Celastrol Nanoemulsion Induces Immunogenicity and Downregulates PD-L1 to Boost Abscopal Effect in Melanoma Therapy

Abstract

Programmed cell death-ligand 1 (PD-L1)-based immune checkpoint blockade therapy using the anti-PD-L1 antibody is effective for a subset of patients with advanced metastatic melanoma but about half of the patients do not respond to the therapy because of the tumor immunosuppressive microenvironment. Immunogenic cell death (ICD) induced by cytotoxins such as doxorubicin (DOX) allows damaged dying tumor cells to release immunostimulatory danger signals to activate dendritic cells (DCs) and T-cells; however, DOX also makes tumor cells upregulate PD-L1 expression and thus deactivate T-cells via the PD-1/PD-L1 pathway. Herein, we show that celastrol (CEL) induced not only strong ICD but also downregulation of PD-L1 expression of tumor cells. Thus, CEL was able to simultaneously active DCs and T-cells and interrupt the PD-1/PD-L1 pathway between T-cells and tumor cells. In a bilateral tumor model, intratumorally (i.t.) injected celastrol nanoemulsion retaining a high tumor CEL concentration activated the immune system efficiently, which inhibited both the treated tumor and the distant untreated tumor in the mice (i.e., abscopal effect). Thus, this work demonstrates a new and much cost-effective immunotherapy strategy — chemotherapy-induced immunotherapy against melanoma without the need for expensive immune-checkpoint inhibitors.

1. Introduction

Advanced melanoma at the metastatic stage is very aggressive and has a poor prognosis [1]. Recently, metastatic melanoma treatments have been evolving rapidly, leading to the most notable immune checkpoint blockade therapies targeting cytotoxic T-lymphocyte antigen 4 (CTLA-4) and programmed cell death 1 (PD-1) [2, 3]. After FDA approval in 2014, PD-1/PD-L1 inhibition became the standard-of-care first-line therapy for metastatic melanoma [4]. While approximately 40% of patients achieve a partial or complete response, approximately 40 to 45% of patients do not respond to the therapy due to the tumor’s inadequate T-cell infiltration and immunosuppressive microenvironment. Many strategies have been explored to overcome these obstacles via stimulating tumor antigen-specific T cells and improving T-cell trafficking into the tumor [2–4].

The success of the checkpoint blockade immunotherapy for melanoma heavily depends on the recruitment of effector CD8⁺ T cells to its microenvironment. Thus, treatments enabling T-cell infiltration into tumors have been extensively explored. One such strategy is to induce immunogenic cell death (ICD), during which damaged dying tumor cells release immunostimulatory danger signals, including high mobility group box 1 (HMGB1), calreticulin (CRT) and ATP that activate DCs and T-cells and thus abrogate immune tolerance [5–8]. A clinic example approved for the treatment of melanoma is talimogene laherparepvec (T-VEC), whose therapeutic activity is thought to result from the induction of ICD in T-VEC-infected melanoma cells [9]. Cytotoxic stress-inducing treatments (e.g., radiotherapy [10–12]) and some chemotherapeutic drugs (doxorubicin, mitoxantrone, paclitaxel, oxaliplatin, and cyclophosphamide) [13–15] and molecular therapy drug crizotinib [16, 17] are effective ICD-inducers that significantly sensitize tumors to PD-1/PD-L1 blockade immunotherapy. However, the ICD-inducers generally upregulate tumor cells’ PD-L1 expression and must combine with PD-1/PD-L1 blockade to gain therapeutic efficacy via either adding [16, 18, 19] or in-situ expression of the antibodies [20].

We hypothesized that an ICD-inducer that simultaneously downregulates the PD-L1 expression in tumor cells would eliminate the need for the expensive anti-PD-1/L1 antibodies. Celastrol is recognized for its anti-inflammatory effects via suppressing macrophage M1 polarization and as a cytotoxin inhibiting proliferation of tumor cells [21–23]. Herein, we show that celastrol not only induced strong ICD in tumor cells but also, surprisingly, reduced their PD-L1 expression. In vivo, the i.t. injected its nanoemulsion (CEL NE) retained a persistent concentration of celastrol in melanoma induced strong ICD and downregulation of PD-L1 (Scheme 1), and thus without anti-PD-L1 antibody effectively inhibited the growth of both treated tumors and distant untreated tumors in a bilateral melanoma model.

Scheme 1. |. Schematic illustration of the intratumorally injected celastrol nanoemulsion (CEL NE) simultaneously inducing immunogenic cell death (ICD) and PD-L1 downregulation, boosting the systemic abscopal effect on B16F10 bilateral tumor model.

CEL NE i.t. injected in the subcutaneous tumor on one side continuously released CEL and induced tumor cells to expose calreticulin (CRT) and release HMGB1 as the tumor-associated antigens, which were engulfed by antigen-presenting cells (DC cells) and primed CD8⁺ T cells infiltration and activation. Meanwhile, CEL NE also effectively downregulated PD-L1 expression in tumor cells. The synergy of strong ICD and PD-L1 reduction activated the tumor immunosuppressive microenvironment and effector CD8⁺ T cells, giving potent tumor inhibition of both primary tumor and distant contralateral tumor as well as long-lasting systemic tumor suppression.

2. Results and discussion

2.1. Systemic PD-L1 blockade was inefficient in B16F10 melanoma because of the immune-suppressive microenvironment and lack of T cell infiltration

It is known that approximately 40 to 45% of clinical melanoma do not respond to anti-PD-1/PD-L1 therapy [2–4]. Indeed, the anti-PD-L1 mAb (αPD-L1) treatments did not significantly slow down the B16F10 melanoma growth even though the PD-1⁺ population was about 17% or 27% of the T cells infiltrating in subcutaneous melanoma tumor, B16F10 or BPD6 (Supporting Information Figures S1A, S1B). Detailed analysis showed that intraperitoneally (i.p.) injected anti-PD-L1 mAb blocked PD-1/PD-L1 interaction, the cytotoxic CD8⁺ T cells and NK cells in the tumor were still very limited, and the immune-suppressive cell populations (e.g., MDSCs and Tregs) were unchanged (Supporting Information Figure S1C). So, the immunosuppressive microenvironment curbed the anti-PD-L1 mAb therapy against melanoma.

2.2. CEL triggered autophagy and thus induced ICD in both human and murine melanoma cell lines

We previously showed that CEL (Figure 1A) worked in synergy with mitoxantrone to induce ICD [24]. Our further study showed that CEL induced autophagy of melanoma cells (Figure 1B), as reported in other cells [25, 26]. Because autophagy is one type of stress that precedes ICD, we therefore hypothesized that CEL alone might serve as an effective ICD inducer.

Figure 1. Celastrol (CEL) induces ICD and down-regulates PD-L1 expression in melanoma in vitro and in vivo.

(A) Chemical structure of CEL. (B) CEL triggered autophagy in melanoma cells. Mouse B16F10 or BPD6 melanoma cells were incubated with CEL (0.1~8 μM) for 12 h and then lysed for Western blot analysis of LC3B, whose subunit transition from LC3B I to LC3B II is a marker for activated autophagy. (C) Immunofluorescent imaging of CRT, HMGB1, and PD-L1 in mouse (B16F10 and BPD6) and human (M10 and A375) melanoma cells after treated with CEL at 1 μM; The treatment time was 4 h for CRT and 24 h for HMGB1 and PD-L1 detection. Cell nuclei were stained with DAPI. Scale bar indicates 20 μm. Each value was quantified in 5 randomly selected fields. Each sample was repeated 3 times. (D) Flow cytometry analysis of CRT⁺ melanoma cells after treated with CEL or positive control, doxorubicin (DOX) and mitoxantrone (MIT), for 4 h at their IC₅₀. (E) The released HMGB1 in the cell culture medium 24 h after incubation with CEL, DOX or MIT at their IC₅₀ doses (n = 4). (F) RT-PCR analysis of the PD-L1 mRNA levels in B16F10 and BPD6 cells after incubation with CEL, DOX or MIT at IC₅₀ for 24 h (n = 6). (G) Western blot analysis of PD-L1 expression in tumors at 48 h after i.t. injection with CEL (0.15 mg/kg), DOX (0.01 mg/kg) or MIT (0.5 mg/kg) or i.p. injection with αPD-L1 (5 mg/kg). All data are shown as mean ± SD. *p < 0.05, **p < 0. 01, ***p < 0. 001, NS: not significant.

The ability of CEL inducing ICD in melanoma cell lines was first tested by imaging two typical ICD markers, CRT and HMGB1, in human melanoma (M10 and A375) and mouse melanoma (B16F10 and BPD6) cell lines at 1 μM, which was well below the dose inducing 50% cell death (IC₅₀, Supporting Information Figure S2A). CEL induced all the 4 melanoma cell lines undergoing intense CRT exposure and cytosolic HMGB1 release at the concentration of 1 μM (Figure 1C). The higher resolution images confirmed that CRT was on the surface (Supporting Information Figure S2B). Furthermore, western blot analysis of the B16F10 cell lysates showed that CEL treatment did not change the overal HMGB1 expression but just induced HMGB1 to translocate from nucleus to cytosol (Supporting Information Figure S2C). The ICD-inducing capacity of CEL was compared with known ICD inducers, doxorubicin (DOX) and mitoxantrone (MIT), at their IC₅₀ doses. CEL induced the cells to have more than 3–10 times more CRT⁺ population (Figure 1D, Supporting Information Figure S2D) and release more HMGB1 in supernatants than DOX and MIT (Figure 1E).

DC recruiting into the tumor bed (stimulated by the released ATP) followed by engulfing tumor antigens (stimulated by CRT) and presenting the antigens to T cells (stimulated by HMGB1) is the critical process in ICD remodeling tumor immunosuppressive microenvironment [9]. We thus tested whether CEL-induced ICD could activate and mature DCs. As shown in Supporting Information Figure S2E, the supernatants of the 4 melanoma cells pulsed with CEL could activate DC2.4 cells, as seen from their morphological changes. The CEL-treated B16F10/BPD6 cells’ condition medium could activate DC2.4 cells even more efficient than that of the positive control, lipopolysaccharides (LPS) at its commonly used dose. This DC activation effect was not induced by the tumor-culture conditioned medium because the medium alone did not change cell phenotype.

The in vivo ICD induction by CEL was confirmed by an immune activation test using a “vaccination method” [5]. B16F10 cells were first treated with CEL or MIT for 24 h to induce ICD and then subcutaneously (s.c.) inoculated to immunocompetent mice as a tumor vaccine. The innoculated CEL-treated cells could not form tumors. After 7 days, the mice were re-challenged with un-treated B16F10 cells on the contralateral side (Supporting Information Figure S2F). The mice that pre-vaccinated with CEL-treated cells significantly slowed contralateral-side tumor formation and progression, while MIT-treated cells did not have such an ability. The M1/M2 macrophage ratio also increased in the B16F10 tumor treated with CEL (Supporting Information Figure S2G), as reported [27].

Surprisingly, CEL also efficiently downregulated the expression of constitutive (Figure 1C) and inducible (Supporting Information Figure S2H) PD-L1 of human and mouse melanoma cells. RT-PCR analysis showed that the PD-L1 downregulation in both B16F10 and BPD6 cells by CEL resulted from the reduced PD-L1 mRNA levels (Figure 1F). Western blot analysis of B16F10 and BPD6 tumors treated with CEL, DOX or MIT confirmed that CEL treatment reduced the PD-L1 expression to 1/2 or 1/3 of the PBS control, while MIT treatment increased the PD-L1 expression. DOX treatment increased PD-L1 mRNA expression but did not affect protein expression in B16F10 tumor (Figure 1G). The effect of CEL on PD-L1 expression in DC2.4 cell lines (mouse dendritic cell lines) was analyzed. CEL decreased their PD-L1 expression, especially in LPS-activated DCs, but much less extensive than in melanoma cells (Supporting Information Figure S2H).

2.3. Celastrol nanoemulsion (CEL NE) formulation and its immune modulation

CEL is hydrophobic and generally cleared once administered. So, we loaded CEL into a nanoemulsion using an ultrasonic emulsification method [28] with FDA approved excipients (Supporting Information Figure S3A). It consisted of an internal sesame oil/celastrol core surrounded by a surfactant layer of soybean lecithin and Pluronic F68. The encapsulation efficiency of CEL in the nanoemulsion was 90 ± 2% (Supporting Information Figure S3B). As shown in Figure 2A, the average size of the CEL NE was 91 nm in diameter with a polydispersity index of 0.153 and a nearly neutral surface charge. The transmission electron microscopic image shows CEL NE having a uniform and spherical shape. The nanoemulsion was very stable at room temperature and showed a pH-insensitive release of CEL (Supporting Information Figures S3C, S3D).

Figure 2. Characterization of celastrol nanoemulsion (CEL NE) and its ICD inducing and PD-L1 downregulation on melanoma.

(A) The size distribution of CEL NE characterized by dynamic laser light scattering and transmission electron microscopy (inserts). (B) The dose-dependent curve of CRT⁺-population of B16F10 cells incubated with free CEL or CEL NE for 4 h. The percentage of CRT⁺ cells was quantified as an average of 5 randomly selected fields. Each sample was repeated 3 times; see Supplementary Figure S4 for the curves of other cells. (C-F) Analysis of in vivo ICD-induction and PD-L1 downregulation. The mice bearing B16F10 tumors received a single i.t. injection of CEL NE or free CEL (0.15 mg/kg) (n = 4) and were sacrificed after 48 h for analysis: (C) Representative immunofluorescent images of CRT, HMGB1 and activated DC markers (CD86, CD11c, and MHC II) in B16F10 tumors after i.t. injection of CEL NE or free CEL. Each value was quantified as an average of 5 randomly selected fields. Scale bar indicates 300 μm. (D) Statistic analysis of co-stimulatory markers, CD80 and CD86, in the draining lymph nodes by flow cytometry (n = 4). (E) Flow cytometry quantitation of PD-L1⁺ melanoma cell percentages (PD-L1⁺MART⁺) in the tumors. (F) RT-PCR analysis of PD-L1 mRNA levels in the tumors. (G) Western blot analysis of dose-dependent downregulation of NF-κB and PD-L1 in the tumors. The mice with tumors as above were i.t. injected with the indicated doses of CEL NE and analyzed after 48 h (n = 4). All data are shown as mean ± SD. *p < 0.05, **p < 0. 01, ***p < 0. 001, NS: not significant.

The cytotoxicity and dose-dependent ICD induction of free CEL and CEL NE were conducted on mouse and human melanoma cell lines (Figure 2B, Supporting Information Figures S3E, S4). CEL NE showed similar cytotoxicity to free CEL with IC₅₀ values in the range of 2.5~4.5 μM. The empty nanoemulsion showed no toxicity even at the highest doses (Supporting Information Figure S3E). ICD induction ability of CEL NE and free CEL was evaluated by quantitation of CRT⁺ ratios using Image J software. The dose-dependent percentage of the CRT⁺ population was plotted (Supporting Information Figure S4). Unexpectedly, CEL NE induced ICD at much lower doses in the 4 melanoma cell lines. For instance, 100 nM CEL NE induced 90% B16F10 cells, 64% BPD6 cells, 92% A375 cells and 89% M10 cells to expose CRT on cell surfaces, while free CEL only induced 20%, 18%, 21% and 40% cells CRT- exposure, respectively. The dose inducing 20% cells CRT⁺ was herein denoted as the minimal ICD inducing dose (ICD_20%) and was obtained from the dose-dependent plot of the CRT⁺ population (Supporting Information Figure S4). The ICD_20% values of CEL NE were significantly lower, only about 1/10 for B16F10, BPD6 and A375 cells, and a half for M10 cells of the values of free CEL. Furthermore, the human melanoma cell line seemed much more sensitive to CEL to expose CRT than mouse melanoma — The ICD_20% of CEL was about 25 nM for M10 cells, 50 nM for A375 cells but about 100 nM for B16F10 and BPD6 cells. Free CEL is poorly water-soluble so its available concentration was low. CEL NE significantly increased its solubility and cellular uptake, leading to more efficient ICD than free CEL in vitro.

The in vivo ICD induction by CEL NE was tested by one single i.t. injection into B16F10 tumors at a CEL-equivalent dose of 0.15 mg/kg, far below its chemotherapy dose (2 ~5 mg/kg), using PBS and free CEL as controls. The immunofluorescence images of the sectioned tumors are shown in Figure 2C. Consistent with the in vitro findings, CEL NE-treated tumor had much stronger CRT and HMGB1 immunofluorescence and higher positive-cell ratios as calculated by Image J than PBS and free CEL. The activated DC markers—including CD86, MHC Class II, and CD11c also increased. Flow cytometry analysis of the draining lymph nodes further confirmed the upregulation of co-stimulatory molecules CD80⁺ and CD86⁺ (Figure 2D). Furthermore, both PD-L1⁺ melanoma cells and the total PD-L1 mRNA level decreased (Figure 2E, 2F), especially in CEL NE treated tumors. The dose-dependent coincidence of PD-L1 and NF-κB downregulation (Figure 2G) indicates that the CEL NE induced-PD-L1 downregulation probably resulted from the NF-κB inactivation because PD-L1 expression in melanoma cells was reported to be dependent on NF-κB activation [29].

2.4. CEL biodistribution and pharmacokinetics in B16F10 bilateral tumor model after single-site i.t. injection of free CEL or CEL NE

The systemic immunotherapy of CEL NE -induced ICD was evaluated by observing the abscopal effect, i.e., the therapeutic effect on the untreated tumor when the treatment was locally administered to a distant tumor, using a B16F10 bilateral tumor model via inoculating cells at the mouse right and left armpits. We used an i.t. injection of free CEL or CEL NE at the dose of 0.15 mg/kg, which did not induce apparent toxic effect even upon intravenous administration.

The biodistribution of the CEL NE was first investigated by tracking the DiD-loaded CEL NE after i.t. injection to the left site tumor. Both treated/un-treated tumors and main organs (heart, liver, spleen, lung, and kidney) were dissected at 4 h, 12 h, or 24 h post-treatment and imaged using IVIS based on the fluorescence intensity of DiD. As shown in Figure 3A, DiD-loaded CEL NE was mostly retained in the injected tumor even after 24 h, and the fluorescence intensity in the untreated tumors and organs was negligible except for the lung at 24 h post-administration.

Figure 3. CEL biodistribution in B16F10 bilateral tumor model after single-site i.t. injection of free CEL or CEL NE.

(A) IVIS imaging and normalized fluorescent intensity of DiD distribution in tumors and main organs. Mice (n = 3) were injected with DiD (0.75 mg/kg) – loaded CEL NE (CEL: 0.15 mg/kg) to the left side tumor. At 4 h, 12 h or 24 h post injection, the mice were sacrificed and imaged using an IVIS system with excitation at 640 nm and emission at 670 nm. **p < 0.01, ***p < 0.001. (B, C) HPLC analysis of CEL concentrations in plasma (B) and the treated and un-treated tumors (C) after single i.t. injection with free CEL or CEL NE (0.15 mg/kg, n = 4). (D) Anti-cancer effects and survival curves of i.t. injection with CEL NE (0.15 mg/kg) and free CEL (0.15 mg/kg), n = 4. For individual comparisons, the two-tailed, unpaired Student’s t-test was performed. Significant differences of survivals were assessed using log rank test. ***p < .0001, **p < 0.01, *p < 0.05, NS: not significant.

The CEL concentrations in plasma and treated/un-treated tumors after a single i.t. injection of free CEL or CEL NE were further quantified using HPLC analysis (Figure 3B, 3C), i.t.-injected free CEL gave much higher serum CEL concentrations, and thus a larger area-under-curve value than CEL NE, indicating that i.t.-injected free CEL was leaked more quickly into the bloodstream, while CEL NE retained the drug in the injected tumor and gave very low serum CEL concentration at all the times. Accordingly, the CEL concentrations in the free CEL-treated tumor were lower than that of CEL NE-treated tumors at all times, while the untreated tumors in the CEL NE group had almost negligible amounts of CEL (Figure 3C). The CEL concentration in the CEL NE-treated tumor was found higher than the minimal ICD-inducing dose (ICD_20%), but the concentration in the free CEL-treated tumor was much lower than the ICD_20%. Notably, the CEL concentration in the CEL NE un-treated tumor was less than 1/10 of that of the treated one, which was too low to induce ICD and cytotoxicity. The retained high drug concentration in the treated tumor after the i.t. injection of CEL NE efficiently inhibited tumor growth of B16F10 melanoma and prolonged mouse survival, compared with blank NE and free CEL-treatments (Figure 3D).

2.5. Low dose CEL NE remodeled tumor immune microenvironment, leading to efficient tumor inhibition and systemic abscopal effect independent of αPD-L1

The therapeutic efficacy of CEL NE was compared with anti-PD-L1 mAb (αPD-L1) blockade on the B16F10 bilateral tumor model in immunocompetent C57BL/6 mice with the treatment schedule shown in Figure 4A. The treatment was initiated by i.p. injection of 5 mg/kg αPD-L1, i.t. injection of CEL NE at 0.15 mg/kg CEL into the left side tumor or their combination (the combo group) when the tumor volumes reached around 70 mm³.

Figure 4. Therapeutic results of αPD-L1, CEL NE, and their combo in the B16F10 bilateral subcutaneous tumor model.

(A) Tumor inoculation and treatment schedule. B16F10 cells (1 ×10⁶) were subcutaneously injected into the left and right armpits of C57/BL6 mice. On day 5, the treatments were initiated by i.t. injection of CEL NE (0.15 mg/kg) into the left side tumor on an every-3-day schedule for 4 times. For the mouse αPD-L1 therapy, on day 8, αPD-L1 (5 mg/kg) was i.p. injected on an every-3-day schedule for 3 times. The combo therapy combined CEL NE and αPD-L1 treatments. (B) B16F10 tumor growth curves of the treated and untreated tumors after different treatments. n = 5. (C) Representative tumors images of each group on day 20. (T: i.t-treated tumors; UT: un-treated tumors). (D) The survival curves of B1F10 tumor-bearing mice after treatments as in Figure 4A. The added controls are blank NE and free CEL treated mice with the equivalent doses of NE and CEL; n = 5. (E) Western blot analysis of the treated and untreated tumor lysates in C. (F) Flow cytometry analysis and quantitation of PD-L1⁺ tumor cell populations in the treated and untreated tumors (n = 4). The PD-L1⁺ cells were from the total tumor. (G) Tumor apoptosis analysis by TUNEL assay of the tumors in C. The percentages of apoptotic cells were quantified using Image J as the average of 5 the randomly selected fields. Scale bar represents 300 μm. All data are shown as mean ± SD. For individual comparisons, the two-tailed, unpaired Student’s t-test was performed. Significant differences of mouse survival were assessed using log rank test. ***p < 0.001, **p < 0.01, *p < 0.05, NS: not significant.

As discussed above (Supporting Information Figure S1), αPD-L1 alone could not slow down B16F10 tumor growth (Figure 4B). Compared with PBS and αPD-L1 treatments, CEL NE alone markedly inhibited the growth of both treated and untreated tumors, though slightly less efficiently on the untreated tumor. The combo group showed that adding αPD-L1 did not significantly enhance the therapeutic efficacy of CEL NE during the treatment period (before day 14). After the treatment withdrawal, the combo treatment seemed to prolong the therapeutic effect slightly, particularly in the untreated tumors. Thus, at the endpoint of the experiment, the tumors in the combo group were smaller than those of CEL NE (p < 0.05 in treated tumors; p < 0.01 in un-treated tumors) (Figure 4C). The survival of the mice after the treatments was further followed up (Figure 4D). αPD-L1 alone marginally prolonged the survival; free CEL treatments only slightly extended the overall survival; all the mice died out before Day 25. However, CEL NE significantly prolonged the Median Survival Time (MST) to almost 2 times of those of the PBS and blank NE groups; even on Day 40, 60% of the mice still survived. Combining αPD-L1 initially slowed the death rate but did not change the survival later. Thus, CEL NE did show the therapeutic advantages over free CEL and αPD-L1.

It is rather remarkable that the CEL NE or CEL NE/αPD-L1 combo treatment effectively inhibited the growth of both treated and untreated tumors (Figure 4B–D). As the CEL concentration in the untreated tumors was much below the minimal dose required to induce ICD or cytotoxicity (Figures 3C), it must be the boosted systemic immune response by the CEL NE-induced ICD and DC activation in the treated tumor that inhibited the untreated tumor growth, so-called the abscopal effect. Therefore, the tumors at the endpoint were further analyzed. The CEL NE treatments downregulated AMPK and NF-κB, the melanoma progression-related markers [30, 31], and PD-L1 in both treated and untreated tumors (Fig. 4E). αPD-L1 alone had weak effects on the AMPK and NF-κB expression; the weak PD-L1 bands were due to its blockage to immunofluorescent antibody rather than PD-L1 downregulation. Thus, the combo had similar effects on AMPK, NF-κB, and PD-L1 except for that αPD-L1 enhanced AMPK expression in the untreated tumor. Flow cytometry analysis confirmed the significant reduction of the PD-L1⁺ population in the tumors, from ~17% in the PBS group to only 5% in both CEL NE-treated and untreated tumors (Figure 4F). CEL NE also downregulated PD-L1 expression in activated DC cells (CD11C⁺MHCII⁺ DCs) after intratumoral injection, but much less extensive than in the tumor cells (Supporting Information Figure S5A). Further more, CEL NE reduced the PD-1⁺ population in the intratumoral T cells (Supporting Information Figure S5B). As CEL concentration was very low in the untreated tumors (Figure 3C), the PD-L1 downregulation in untreated tumors must not come directly from CEL NE. CEL NE-induced PD-L1 downregulation in the untreated tumors may be through the NF-κB inactivation (Figures 2G, 4E). H&E staining showed a large area crack and acellular scar structure in both CEL NE- and combo treated tumors (Supporting Information Figure S6). The TUNEL assay revealed that CEL NE/αPD-L1 treatment exhibited the most effective killing effects, inducing nearly 90% of total cell apoptosis in both treated and un-treated tumors with apoptotic cells nearly 15-fold or 10-fold more of the control or αPD-L1-treated tumors, respectively. CEL NE alone also induced about 70% cell apoptotic in the treated tumors and 35% cells apoptotic in the untreated tumors (Figure 4G).

The modulation of the tumor immune microenvironment was further analyzed (Figure 5 and Supporting Information Figure S7–S11). Flow cytometry analysis showed that the CD3⁺ cell population was only about 1% in the control tumors but increased to 13%, 26%, or 30% after treated with αPD-L1, CEL NE or the combo (Figure 5A and Supporting Information Figure S7). Surprisingly, the CD3⁺ cell populations in the untreated tumors were at the same levels as their corresponding treated ones. The proportion of natural killer cells (NK cells), another type of cytotoxic lymphocytes crucial to the innate immune system, increased by 4- to 10-fold in the treated and untreated tumors with the CEL NE or combo, but unchanged upon treatment with αPD-L1 alone.

Figure 5. Tumor immune microenvironment analysis in the B16F10 bilateral model after the treatments shown in Figure 4.

(A) Proportions of immune cells (CD3⁺ T cells, NK1.1⁺ cells, CD3⁺CD8⁺ T cells, activated DCs, MDSCs and Tregs) in the treated and untreated tumors (Figure 4C) analyzed by flow cytometry (n = 4). (B) Confocal microscopy images and quantitation of tumor-infiltrating CD8⁺ T cells and activated DCs (CD11c⁺MHCII⁺) in tumor sections after each treatment. The number shown in white is the average of each cell type in the tumors (n = 4). (C) Relative mRNA levels relative to the PBS control of various cytokines in treated and untreated tumors detected by quantitative RT-PCR (n = 4). All data are shown as mean ± SD. *p < 0.05, **p < 0. 01, ***p < 0. 001, NS: not significant.

It is known that tumor cells undergoing ICD release tumor-associated antigens, which are engulfed by antigen-presenting cells (APCs) like DCs and presented to cytotoxic CD8⁺ T lymphocytes [8, 10]. The CD8⁺ T cells recognize and attack primary and metastatic tumors. As shown in Figures 5A, 5B and Supporting Information Figure S8, αPD-L1 treatment increased the CD3⁺ T cell population but did not recruit CD8⁺ T cells into the tumor, explaining why αPD-L1 alone could not inhibit tumor growth. However, CEL NE treatment increased about 10-fold more CD8⁺ T cells in both treated and untreated tumors, when compared with the PBS group. Notably, the addition of αPD-L1 to CEL NE even decreased the CD8⁺ T cell population in the treated tumor.

Furthermore, CEL NE or the combo treatments increased MHCII⁺ and CD11c⁺ (activated DC markers) cell populations in the tumors by 5- to 10-fold relative to the PBS control, while αPD-L1 treatment alone caused no noticeable DC activation. Therefore, inefficient DC activation is another reason for anti-PD-L1 mAb therapy failure (Figures 5A, 5B and Supporting Information Figure S8).

We also analyzed immune-suppressive cell populations. Myeloid-derived suppressor cells (MDSCs, identified as CD11b⁺Gr-1⁺ in mice) inhibit T cell proliferation and activation and are associated with poor cancer prognosis and therapeutic resistance. All three treatments significantly decreased the MDSC population in both treated and untreated tumors to about 50% by αPD-L1, 30% by CEL NE, and 12% by the combo compared to the PBS control (Figures 5A, Supporting Information Figure S9). The regulatory T cells (Tregs, identified as CD4⁺FoxP3⁺) developing from CD4⁺ cells maintain immune tolerance to tumor antigens and suppress effector T cells proliferation, contributing to tumor’s immunosuppressive microenvironment. CEL NE- or the combo decreased the CD4⁺FoxP3⁺ cell population to 50% or 40% relative to the PBS control, while αPD-L1 treatment did not change the population, in both treated and untreated tumors (Figures 5A, Supporting Information Figure S9).

Memory T cells retain specific memory of their encountered antigens. We evaluated memory CD8⁺ T-cell (identified as CD8⁺CD44⁺) and memory CD4⁺ T-cell (identified as CD4⁺CD62L⁺) populations in both tumors and spleens after the treatment. As shown in Supplementary Figure S10, the αPD-L1 treatment alone did not change the memory CD8⁺ T-cell population, but CEL NE and the combo treatments enriched the cell population by 10-fold in the treated and untreated tumors. Memory CD4⁺ T-cell population increased by 4- to 10-fold in both treated and un-treated tumors after the three treatments. Similar trends were found in the memory T cell population in the spleens (Supporting Information Figure S11). The cell populations with positive co-stimulatory molecules CD80 and CD86 also increased about 20-fold in draining lymph nodes after CEL NE or the combo treatments. Thus, CEL NE therapy, independent of αPD-L1, could evoke both immune response and long-lasting immune memory effect against cancer.

Thus, the activation of DC cells, increased infiltration of effector T cells and NK cells, decreased MDSCs and Tregs induced by CEL NE treatment all contributed to its potent systemic-immune therapy against both treated and untreated tumors.

The immune response-related cytokines induced by the treatments were further analyzed (Figure 5C). CEL NE treatments increased T helper 1 (Th1) cytokines TNF-α by 3-fold and IFN-γ by 2-fold in the treated tumors. αPD-L1 alone had no such an effect while adding αPD-L1 to CEL NE weakened CEL NE’s ability to induce TNF-α expression. The increased expression levels of TNF-α and IFN-γ promoted DC maturation and antigen presentation. IL12 is a T cell-stimulating factor [32] involved in the differentiation of naive T cells into Th1 cells, stimulating TNF-α and IFN-γ production from T cells and NK cells and thus enhancing the cytotoxic activity of NK cells and CD8⁺ cytotoxic T lymphocytes. The three treatments, particularly CEL NE and its combo, significantly increased IL12 levels by more than 5-fold in both treated and untreated tumors. The level of IL2, a type I cytokine promoting the differentiation of T cells into effector T cells and memory T cells [33], was increased by 1- to 3-fold after the three treatments in both treated and untreated tumors. These cytokines are all T cell or NK cell stimulators, and their increased expression levels are consistent with the increased T cell and NK cell populations and activation after treatments (Figure 5A). On the other hand, CEL NE and its combo with αPD-L1 decreased the Stat3 mRNA level, an immune-suppressive cytokine [34], in both treated and untreated tumors.

The above results showed that activated DC cells and NK cells increased in both treated and untreated tumors with CEL NE. NK cells express TNFα during killing tumor cells [35–37], while TNFα downregulates PD-L1 translation via miR-155 targeting 3’-UTR of PD-L1 mRNA [38, 39]. Thus, the systemic TNFα upregulation from the activated DCs and NK cells by CEL NE (Figure 5C) might account for the PD-L1 downregulation in the untreated tumors (Figure 4F). In contrast, the PD-L1 downregulation in the CEL NE treated- tumors resulted from the synergetic effect of the CEL-induced PD-L1 downregulation and CEL-induced NF-κB inactivation and TNFα production by activated DCs and NK cells.

2.6. CEL NE-induced systemic abscopal effect was dependent on CD4⁺ and CD8⁺ T cells and possessed long-lasting memory effect

It is reported that the effectiveness of ICD depends on CD4⁺ and CD8⁺ T cells [5]. To further confirm this, we depleted CD4⁺ and CD8⁺ T cells in the B16F10 tumor by anti-CD8 or anti-CD4 antibody during the CEL NE treatment. As shown in Figure 6A, depletion of either CD4⁺ or CD8⁺ T - cells abolished the tumor inhibition effect of CEL NE on both treated and untreated tumors, whereas an isotype IgG control did not affect the tumor inhibition of CEL NE. Thus, the dependence on CD4⁺ and CD8⁺ T-cells of CEL NE’s antitumor activity further supported the ICD effect of CEL NE. It should be noted that the activity of CEL NE in the treated tumor was also dependent on the CD4⁺ and CD8⁺ T - cells, indicating that the inhibition of tumor growth was not due to the direct cytotoxicity of CEL but immunological effect.

Figure 6. CD4⁺ and CD8⁺ T cell- dependence and memory effect of the CEL NE treatment.

(A) CD8⁺ and CD4⁺ T-cells depletion assay. Bilateral B16F10 tumor-bearing C57/BL6 mice were treated with i.p. injections of anti-CD8 (10 mg/kg) or anti-CD4 (10 mg/kg) antibody or isotype control (10 mg/kg) started from 2 days after inoculation for 5 treatments. CEL NE treatment was initiated by i.t. injection at left-side tumor (CEL, 0.15 mg/kg) on an every-3-day schedule for 4 times. The tumors were monitored and dissected and imaged on day 17 (T: treated tumors. UT: Un-treated tumors. n = 4; *p < 0.05, **p < 0. 01, ***p < 0. 001, NS: not significant.) (B) Long-term immune surveillance and memory immunity. B16F10 cells (1×10⁶) were inoculated to each mouse; On day 5 post-inoculation, the tumors were either i.t. injected with PBS or CEL NE (0.15 mg/kg) for 4 times. On day 20, the tumors in the mice were surgically removed and the mice were maintained. After 20 days of surgery, the mice were re-challenged by inoculation of 1×10⁶ B16F10 cells on the other side. The re-challenged tumors were monitored and dissected, weighed, and imaged on day 14 post-re-challenge. PBS^pre: PBS pre-treated mice. CEL NE^pre: CEL NE pre-treated mice. All data are shown as mean ± SD. n = 8. *p < 0.05, **p < 0. 01, ***p < 0. 001, NS: not significant.

We also surgically removed the CEL NE treated tumors, and 20 days later re-challenged the mice with the second batch of B16F10 tumor cells (Figure 6B). The cells formed new tumors, but the tumor burden of the mice whose tumors were previously treated with CEL NE was less than that in the PBS pre-treated mice. This suggests a long-lasting immune surveillance and memory immunity induced by the ICD effect of CEL NE.

2.7. CEL NE was effective in other melanoma models and could be systemically administered

CEL NE was also tested on another desmoplastic mouse melanoma tumor model, murine BRAF-mutant melanoma cell line BPD6 (Supporting Information Figure S12). Strikingly, a single i.t. injection of CEL NE at the low dose (0.15 mg/kg) arrested tumor growth by 50% compared to the PBS group for at least 12 days. Masson’s trichrome staining showed a large necrotic area and loss of fiber and collagen after CEL NE treatment. Thus, one shot of CEL NE significantly prolonged the overall survival.

We further tested the effectiveness of the intravenously (i.v.) injected CEL NE against B16F10 tumors at the same dose (Supporting Information Figure S13). CEL NE (loaded with DiD dye) mainly accumulated in the liver and lung, but a significant amount also accumulated in the tumor. Three i.v. injections of this low dose of CEL NE suppressed tumor growth to 50% of that of the PBS group. More importantly, the i.v. injected CEL NE effectively prevented melanoma spontaneous metastasis to the lung, which was significant in the PBS group. Splenomegaly was also alleviated after CEL NE treatments as a result of the less tumor burden and metastasis.

2.8. Evaluation of therapy safety

CEL is indeed very cytotoxic [40]. Its maximum tolerance dose of i.v. injected CEL was reported to be 4 mg/kg [41], so the most used dose was 2 mg/kg [42]. In this study, the CEL dose was as low as 0.15 mg/kg and was locally injected into the tumors. The resulting plasma CEL was extremely low. Therefore, the CEL NE treatments did not cause visible signs of toxicity, neither body weight loss (Supporting Information Figure S14A) nor abnormal changes in blood routine or hepatorenal functions (Supporting Information Figure S14B) during the therapies. The H&E analysis also showed no apparent histological damages to the main organs (heart, liver, spleen, lung and kidneys) after the treatments. Notably, spontaneous liver metastasis and lung metastasis were observed in the PBS- and αPD-L1-treated groups, while no metastasis was observed in the mice treated with CEL NE or CEL NE/αPD-L1 (Supporting Information Figure S14C). These results indicate that the “treat locally and act systemically” strategy with very low dose of CEL nanoemulsion formulation is effective and without side effects.

3. Discussion

PD-1/PD-L1 inhibition-based immunotherapy is now the standard-of-care first-line therapy for metastatic melanoma and has been extensively tested in the clinic for many other cancers. However, it has two major drawbacks: (1) Only partially effective: while some melanoma patients achieve a partial or complete response, about half of patients do not respond to the therapy [2] [3]. (2) Too costly: a typical cost using the most popular immunotherapy antibodies, Keytruda and Opdivo, is $150,000 per patient per year in the US.

Immunogenic cell death (ICD), a process that dying tumor cells release immunostimulatory danger signals, can modulate tumors immuno-microenviroment and improve the anti-PD-1/PD-L1 therapy efficacy [20]. However, currently known ICD inducers all promote tumor cells to express more PD-L1 (so helps tumors cells to more efficiently evade immunochecking and killing). Aditional more antiPD-1/PD-L1 antibodies are needed, further increasing the cost and the risk.

The most important innovation of this work is the demonstration of the first and much cost-effective immunotherapy strategy, chemotherapy induced immunotherapy against melanoma without the need of expensive immune checkpoint inhibitors. The significances include: (1) Effectively inducing ICD: Celastrol effectively induces ICD at very low doses (thus not inducing toxicity) via promoting autophagy, and thus modulates tumor immuno-microenvironment, boosting systemic-immune therapy. (2) Reducing tumor cell surface PD-L1: Here for the first time celastrol is found also to downregulate PD-L1 expression in tumor cells through NF-κB inhibition. So, the tumor cells are now readily recognized by T-cells and killed. (3) Needing no anti-PD-L1 antibody to boost the abscopal effect : The combination of above two characters makes celastrol alone effectively inhibit the growth of both treated and remote untreated tumors (mimicking metastatic tumors). (4) The drug nanoemulsion is the key: the developed celastrol nanoemulsion reduces minimal ICD inducing concentration in vitro and enables to maintain the celastrol concentration in tumor sufficiently high for effective ICD-inducing and PD-L1 downregulation. While administered free drug quickly leaks into bloodstream and cannot achieve these effects. (5) Low cost: The cost of celastrol would be only a small fraction of the antibody’s cost. This work may open a new area, i.e. cancer immunotherapy without using antibody.

4. Conclusion

In summary, we herein demonstrate an anti-PD-L1 antibody-free immunotherapy strategy — celastrol nanoemulsion-based immunotherapy against melanoma. CEL in the tumor effectively induces ICD, thereby activates DCs, recruits cytotoxic CD8⁺ T cells and NK cells, and increases immune-boosting cytokine levels but reduces immunosuppressive MDSCs and Tregs, leading to a potent systemic-immune therapy. More importantly, different from other ICD-inducing agents that increase PD-L1 expression in tumor cells and thus must rely on more anti-PD-L1 mAb to exert therapeutics, CEL simultaneously downregulates PD-L1 expression of tumor cells and thus directly activates the recruited T-cells. The developed nanoemulsion enables to maintain a persistent tumor CEL concentration for effective ICD - induction and PD-L1 downregulation. Therefore, low doses of CEL NE effectively inhibits the growth of both the treated and untreated tumors and prolongs the mouse survival.

5. Materials and methods

Materials.

Celastrol was purchased from Shanghai Tauto Biotech Co. Ltd (Shanghai, China). Soybean lecithin was purchased from Santa Cruz Biotechnology, Inc (Dallas, Texas). Pluronic F68 was provided by BASF (Florham Park, NJ). Sesame oil was purchased from Sigma-Aldrich, MO, USA.

Cell lines.

Murine melanoma B16F10, human melanoma A375 and M10 cell lines were purchased from American Type Culture Collection (ATCC), and cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Gibco, USA) in a humidified atmosphere with 5% CO₂ at 37 °C. Murine BRAF-mutant melanoma cell line BPD6 (BRAF^V600EPTEN^−/−, syngeneic with C57BL/6 mice) was provided by Brent Hanks (Duke Cancer Institute, Durham, NC) and cultivated in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin (Gibco, USA) in a humidified atmosphere with 5% CO₂ at 37 °C.

Antibodies.

Anti-mouse PD-L1 (clone B7-H1), anti-mouse CD8α (clone Lyt 2.1), anti-mouse CD4 (clone GK1.5), and IgG2a isotype were purchased from BioXcell (West Lebanon, NH). Primary antibodies, secondary antibodies for flow cytometry, and immunofluorescence staining are listed in Supplementary Table 1.

Preparation and characterization of celastrol nanoemulsion (CEL NE)

CEL NE was prepared by an ultrasonic emulsification method. Briefly, celastrol (1 mg) was dissolved in ethanol (10 μL) and then mixed with sesame oil (25 mg) and soybean lecithin (25 mg). Pluronic F68 solution (1 mL, 100 mg/mL) was added dropwise into the above drug mixture under stirring for 5 min at room temperature, then ultrasonicated on an ice bath for 5 min, and finally dialyzed against 5% glucose for 3 h to adjust the osmotic pressure and remove the ethanol.

The size and zeta potential of the CEL NE were determined by the Malvern ZetaSizer Nano series (Westborough, MA). Transmission electron microscopy (TEM, JEOL 1230) images were acquired after CEL NE nanoparticles were negatively stained with 2% phosphotungstic acid. The encapsulation efficiency (EE) of CEL NE was calculated as the percent of the amount of CEL loaded in the CEL NE analyzed using the HPLC system (Shimadzu LC-20AT, Kyoto, Japan) over the original feeding amount. Thermo Scientific C18 column (100 mm × 4.6 mm, 2.6 μm, Thermo Fisher Scientific, Waltham, MA USA) with 44:44/12 acetonitrile/methanol/water (1% formic acid) as the mobile phase (0.2 mL/min) at room temperature was used. Pure CEL appeared at 9.2 min was used as a standard.

In vitro CEL release assay

In vitro CEL release was studied by dialyzing 1 mL of CEL NE solution loaded in a sealed dialysis tube with a molecular weight cut off of 3.5 kDa against 30 mL of PBS at different pHs (pH 7.4, pH 6.5, pH 5.5 containing 0.5% w/v Tween80 to achieve sink conditions) and incubated at 37 °C under shaking (70 rpm). At timed intervals, 1 mL of the medium outside of the dialysis tube was collected and displaced with a fresh buffer solution. The CEL released at each time interval was detected by the HPLC system (Shimadzu LC-20AT, Kyoto, Japan), and the tests were conducted thrice.

Autophagy induction by free CEL

Mouse melanoma B16F10 and BPD6 cells were separately seeded on 6-well cell culture plates at 200,000 cells per well and incubated with CEL (0.1 ~ 8 μM) for 12 h. Cells without CEL treatment were served as the controls. The cells were then washed thrice with cold PBS, lysed with RIPA buffer (containing 1% protease inhibitor cocktail and 1% phosphatase inhibitor), and subjected to western blot analysis. GAPDH was used as an internal reference.

ICD detection and dose-dependent curve of free CEL and CEL NE

Mouse melanoma cell lines B16F10 and BPD6 and human melanoma cell lines A375 and M10 were separately plated onto glass-bottom petri dishes at 100,000 cells per dish in 1.5 mL of 10% FBS-containing cell culture medium and incubated for 24 h before use. For the detection of cellular surface CRT marker, the cells were incubated with free CEL or CEL NE for 4 h at concentrations ranging from 10 nM to 1 μM. Afterword, the cells were rinsed twice with PBS and fixed in 0.25% PFA for 5 min and then rinsed twice with PBS, incubated with CRT primary antibody (1:500, diluted in 1% BSA blocking buffer) for 30 min. Then, the cells were again rinsed 3 times with PBS and incubated with a rabbit anti-mouse secondary antibody for 30 min diluted in a cold blocking buffer. The cells were fixed with 4% PFA for 20 min and then mounted with Prolong® Diamond Antifade Mountant with DAPI. The ICD dose-dependent-curves of the 4 melanoma cell lines were drawn by calculating the percentage of CRT- positive cells using Image J software versus the concentration of free CEL or CEL NE. CRT- positive cells were also analyzed by flow cytometry.

For intracellular HMGB1 staining, cells were incubated with free CEL or CEL NE for 24 h at a concentration of 1 μM. The cells were then rinsed with PBS, fixed with 4% PFA for 20 min, permeabilized with 0.1% Triton X-100 for 10 min and rinsed 3 times with PBS. The nonspecific binding sites were blocked with 1% BSA in PBS for 30 min. HMGB1 primary antibody was added and incubated with cells for 1 h. Finally, the cells were rinsed 3 times with PBS and incubated with a secondary antibody for 30 min and mounted with Prolong® Diamond Antifade Mountant with DAPI. Primary and secondary antibodies are listed in Supplementary Table 1.

The immunofluorescence images were taken on a laser scanning confocal microscope (Zeiss LSM 700). Each sample was imaged for at least 5 randomly selected fields. Supernatant released HMGB1 was quantitated according to HMGB1 ELISA kit (orb409067 and orb406327, Biorbyt, Ltd.).

For in vivo ICD observation, 1×10⁶ B16F10 cells were inoculated in C57/BL6 female mice. When the tumor volume reached 500 mm³, 10 μL of CEL NE (CEL, 0.15 mg/kg) or PBS was i.t. injected into the tumor. After 24 h, the tumors were dissected and analyzed for ICD markers (CRT and HMGB1) using immunofluorescence staining.

DC cell activation determination

For in vitro determination of DC activation, mouse melanoma (B16F10 and BPD6) and human melanoma (M10 and A375) cells were separately seeded on 6-well cell culture plates at a density of 150,000 cells per well in 2 mL of 10% FBS-containing cell culture medium and incubated for 24 h before use. Then, CEL was added to each well at a concentration of 1 ıM and incubated with cells for 8 h. The medium was replaced with fresh medium and cultured for another 12 h. The supernatant medium was isolated as the conditioned medium which contained tumor released antigens (e.g., HMGB1) and ATP. DC2.4 cells cultured in CEL pre-pulsed melanoma cell-conditioned medium are denoted as melanoma/CEL^con. DC2.4 cells cultured in melanoma cell- conditioned medium are denoted as melanoma^con. DC2.4 cells were incubated in the above conditioned medium for 12 h. The positive control was the DC2.4 cells cultured with 0.5 μM LPS in a fresh culture medium. After culturing for 12 h, the cells were fixed with 4% PFA for 20 min and then permeabilized with 0.1% Triton X-100 for 10 min and rinsed 3 times with PBS. Then, phalloidin (Sigma, USA) was added to cells at the concentration of 50 μg/mL for 30 min. Afterword, the cells were rinsed 3 times with PBS and mounted with Prolong® Diamond Antifade Mountant with DAPI.

For in vivo determination of DC activation in the B16F10 tumor, tumor single-cell suspensions were prepared from the tumors for flow cytometry analysis of CD11c⁺ and MHCII⁺ double-positive markers. At the same time, tumors were also sectioned for immunofluorescence staining of CD86, CD11c, and MHCII.

Animal model establishment

Female C57BL/6 mice (6 to 8 weeks old) were purchased from Charles River Laboratories (Wilmington, MA). All animal procedures were carried out under protocols approved by the Institutional Animal Care and Use Committees at the University of North Carolina at Chapel Hill. B16F10 and BPD6 subcutaneous tumor models were established by inoculating 50 μL of 1×10⁶ B16F10 or BPD6 cells suspension into mouse left and right armpits (for bilateral tumor model) or single site (unilateral tumor) via a 28G insulin-gauge syringe.

The tumor volume was measured by caliper and calculated by V (mm³) = 0.5×a×b² (a represents the length and b is the width). Mice body weights were recorded every 3 days.

In vivo anticancer experiments

For αPD-L1 single therapy on the B16F10 tumor model, 5 days after inoculation, the mice were i.p. injected with αPD-L1 (5 mg/kg) every 3 days for 3 times. Tumor volumes were measured every 3 days after inoculation. For free CEL and MIT vaccination study, B16F10 cells were incubated with CEL or MIT at their IC₅₀ doses for 24 h, and then subcutaneously injected into C57/BL6 mice at 5×10⁵ cells per mouse. After seven days, the mice were re-challenged with 1×10⁶ B16F10 cells per mouse on the contralateral side, and the tumor volumes were measured every day after 5 days inoculation.

For CEL NE anti-cancer study, the mice were intratumorally injected with CEL NE (0.15 mg/kg) or free CEL (0.15 mg/kg) (n = 5). Tumor volumes were measured every day on Day 8 post-inoculation, and the tumor area was photographed on 12 d, 14 d, 16 d, and 18 d post-inoculation. For long-term immune surveillance and memory immunity. B16F10 cells (1×10⁶) were inoculated to each mouse; On Day 5 post-inoculation, the tumors were i.t. injected with PBS or CEL NE (0.15 mg/kg) for 4 times. On Day 20, the tumors in the mice were surgically removed, and the mice were maintained. After 20 days of surgery, the mice were re-challenged by inoculation of 1×10⁶ B16F10 cells on the other side. The tumors were monitored after re-challenge. The tumors were dissected, weighed, and imaged on day 14 post-re-challenge.

For B16F10 bilateral tumor inhibition study, 5 days after inoculation, the mice were grouped randomly (n = 5) and the treatments were initiated as follows: (a) PBS control, (b) αPD-L1 (5 mg/kg, i.p.), (c) CEL NE (0.15 mg/kg, i.t.), (d) CEL NE (0.15 mg/kg, i.t.) and αPD-L1 (5 mg/kg, i.p.) combo therapy. The treatments were initiated via i.t. injection of 10 μL of either PBS (a, b) or CEL NE (c, d) at first and 3 days after the first administration, αPD-L1 was i.p. injected (b, d) every 3 days for 3 times. For (c, d), CEL NE was given 4 times in total. The tumor volumes and mouse body weights were measured at the time of injection. The endpoint for each mouse was defined as the tumor volume reached 2000 mm³. Six days after the last injection, the animals were sacrificed, and their tumors and other organs were dissected and analyzed using H&E staining, Masson’s trichrome staining, immunofluorescence staining, flow cytometry analysis, Western blot, and RT-PCR assay.

For the CD8⁺ and CD4⁺ T-cells depletion assay, B16F10 bilateral tumor-bearing C57/BL6 mice (n = 4) were treated with three daily i.p. injections of an anti-CD8 or anti-CD4 antibody or isotype control (10 mg/kg) started from 2 days after inoculation for 5 times in total. CEL NE treatment was initiated by i.t. injection at the left-side tumor (CEL, 0.15 mg/kg) on an every-3-day schedule for 4 times.

For i.v. injected CEL NE anticancer activity study, CEL NE (0.15 mg/kg) or PBS was injected via the tail vein to the B16F10 unilateral tumor-bearing mice on Day 7 post-inoculation on an every-3-day schedule for 4 times. On 28 days post-inoculation, the mice were euthanized, and their tumors, spleens, and lungs were dissected. Spontaneous metastatic tumor nodules in the lungs were counted and statistically analyzed.

For the BPD6 bilateral tumor inhibition study, 10 days after inoculation, CEL NE or PBS was i.t. injected into the one side tumor. The treated and un-treated tumor volumes were measured every 3 days. On 12 days post-inoculation, the tumors were dissected for Masson’s Trichrome staining.

For survival studies, the mice were maintained after treatments until the endpoint. The survival curves were drawn according to mice survival percentages with time elapse.

Immunofluorescence staining

The tumor tissue was dissected and soaked in 4% neutral-buffered paraformaldehyde for 48 h, dehydrated in 15% sucrose and 30% sucrose solution, respectively, and then cryo-embedded in OCT compound- embedding medium (SAKURA). Tumor sections of 5-μm-thick were cut using a cryostat (Leica CM1950; Leica Microsystems CMS GmbH, Wetzlar, Germany). The sections were rinsed 3 times with PBS, blocked in 1% BSA in PBS at room temperature for 1 h, rinsed 3 times with PBS, and then incubated with primary antibodies at 4 °C overnight. Finally, the slices were incubated with fluorescent secondary antibodies at 37 °C for 1 h and mounted with Prolong® Diamond Antifade Mountant with DAPI (ThermoFisher Scientific). Primary antibodies and fluorescent primary/secondary antibodies used for immunofluorescence staining are listed in Supplementary Table 1. Immunofluorescence images were taken on a laser scanning confocal microscope (Zeiss LSM 700). Each section was imaged for at least 5 randomly microscopic fields and quantified by ImageJ software.

H&E staining and Masson’s Trichrome staining

Tumor tissue was fixed in 4% neutral-buffered PFA, paraffin-embedded, and sectioned into 5-μm-thick slices for hematoxylin and eosin (H&E) staining or Masson’s Trichrome staining (Abcam) for immunohistochemical studies. The stained sections were imaged by a microscope (BX61, Neville).

TUNEL assay

Tumor slides were conducted under the DeadEnd fluorometric TUNEL system (Promega, Madison, WI) instructions, mounted with Prolong Diamond Antifade Mountant with DAPI (ThermoFisher Scientific), and then subjected to confocal microscopy imaging (Zeiss, LSM 700). Apoptotic cells with fragmented DNA (FITC-positive) indicate TUNEL-positive nuclei. The TUNEL positive cells were quantitatively analyzed using ImageJ software.

Biodistribution and pharmacokinetic study

DiD (0.75 mg/kg)-loaded CEL NE (0.15 mg/kg) was prepared via the same method as described. For the B16F10 bilateral tumor biodistribution study, mice (n = 3) were i.t. injected with DiD-CEL NE at one side tumor and sacrificed after 4 h, 12 h, and 24 h. The treated/un-treated tumors and main organs were dissected, rinsed with PBS, and imaged with an IVIS kinetics optical system (PerkinElmer, CA) (Ex = 640 nm, Em = 670 nm). For the B16F10 unilateral tumor biodistribution study, mice (n = 3) were i.v. injected with DiD-CEL NE (0.15 mg/kg) and sacrificed after 12 h, 24 h, and 48 h for collection of tumors and main organs, rinsed with PBS and imaged under IVIS kinetics optical system. Acquired images were obtained by superimposing the emitted light over the grayscale photographs of the animals. Quantitative analysis was performed using the Living Image 4.5.2 software.

For measurements of CEL concentrations in plasma and treated/un-treated tumors, mice bearing B16F10 bilateral tumors were i.t. injected with either free CEL or CEL NE at one side tumor. At predetermined times the mouse blood samples were collected and centrifuged at 3000 rpm for 10 min. The separated plasma was placed in heparin-rinsed Eppendorf tubes and stored at −20 °C for analysis. For plasma sample preparation, 50 μL of plasma sample was mixed with 200 μL ethyl acetate and vortexed for 10 min and then centrifuged at 15,000 rpm for 10 min at 4 °C. The extracted CEL solution was evaporated dry and reconstituted in 100 μL of 50% acetonitrile/50% methanol supplemented with 1% formic acid. For tumor sample preparation, the tumor tissue was homogenized in saline (the ratio of sample weight and volume was 1:1), and its proteins were precipitated with 5 volumes of 50% acetonitrile/50% methanol supplemented with 1% formic acid. The mixture was centrifuged at 15,000 rpm for 5 min at 4 °C. The supernatant was filtered through a 0.22 μm filter and subjected to the HPLC system (Shimadzu LC-20AT, Kyoto, Japan) for analysis.

Flow cytometry analysis of immune cell populations in tumors

Tumor tissue was dissected and digested with collagenase IV and DNAase at 37 °C for 1 h, dispersed in MACs buffer (1× PBS + 2 mM EDTA + 0.5% BSA, filter sterile), and centrifuged at 1,400 rpm for 5 min. Then, the supernatant was discarded, and the ACK buffer was added to dissociate red blood cells. The cells were then dispersed in MACs buffer and centrifuged at 1,400 rpm for 5 min. Then, 2×10⁶ cells were incubated with fluorescently labeled antibodies for 30 min at 4°C for surface marker expression analysis. For staining intracellular markers, the cells after centrifuge were incubated with permeabilized buffer (BD, Franklin Lakes, NJ) for 30 min at room temperature, washed twice with the buffer, incubated with intracellular fluorescently labeled antibodies for 15 min at 4 °C, and then washed twice with the buffer again. The cells were then fixed in 4% PFA and subject to flow cytometry analysis via FACS (BD LSR II). The fluorescence conjugated antibodies used for flow cytometry are listed in Supplementary Table 1. Data were analyzed with FlowJo V10 software (TreeStar).

Quantitative real-time PCR (qPCR) assay

Total RNA was extracted from whole tumor tissue using RNeasy^® Microarray Tissue Mini Kit (Qiagen) and then reverse-transcripted with iScript^™ cDNA Synthesis Kit (BIO-RAD). Then cDNA templates were amplified with TaqMan^™ Gene Expression Master Mix using the 7500 Real-Time PCR System. Primers for RT-PCR reactions are listed in Supplementary Table 2. GAPDH was used as an endogenous control.

Western blot assay

B16F10 tumor tissue was homogenized with RIPA lysis buffer. Total protein concentration was determined using Pierce^™ BCA Protein Assay Kit (Thermo Scientific). For each sample, 40 μg total proteins were loaded and separated on 4% – 12% SDS-PAGE gel (Invitrogen) electrophoresis and then transferred onto PVDF membranes (Bio-Rad, Hercules, California). The membranes were blocked with 5% BSA and then incubated with primary antibodies at 4 °C overnight. Then, the membranes were washed 3 times with TBST buffer and incubated with horseradish peroxidase-coupled secondary antibody at a 1:3,000 dilution. Finally, the membrane was rinsed and visualized with the Pierce ECL Western Blotting Substrate (Thermo, Rockford, IL). GAPDH was used as an internal reference. The relative protein expression level was quantified using Image J software.

Biosafety evaluation

At the end of the B16F10 bilateral tumor inhibition study, the mouse whole blood samples were collected and subjected to the UNC histology facility for the analysis of blood cells and serum chemistry. The levels of serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), creatinine (CREAT), and the blood urea nitrogen (BUN) were measured to evaluate the hepatorenal functions. Red blood cells (RBCs), white blood cells (WBCs), platelets (PLTs), hemoglobin (HGB), hematocrits (HCTs), lymphocyte (LYMPH), and monocyte (Mono) counts indicate the immune homeostasis. At the same time, the major mouse organs were collected and fixed in 4% neutral-buffered PFA for analysis of organ toxicity and spontaneous melanoma metastasis using H&E staining. Mouse body weights of each group were monitored and recorded during the whole tumor-suppression study

Statistical analysis

Experiments were repeated at least 3 times, and each measurement was performed in triplicate. For individual comparisons, one-way ANOVA and a two-tailed, unpaired Student’s t-test were performed using Prism 5.0 software. Data were compared with the PBS control group and between groups. Data were presented as means ± SD. Assignments and selections of microscopic inspection fields were made at random. Significant differences of mouse survivals were assessed using log rank test. ***p < .0001, **p < 0.01, *p < 0.05, NS, not significant.