meier Survival Curves. P < 0.05 is considered to be statistically significant. *p <0.05; **p <0.01; ***p <0.001; N.D. (not detected).

Results
DCIL-15-based DNA vaccines elicit potent CD8 C T celldependent therapeutic antitumor immunity in multiple clinically-relevant murine tumor models We have designed a novel DCIL-15-based cancer vaccine platform in which DC specifically express human IL-15 transgene and simultaneously produce tumor Aghsp70, and demonstrated its efficacy in effectively inducing prophylactic antitumor immunity. 10 When incorporating the tumor-associated Ag TRP2 or NeuED, we found that DCIL-15-based combination DNA vaccines elicited potent therapeutic antitumor immunity against distinct murine established tumors (i.e., 8 d after s.c. inoculation of a lethal-dose of the various tumor cells) including syngeneic native melanoma (B16) (Fig. 1A), glioma (GL26, naturally expressing TRP2) 34 (Fig. 1B and C), and spontaneous metastatic breast tumor (4T1.2-Neu, 4T1.2 ectopically expressing activated onco-antigen rat Neu) 35 (Fig. 1D-F, Table 1). Furthermore, antibody-based CD8 C T cell depletion abrogated the protective effects associated with vaccination ( Fig. 1B-E). These results suggest the broad therapeutic potency of this vaccine strategy in eliciting host protective CD8 C T cell responses against primary and metastatic tumors.
Therapeutic antitumor immunity elicited by DCIL-15based DNA vaccines depends on host Batf3 C DC subsets and partially requires the pDC subset of antigen-presenting cells Batf3 ¡/¡ mice are selectively deficient of the antigen crosspresenting CD8a C and CD103 C DC subsets, 44 which are  required for in vivo priming of tumor-specific CD8 C T cells. 45,46 Using these mice as tumor-bearing hosts, we observed that therapeutic antitumor immunity elicited by DCIL-15-based DNA vaccines required host Batf3 C DC subsets in both the GL26 and 4T1.2-Neu models ( Fig. 2A and B). Since IL-15 may mediate cross-talk between pDC and conventional DC (cDC) during the course of specific T cell activation, 47 we next investigated the requirement for pDC in vaccine efficacy. Depletion of pDC during the immunization protocol partially impaired the ability of therapeutic intervention to promote antitumor immunity (Fig. 2C). These data indicate that the optimal therapeutic antitumor immune response promoted by this immunization approach requires host Batf3 C DC and the participation of pDC. DCIL-15 (not rIL-15)-based DNA vaccines elicit robust durable therapeutic CD8 C T cell responses in a clinically-reflective Braf V600E /Pten-driven melanoma model. DCIL-15-based DNA vaccines were therapeutically effective in transplantable murine B16 melanoma, GL26 glioma, and 4T1.2-Neu breast carcinoma models (Fig. 1, Table 1). Given the contention that findings from transplantable tumor models may not provide the most useful information for translation into the clinic, we next evaluated a genetically engineered inducible Braf V600E /Pten-driven melanoma murine model that is more reflective of human disease (e.g., tumor self Ag expression, highly metastatic, and relapses post chemotherapies). 31 Melanoma in this model develops when oncogenic Braf V600E expression in melanocytes is induced by treatment with 4-HT. 31 As observed in many aggressive solid tumors, Foxp3 C CD4 C Treg infiltrated into Braf V600E /Pten-driven melanoma and melanoma-associated tdLN and intratumoral Treg exhibited potent activity of suppressing T cell activation (Fig. S1).
DC genetically-modified to express IL-15 and Aghsp70 are superior to the DC genetically-modified to express only Aghsp70 and cultured with rIL-15 in activating CD8 C T cells IL-15 gene therapy and rIL-15 are being used to generate ex vivo DC vaccines for cancer immunotherapy in preclinical models and clinical trials/studies. 8,10,11 We have previously shown that murine BM DC genetically-modified to express IL-15 and Aghsp70 secrete substantial levels of human IL-15 and produce Aghsp70, in association with their enhanced maturation as evidenced by higher levels of cell surface IL-15Ra expression. 10 Although both murine BM DC and human moDC geneticallyengineered by Aghsp70 (i.e., N7 for murine DC, M7 for human DC) and cultured with rIL-15 (without other maturation factors) (rIL-15-based DC) were able to activate syngeneic/autologous CD8 C T cells without the need for CD4 C Th "help," DC genetically-modified to express both IL-15 and Aghsp70 (DCIL-15-based DC) were much more effective in activating cognate syngeneic CD8 C T cells (Fig. 4). Neutralization of IL-15/IL-15Ra signal using specific anti-IL-15 and -IL-15Ra antibodies during the in vitro DC-CD8 C T cell coculture abolished DCinduced activation of CD8 C T cells (Fig. 4), and such activation circumvented the suppressive effects of Treg (Fig. 4). These results suggest that DCIL-15 is superior to rIL-15 in activating DC for the activation of CD8 C T cells without a need for CD4 C Th "help," but in an IL-15/IL-15Ra-dependent manner that overcomes suppression mediated by Treg.

DCIL-15 (not rIL-15)-based DC attenuate tumor-associated
Treg and induce therapeutic antitumor immunity in an IL-15Ra-or IL-2Rb-dependent manner that does not require endogenous production of IL-15 or CD4 C T cell "help" Without the addition of alternate DC maturation signals, DCIL-15 (but not rIL-15)-based DC were observed to attenuate the suppressive activity of tumor-associated Treg in vitro (Fig. 5A). In vivo, a single low-dose of the DCIL-15 (not rIL-15)-based DC vaccines generated therapeutic antitumor  immunity even in the CD4 C Th-deficient tumor-bearing hosts (Fig. 5B). Therapeutic antitumor immunity induced by DCIL-15-based DC vaccines was significantly associated with expression of IL-15Ra or IL-2Rb on vaccine DC (Fig. 5C and D). Furthermore, DCIL-15-modified IL-15 ¡/-DC were capable of generating effective therapeutic antitumor effects (Fig. 5D). CD14 C DC emigrating from human skin explants genetically-immunized by IL-15 and M7 are more effective than the DC emigrating from the explants genetically-immunized by M7 alone in the presence of rIL-15 in expressing membranebound IL-15/IL-15Ra and activating CD8 C T cells.
Skin harbors multiple DC subsets and is an ideal anatomic site for vaccination of Type-1 immune responses. DC that leave human skin explants in vitro are believed to be analogous to tissue DC that travel to LN in vivo after locoregional antigenic insult. CD14 C DC (not CD14 ¡ DC) emigrating from human skin explants genetically-immunized by IL-15 and M7 enhanced membrane-bound IL-15/IL-15Ra expression when compared to the explants untreated or genetically-immunized by M7 in the presence of rIL-15 ( Fig. 6A and B, data not shown). Although both CD14 C DC and CD14 ¡ DC emigrating from human skin explants geneticallyimmunized by IL-15 and M7 were more effective than the DC emigrating from the explants genetically-immunized by M7 alone in the presence of rIL-15 in activating allogeneic CD8 C T cells (Fig. 6C), CD14 C DC were superior to CD14 ¡ DC in such activation. These results suggest that this vaccination strategy may be useful in potentiating the ability of human skin DC to drive therapeutic CD8 C T cell responses in the cancer setting.

Discussion
IL-15, one of the most promising biological agents for cancer treatment, 6 has been used in the in vitro culture (via rIL-15) or modification (via transgene IL-15) of CD8 C T cells, DC or tumor cells for generation of tumor-specific CD8 C T cells and DC-or tumor cell-based vaccines, and administered systemically in vivo by injection of rIL-15 or expression of transgene IL-15 alone or in combination with other agents/approaches (e.g., IL-12, IL-21, IL-7, GM-CSF, anti-CD40, anti-PD-L1, anti-CTLA-4, adoptive CD8 C T cell transfer, irradiation). [6][7][8][9][10][11][12][48][49][50][51][52] The therapeutic antitumor efficacy of these IL-15-based strategies has been demonstrated in transplantable tumor models. The phase I/II clinical trials with rIL-15 are currently being performed in patients with various forms of cancer. 52 The unique in vivo mechanism of action of IL-15 occurs through cell, notably, DC contact-dependent trans-presentation of a membrane-bound IL-15/IL-15Ra signal to effector cells (e.g., NK and CD8 C T cells expressing IL-15Rb/g receptor) that is required for optimal IL-15-mediated signaling under physiologic conditions. 6,24,25,53 Accordingly, ex vivo-generated DC or tumor cells co-expressing transgene IL-15 and IL-15Ra, and engineered IL-15 agonists (e.g., IL-15/IL-15Ra-Fc complex, IL-15/IL-15Ra fusion protein) have been explored to generate effective antitumor responses. 14,[19][20][21][22][23]51 DCIL-15 activates DC and provides signals equating to those normally provided by CD4 C Th in "licensing" DC for CD8 C T cell activation even though the underlying mechanisms associated with this biology remain largely unknown. 10,[26][27][28][29][30] Directed in vivo delivery of transgene IL-15 into vaccine DC has the potential to enhance the potency of DCIL-15 provided in a trans-presented manner from vaccine DC to responder CD8 C T cells. Regulated provision of transgene IL-15 in this manner may also obviate, or at least reduce, potential adverse events associated with nonspecific overexpression of transgene IL-15 in vivo. [14][15][16][17] We observed that DCIL-15 was superior to rIL-15 in promoting both murine and human DC functionality and therapeutic cancer vaccine efficacy in multiple distinct and clinically-relevant murine tumor models ( 10 , Figs. 1-6), suggesting the broad therapeutic potency of this vaccine strategy in eliciting antitumor immunity against primary tumors and their metastases.
DC-derived intracellular IL-15 interacting with IL-15Ra (in cis) during production within DC results in mutual stabilization and increased bioactivity of membrane-bound IL-15/IL-15Ra on DC that may promote IL-15 action on IL-2Rb (another receptor of IL-15) expressed by CD8 C T cells, leading to efficient IL-15/ IL-15Ra trans signaling in support of enhanced antitumor CD8 C T cell activation. The lack of response to IL-15 stimulation by IL-2Rb ¡/¡ -DC suggests that IL-2Rb expressed by DC may play an important role in IL-15-mediated signaling. 54 IL-15Ra substantially increases the affinity of IL-15 for IL-2Rb, and this allosteric interaction is required for effective IL-15-mediated signaling into T cells. 55 Thus the strategic enhanced production of transgene IL-15 by IL-15Ra C DC likely plays a major role in the superior ability of DCIL-15-based cancer vaccines to promote robust antitumor CD8 C T cell responses including those from central memory CD8 C T cells (CD8 C CD44 hi CD62L hi ) and effector memory CD8 C T cells (CD8 C CD44 hi CD62L lo ) in tdLN (Fig. S3). DC modified by DCIL-15/Aghsp70 produced substantial levels of pro-inflammatory cytokines (e.g., IL-6), 10 which may also assist in the inactivation of Treg. 56 CD4 C Th are generally considered indispensable during the induction of optimal CD8 C T cell responses, however, our analyses of DCIL-15-based vaccines indicated that CD4 depletion did not impair vaccine efficacy in either the prophylactic 10 or therapeutic settings (Fig. 5C). DCIL-15-based vaccines induced durable Type-1 (i.e. IFNg-producing) CD8 C T cells reactive against the melanoma-associated Ag TRP2, even in CD4 C T cell-deficient mice (Fig. S4). The ability of this vaccine to work in the absence of CD4 C T "helper" cells may be translationally important in the context of cancer patients (e.g., HIV-infected and some after chemotherapy) who are frequently deficient in (Type-1) CD4 C Th function. 57,58 CD14 C DC emigrating from human skin explants geneticallyimmunized by IL-15 and Aghsp70 enhanced their cell surface expression of IL-15/IL-15Ra and readily activated CD8 C T cells.
Although it remains to be determined whether HLA-DR C CD14 C IL-15 C /IL-15Ra C DC that emigrate out of human skin explants after genetic-immunization using this vaccine strategy promote the differentiation of melanoma Ag-specific CD8 C T cells in vitro and in humanized (human melanoma-bearing severe combined immunodeficiency) mice, our current data support the notion that this approach could be used to potentiate human skin DC for the effective genetic immunization against cancer.
Furthermore, although the current study focused on the influences of DCIL-15-based cancer vaccines on CD8 C T cells, NK cells are also responsive to IL-15, and the impact of DCIL-15based cancer vaccines on Type-1 NK cells needs to be comprehensively evaluated in future studies.
Although the precise mechanisms behind the DCIL-15-based vaccine strategy need to better delineated, its merits are evident: a) Directed in vivo delivery of transgene IL-15 into vaccine IL-15Ra C DC focuses DCIL-15 on activating DC, substituting for CD4 C Th, and potentiating vaccine efficacy. This may lead to the most efficient utilizing the cytokine IL-15 in vivo while minimizing concerns for off-target toxicities associated with systemic administration of rIL-15 or overexpression of transgene IL-15. b) The vaccines were therapeutically effective even in the CD4 C Th-deficient mice and this induction strategy circumvented Treg-mediated suppression. This may be important to the design of vaccines for cancer patients who are deficiency of CD4 C Th and/or display profound immune suppression. c) The possibility of a single-immunization of either small numbers of ex vivo-generated DCIL-15-based DC or in vivo DC-targeting DNA vaccines required to induce effective antitumor immunity avoids or at least minimizes the need for secondary immunizations that makes this vaccine strategy both feasible and translationallyattractive from a logistics perspective. d) The in vivo mouse data from multiple highly clinically-relevant tumor models including the authentic Braf V600E /Pten-driven melanoma and the in vitro human data from both blood-and skin-derived DC support the translational potential of this vaccine strategy in the clinic.
Given the current priority status for the clinical use of rIL-15 in patients with cancer and our findings for the superior performance of DCIL-15 over rIL-15 in cancer vaccine formulations, and DNA vaccines offering the potential for an off-the-shelf, easily-scalable vaccine platform, this vaccine strategy is both salient and attractive for translation as a future cancer therapeutic modality that could benefits the vast majority of cancer patients. This is particularly compelling for those patients with profound defects in CD4 C Th "helper" responses (or robust suppressor cell populations).

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed. Pittsburgh), X Huang (University of Southern California) for providing plasmids, J White (University of Colorado) for preparing samples, and K Meeth (Yale University) for helping in the Braf/Pten-driven melanoma model.

Funding
This work was supported by Department of Dermatology at The University of Pittsburgh (ZY) and NIH grants R01CA108813 and R01CA108813-04S2 (ZY), P50CA121973, R01AI076060 and R01EB01277 (LDF).

Supplemental Material
Supplemental data for this article can be accessed on the publisher's website.

Author Contributions
YZ, ST, ZL, JZ and MZ performed the experiments; MWB, RMK and TAW provided the critical materials; YZ, ST, ZL, WJS, LDF and ZY analyzed the data; TAW, WJS and LDF reviewed the manuscript; and ZY supervised the study and wrote the paper.