Synthetic vaccine particles for durable cytolytic T lymphocyte responses and anti-tumor immunotherapy

We previously reported that synthetic vaccine particles (SVP) encapsulating antigens and TLR agonists resulted in augmentation of immune responses with minimal production of systemic inflammatory cytokines. Here we evaluated two different polymer formulations of SVP-encapsulated antigens and tested their ability to induce cytolytic T lymphocytes (CTL) in combination with SVP-encapsulated adjuvants. One formulation led to efficient antigen processing and cross-presentation, rapid and sustained CTL activity, and expansion of CD8+ T cell effector memory cells locally and centrally, which persisted for at least 1–2 years after a single immunization. SVP therapeutic dosing resulted in suppression of tumor growth and a substantial delay in mortality in several syngeneic mouse cancer models. Treatment with checkpoint inhibitors and/or cytotoxic drugs, while suboptimal on their own, showed considerable synergy with SVP immunization. SVP encapsulation of endosomal TLR agonists provided superior CTL induction, therapeutic benefit and/or improved safety profile compared to free adjuvants. SVP vaccines encapsulating mutated HPV-16 E7 and E6/E7 recombinant proteins led to induction of broad CTL activity and strong inhibition of TC-1 tumor growth, even when administered therapeutically 13–14 days after tumor inoculation in animals bearing palpable tumors. A pilot study in non-human primates showed that SVP-encapsulated E7/E6 adjuvanted with SVP-encapsulated poly(I:C) led to robust induction of antigen-specific T and B cell responses.


Introduction
Immunotherapy has become one of the most promising fields in cancer research, with checkpoint inhibitors becoming a standard of care against several types of cancer [1][2][3] and engineered T cells scoring successes in early clinical trials [3][4][5]. However cancer vaccines have been a notable exception, with a number of late stage clinical trial failures [6][7][8]

In vivo tumor inoculation
Cells were injected s.c. (intrascapular, 5×10 4 /mouse). Mice were monitored daily for health status and 2-3 times a week for tumor progression. The tumor volume determination was performed by external caliper measurement of length and width of tumor and calculation of the volume by use of the modified ellipsoid formula -½(Length × Width 2 ). Animals were euthanized if one of the measurements exceeded 20 mm or if the animal became moribund. Specifically, humane endpoints were used in all tumor studies (death as an endpoint was never utilized). In addition to tumor measurements being taken 2-3 times a week by trained scientific personnel and animals being euthanized if any measurement exceeded 20 mm (or if an animal was judged as moribund by phenotypical features), all mice have been health-checked by a trained animal technician on a daily basis and if judged moribund, euthanized during the same day (<8 hours). Moreover, in all the studies involving TC-1 cell line, all mice were weighed prior to and during the experiment and the weight loss of >15% was used as another criterion for euthanasia (EG.7-OVA and B16-F10 lines are not known to cause a weight loss within defined tumor size and they never did in any of studies reported herein). No animals died before meeting euthanasia criteria with exception of extremely rare events not exceeding spontaneous mortality rate in laboratory mice of the same age and strain. Duration of each experiment is shown in each figure in Results section, but overall short-term experiments involving aggressive tumor cell lines or SVP of limited immunogenicity lasted 45-80 days, while long-term experiments using TC-1 cell line and highly immunogenic SVP lasted 100-500 days. All short-term tumor treatment experiments used 5 mice per group in a single study, while long-term TC-1 studies involved 7-8 mice per group with their mortality rate shown in corresponding figures. All euthanasia procedures were performed by cervical dislocation under isoflurane anesthesia according to AVMA Guidelines for the euthanasia of animals (Section 2.2). Most animals surviving tumor challenge were not euthanized at the end of the study, but kept under daily observation and used later in long-term immune memory studies as described in Results.

Vaccination
Mice were injected s.c. in both hind limbs (30 μl volume per injection site, 60 μl total) with PBS vehicle containing SVP-formulated or free antigens and adjuvants. A single time-point injection was used in cytokine production and in vivo CTL assay experiments, and primeboost regimens starting at d3-14 post tumor inoculation with the 1 st boost on day 3 or 4 after the prime followed by two additional weekly boosts were used when assessing anti-tumor efficiency.

In vivo cytotoxicity
Antigen-specific cytolytic activity in vivo was determined as described [15] at different timepoints after a single SVP immunization. Other immunological readouts are described in detail in the Supporting information Materials and Methods (S1 Methods).

NHP studies
All procedures in NHP study were in full compliance with the U.S. Department of Agriculture's (USDA) Animal Welfare Act ( (Fig 1A and 1B). Further analysis showed that the SVP[OVA]-PLGA formulation induced a significant increase in the percentage of effector memory T cells (CD62L low CD44 + ) in the spleen (>3-fold) and in local lymph nodes (>2.5-fold) compared to the SVP[OVA]-PLA formulation (Fig 1C and 1D). No difference was noted between the SVP formulations in the induction of central memory T cell populations (CD62L high CD44 + ).

Nanoparticle encapsulation of model antigen in PLGA leads to induction of durable effector CD8 + CTL responses and promotes sustained anticancer activity in vivo
Both SVP formulations induced similarly potent CTL activity in vivo at 4-7 days after administration, with nearly complete target elimination seen at day 7 ( Fig 1E). However, the level of CTL activity induced by SVP[OVA]-PLA declined by day 10 and continued to decrease through day 17. In contrast, the CTL activity induced by SVP[OVA]-PLGA was sustained, with >90% elimination of target cells through day 17 ( Fig 1E) and >30% activity at day 31 after a single SVP injection (S1 Fig).
SVP[OVA]-PLGA was similarly superior to SVP[OVA]-PLA when used in combination with SVP[R848] for therapeutic treatment of EG.7-OVA tumors (Fig 1F and 1G). This was especially pronounced if surviving animals were re-challenged with the same tumors without any additional treatment ( Fig 1H). In this case, the majority of animals previously treated with SVP  In additional studies, SVP[OVA]-PLGA also demonstrated an ability to induce antibody responses to OVA (S4 Fig, note that from this point on, only PLGA-based SVP formulations were used to encapsulate test antigens). In particular, three immunizations of SVP[OVA] adjuvanted with SVP-encapsulated TLR agonists were capable of maintaining high anti-OVA antibody titers for 1-2 years after the initial immunization (S4A and S4B Fig). At this point, SVPencapsulated CpG oligonucleotides (ODN), potent TLR9 agonists, were tested and compared to SVP[R848]. Natural CpG ODNs contain a phosphodiester (PO) backbone (PO-CpG), which is susceptible to rapid hydrolytic cleavage by nucleases in vivo. Nuclease-resistant CpG sequences with a phosphorothioate (PS) backbone (PS-CpG) have been shown to have superior activity to PO-CpG in vivo and have been utilized in many preclinical and clinical studies [43]. Notably, similar levels of antibody to OVA were produced when SVP[OVA] was combined with either SVP-encapsulated PS-or PO-CpG ODNs, exceeding titers induced by SVP

Therapeutic activity of SVP encapsulated peptide vaccines
We next evaluated CTL induction against E7.I.49, the dominant MHC class I epitope of the HPV-16 E7 oncoprotein. Encapsulating both E7.I.49 and R848 within the SVP led to the dosedependent induction of robust CTL activity, while free (non-encapsulated) E7.I.49 and R848 were inactive (Fig 2A). Next we assessed the efficacy of SVP[E7.I.49] administered with free CpG versus SVP-encapsulated CpG using the mouse-specific CpG sequence 1826. SVP [PS-CpG]-and SVP[PO-CpG]-adjuvanted vaccines induced more rapid onset of CTL activity than the free PS-CpG-adjuvanted vaccine ( Fig 2B). Moreover, while CTL activity induced with the free PS-CpG-adjuvanted vaccine decreased after 7 days, the SVP[PS-CpG]-and SVP [PO-CpG]-adjuvanted vaccines induced sustained CTL activity, with nearly complete elimination of target cells observed through day 21 and 85-95% activity for at least one month. Notably, free nuclease-susceptible PO-CpG was completely inactive (not shown).
We next evaluated the efficacy of SVP[E7.I.49] against TC-1 tumors in vivo. This mouse tumor line was transduced to express the E6 and E7 oncogenic proteins of HPV-16 [44]. SVP [E7.I.49] was more effective in promoting survival in TC-1 bearing animals when adjuvanted with SVP[PS-CpG] than with SVP[R848] (Fig 3A and 3B). This difference became even more pronounced when treatment was initiated at 6 days versus 3 days after tumor inoculation (compare Fig 3B to Fig 3A). Moreover, SVP[PS-CpG] was more effective than free PS-CpG ( Fig 3C). SVP-encapsulated PO-and PS-CpG were equally potent, while free PO-CpG provided no benefit ( Fig 3D).

Therapeutic efficacy of SVP vaccines against HPV-16 oncogenic proteins in mice with established palpable TC-1 tumors
Next we proceeded to construct protein-based HPV antigens capable of inducing broad and long-term immune response to HPV-16 oncogenic proteins. Two genetically-modified antigens, mutated E7 protein (E7 Ã , bearing two point mutations abolishing E7 binding to p105-Rb) and a fusion of mutated E7 and E6 proteins (E7/E6 Ã , additionally bearing mutations abolishing E6 binding to p53) were constructed, encapsulated into SVP and tested for their immunogenicity and therapeutic activity. Both SVP[E7 Ã ] and SVP[E7/E6 Ã ] were immunogenic in vivo, as assessed by their ability to induce antibody responses (S7 Fig) and cytotoxic responses to E7 (Fig 4A and 4B) Fig 3B), resulting in long-term survival rates of 150 days in 60-70% of mice (not shown). Moreover, this difference was even more pronounced if the start of treatment was delayed until day 10 after tumor inoculation, at which point most TC-1-inoculated mice exhibited palpable tumors. In this case, SVP[E7.I.49] combined with SVP[R848] delayed tumor growth but provided minimal long-term survival advantage over the mock-treated control ( Fig 4C). In contrast, treatment with SVP[E7 Ã ] or SVP[E7/E6 Ã ] led to statistically higher lifeexpectancy and survival rates, with 44% and 56%, respectively, of mice surviving at least 500 days ( Fig 4C).  Fig 5A). However, with SVP[R848], antigen-specific cytotoxicity exceeded 90% at day 7 in only 2 out of 8 mice, with an average activity of 68%, and then dropped precipitously by week 2 (Fig 5A). In contrast, utilization of either SVP [CpG] or SVP[poly(I:C)] led to near-complete elimination of target cells at day 7 with sustained high levels of antigen-specific cytotoxicity persisting for several more weeks before falling to 40-60% activity at 4 weeks ( Fig 5A).

Comparison of the therapeutic activity of SVP[E7/E6
Next SVP[E7/E6 Ã ] was tested with the same SVP adjuvant formulations for therapeutic activity in mice with palpable TC-1 tumors, with the start of treatment delayed until 10 days after tumor inoculation. Both SVP[CpG]-and SVP[poly(I:C)]-adjuvanted vaccines led to >80% long-term survival, while SVP[R848] induced 65% long term survival (Fig 5B). When the start of treatment was further delayed until day 13 or 14 after tumor inoculation, all three SVP adjuvant formulations showed similar efficacy through 60 days (Fig 5C-5F). However, the SVP[poly(I:C)]-adjuvanted vaccine maintained a long-term survival rate of almost 60% through day 150, compared to 31% and 27% for SVP[CpG]-and SVP[R848]-adjuvanted vaccines, respectively ( Fig 5C). The activity of the the SVP[poly(I:C)]-adjuvanted was also associated with a significant elevation of tumor-infiltrating lymphocytes (3.8% vs. 0.5% in mocktreated mice).
Remarkably, delayed therapeutic treatment with the SVP vaccines enabled complete regression of tumors up to~2000 mm 3 in size, although generally tumors of >1000 mm 3 were more difficult to control than those in the 500-1000 mm 3 range (Fig 5D-5F). Individual early tumor growth curves aligned well with overall survival data with use of SVP[R848] being slightly inferior to SVP[CpG] and SVP[poly(I:C)] (Fig 5D-5F). We next re-challenged 26 surviving mice that had been treated with SVP[E7/E6 Ã ] and SVP[poly(I:C)] with TC-1 cells in the absence of any additional treatment five months after the initial tumor inoculation (and four months after the last SVP treatment). Notably none of the mice showed tumor growth for an additional five months after tumor re-challenge, at which time they were challenged with TC-1 for a third time. The treated mice remained tumor free after the third inoculation, while all control naive mice inoculated at the time of both re-challenges succumbed to tumor growth within 24-36-days (Fig 6A). Similarly, adoptive transfer of splenocytes from SVP-treated long-term TC-1 survivors led to complete protection of otherwise untreated syngeneic recipients from a challenge with TC-1 (Fig 6B), with seven of eight recipients remaining tumor-free during the 90-day observation period. The eighth animal developed a small (14 mm 3 ) tumor that was subsequently eliminated (not shown). We previously demonstrated that SVP-encapsulated R848 and CpG prevented systemic cytokine production [15]. Here, we similarly tested SVP-encapsulated poly(I:C).

Synergy of immune checkpoint inhibition or chemotherapy with SVP treatment
As demonstrated above, even suboptimal SVP immunotherapeutic treatment regimens incorporating only a single peptide epitope led to a discernible benefit against both TC-1 and B16-F10 tumors (Fig 3 and S6 Fig). These suboptimal regimens were selected to evaluate possible synergy between SVP treatment and immune checkpoint inhibitors, specifically antibodies against CTLA-4, PD-1 and PD-L1. The combination of SVP[Trp2.180-188 peptide] + SVP[PO-CpG] with anti-PD-L1 antibody showed profound synergy in the B16-F10 model (Fig 7A), leading to more than 50% long-term survival. Both anti-PD-1 and anti-CTLA-4 also demonstrated synergistic effects with SVP, but were inferior to anti-PD-L1 (not shown). None of the antibodies tested were effective without SVP (Fig 7A and not  SVP treatment showed even more profound synergy with chemotherapy in the TC-1 model. Cisplatin, a standard-of-care chemotherapeutic agent used in many cervical cancer treatment modalities, was partially effective as monotherapy, with~40% overall long-term survival when administered on day 5 and 12 after tumor inoculation (Fig 7B). Similarly, SVP[E7. I.49] + SVP[CpG] monotherapy showed~50% long term survival when treatment was initiated on d14 after tumor inoculation (Fig 7B, consistent with earlier data; see Fig 5C), but onlỹ 17% survival when treatment was delayed until day 21. In contrast, combining cisplatin and SVP treatment regimens resulted in substantially better survival, with 100% overall survival at 166 days when SVP vaccination was initiated on day 14, and~85% survival when SVP vaccination was delayed until day 21 after tumor inoculation.

SVP immunogenicity in non-human primates
A small pilot study was conducted to translate the findings in mice to nonhuman primates. In particular, since the TLR distribution differs between mice and monkeys, we were interested to compare SVP formulations of R848, PO-CpG and poly(I:C). Twelve cynomolgus monkeys (Macaca fascicularis) were immunized with two different dose levels of SVP-encapsulated HPV-16 E7/E6 Ã fusion protein combined with two different dose levels of SVP[R848], SVP [PO-CpG] or SVP[poly(I:C)] (see Table 1 for details), with each animal receiving a different antigen/adjuvant dose combination (high dose E7/E6 Ã + high dose adjuvant; high dose E7/E6 Ã + low dose adjuvant; low dose E7/E6 Ã + high dose adjuvant; or low dose E7/E6 Ã + low dose adjuvant; Table 1). An additional control animal was immunized with high dose of free E7/E6 Ã and free PS-7909. The vaccination scheme is shown in Fig 8. The various vaccine combinations were generally well tolerated. At one week after the 3 rd injection (day 49), nodules were observed at the immunization sites (thighs) of three animals that received the SVP[poly(I:C)] adjuvant. The animals showed no discomfort and thus no treatment was required.
Animals were assessed for antigen-specific T cell recall by interferon-γ (IFN-γ) ELISPOT, CD8 + T cell expression of granzyme B, IFN-γ, and TNF-α, CTL activity against E6 and E7 and anti-E6 and anti-E7 antibody responses (see summaries for IFN-γ ELISPOT, granzyme B and IFN-γ intracellular staining and humoral response to E6 and E7 in Fig 9, data for individual animals is shown in S10 and S11 Figs; results for each readout in individual animals are summarized in Table 1). Monkeys immunized with SVP[E7/E6 Ã ] adjuvanted with SVP[poly(I:C)] showed the most robust T cell recall responses, which were detectable as early as day 35 and increased over time (Fig 9A and 9B, and S10 and S11  (Table 1). Moreover, SVP  Table 1). These results indicate that the SVP[poly(I:C)] is the superior adjuvant in NHP for the induction of antigen-specific CD8 T cell and B cell responses.   Polymer nanoparticles induce CTLs and suppress tumor growth

Discussion
Nanoparticles have been shown to be effective vehicles for vaccine delivery, as they are capable of trafficking via the draining lymphatics to lymph nodes where they are selectively taken up by antigen presenting cells ( [45], reviewed in [46,47]). Here we describe the development of a biodegradable polymer-based nanoparticle formulation for efficient induction of CTL responses. Our objectives were to optimize therapeutic vaccines by assessing different polymer formulations, TLR ligands, tumor antigens, and synergy with checkpoint inhibitors and chemotherapeutic agents. Finally, we have translated key findings in a small pilot study in nonhuman primates. SVP are comprised of biodegradable polymers. We assessed SVP made with poly(lactic-coglycolide) (PLGA) and poly(lactic acid) (PLA), which are predicted to have different properties with respect to their degradation time in vivo. SVP vaccines comprised of PLA as a core polymer induced potent CTL responses in vivo with kinetics that are typical of most vaccines, namely with CTL activity peaking at day 7 and then rapidly contracting. In contrast, SVP vaccines comprised of PLGA as a core polymer induced durable CTL activity that was sustained for up to several weeks (Figs 1A-1E, 2B and 5A and S1  (Fig 1). This is likely due to the faster degradation rate of PLGA compared to PLA under acidic conditions such as those found in the endosomes of antigen presenting cells. The enhanced processing of antigen in the SVP-PLGA formulation may thus be related to the ability of professional antigen presenting cells to more completely degrade the polymer matrix and encapsulated antigen within SVP-PLGA compared to SVP-PLA. In addition, the PLGA polymer core provides higher antigen loads, likely producing a denser protein corona on the nanoparticle surface, which would be more readily accessible to endosomal protein processing machinery. Collectively, it is likely that PLGA formulations improve cellular immune responses by promoting prolonged processing and presentation of the encapsulated antigen. Polymer nanoparticles induce CTLs and suppress tumor growth TLR ligands have potent adjuvant properties by activating innate antigen presenting cells and promoting the production of Th1 cytokines. One of the concerns of using TLR agonists as adjuvants is the production of systemic inflammatory cytokines. We focused here on TLR3, TLR7/8, and TLR 9 agonists, as these receptors reside within the endosomal compartment. When SVP are endocytosed, they can release their payload directly to adjacent receptors. We have previously shown that multiple immunizations with SVP[R848] and SVP[CpG] formulations initiate local, but not systemic, cytokine induction [15] and induce minimal inflammation at the site of vaccine administration compared to free TLR agonists. We extend these findings here to SVP[poly(I:C)], a TLR 3 agonist (S8 Fig). All three formulations were effective adjuvants for promoting potent CTL responses in mice (see below for discussion of relative activity in nonhuman primates), although SVP[CpG] and SVP[poly(I:C)] induced more sustained effector CD8 T cell activity (Fig 5A). Importantly the SVP-formulated adjuvants supported potent therapeutic vaccine efficacy in tumor models. Even mice with palpable tumors up to 2000 mm 3 in size showed substantial tumor regression and long term survival with therapeutic SVP vaccination (Fig 5D-5F).
Both peptide antigens and protein antigens can be encapsulated in SVP vaccines. We have previously demonstrated that a MHC class II-binding peptides can be encapsulated to provide T cell help to B cells [48]. Here we focused on the ability of SVP vaccines encapsulating MHC class I-binding peptides to induce CD8 T cell responses. An SVP vaccine encapsulating the Trp2.180-188 peptide adjuvanted with SVP[PO-CpG] or SVP[R848] provided therapeutic protection in the B16-F10 melanoma model (Fig 6A and S6C Fig). Similarly, SVP[E7.I.49] encapsulating the dominant MHC class I epitope of the HPV E7 antigen, provided significant benefit when dosed therapeutically in the TC-1 tumor model. A major concern of using peptide vaccines is the incomplete coverage of potential epitopes and the MHC heterogeneity in the human population. However, protein vaccines are often limited by inefficient cross-presentation of endocytosed antigens. SVP vaccines containing E7 Ã protein or E7/E6 Ã fusion protein induced potent CD8 T responses to both major MHC class I epitopes as well as subdominant epitopes (Fig 4A and 4B). In addition, SVP[E7 Ã ] and SVP[E7/E6 Ã ] vaccines provided superior therapeutic activity compared to the SVP[E7.I.49] peptide vaccine (Fig 4C). In particular, SVP[E7/E6 Ã ] adjuvanted with SVP[poly(I:C)] provided~80% survival even when treatment was initiated 10 days after tumor inoculation and more than 50% survival when treatment was delayed until day 14, at which time mice had palpable tumors of approximately 500 mm 3 (Fig 5B, 5C and 5F). It is noteworthy that the SVP injections were administered into tumor-distal sites (hind limbs), not peritumorally. Our results compare favorably to several other experimental HPV therapeutic vaccine candidates tested in the TC-1 tumor model, which typically used more aggressive treatment regimens [49][50][51][52][53][54].
Tumors can create an immunosuppressive microenvironment, and render the tumor refractory to treatment [9,10]. The recent approval of immune checkpoint inhibitors have researchers pursuing combination therapy strategies. We show here that SVP vaccines are synergistic with checkpoint inhibitors, such as anti-PD-L1 antibodies (Fig 6A and S9B Fig). A suboptimal single-peptide SVP vaccine showed significant synergy with anti-PD-L1 antibodies, leading to long term survival of greater than 50% mice with durable immune memory in the aggressive B16-F10 melanoma tumor model. Even more striking therapeutic efficacy was observed when the SVP therapy was combined with cisplatin, a chemotherapeutic agent that is often used to treat cervical cancers. Cisplatin alone was only partially effective in the TC-1 model, in agreement with other studies [55,56]. However, cisplatin was combined with SVP therapy provided 100% long term survival even with SVP treatment was delayed until 13/14 days after tumor inoculation and >80% survival when SVP treatment was delayed until 21 days (Fig 6B). Re-inoculation of SVP[E7/E6 Ã ]-immunized survivors with TC-1 tumor cells led to complete tumor rejection reflecting durable immune memory to E7 antigen (Fig 5G and  5H).
One of the hurdles to cancer vaccine development is that data in mice have not translated to humans [9]. We conducted a small pilot study to assess the immunogenicity of SVP[E7/ E6 Ã ] in non-human primates. In particular, we focused on evaluating different TLR agonists, as TLR expression differs between mice and humans [57] and the subset of dendritic cells implicated in cross presentation is also different [57]. We evaluated three SVP-encapsulated TLR agonists, R848, PO-CpG and poly(I:C). While antigen-specific T cell activity and antibody production was seen in animals from all experimental arms, formulations adjuvanted with SVP[poly(I:C)], a TLR3 agonist, generated the most robust responses. Notably, CD8 T cells from all four animals treated with SVP[E7/E6 Ã ] adjuvanted with SVP[poly(I:C)] were positive for granzyme B and interferon-γ expression and three of the four animals demonstrated HPV E6/E7-specific CTL activity (Table 1). These results indicate that the SVP[poly(I:C)] is the superior adjuvant in NHP for the induction of antigen-specific CD8 T cells. These results are consistent with TLR3, but not TLR 7, 8, or 9, being highly expressed on the subpopulation of human dendritic cells capable of efficient antigen cross-presentation [57].
Our data indicate the therapeutic potential of an SVP-based HPV vaccine. Notably, SVP technology shows the potential to augment the immunogenicity of the target antigen via its particulate delivery and sustained in vivo release, resulting in robust and sustained production of effector CD8 T cell activity with durable immunological memory. In addition SVP adjuvants can deliver potent TLR agonists, such as poly(I:C) to their cognate receptor in the endosome, thereby reducing the potential for systemic side-effects. Moreover, GMP manufacturing of SVP nanoparticles has been also shown to be scalable and reproducible [58]. Therefore a therapeutic SVP platform-based vaccine against HPV-induced malignancies is an attractive candidate for further clinical evaluation.  Table 1 (high dose-1, low dose-2), data grouped in columns per adjuvant used (indicated on top of each set); peptide pools used for PBMC stimulation are shown. Y-axis scale for R848 is 8 times smaller than for CpG and poly(I:C) with the latter two being equal (with the exception of low doses of both E7/E6 Ã and an adjuvant). (DOCX) S11 Fig. Granzyme