Background
In 2018, approximately 41,400 breast-cancer-related deaths will occur in the USA [
1]. About 90% of these deaths will be due to metastases. Since only about 4% of the 265,000+ new patients with breast cancer are typically diagnosed with stage IV metastatic cancer, the vast majority of breast-cancer-related deaths are due to the recurrence and progression of breast cancers initially diagnosed at stages I–III. In an attempt to prevent tumor recurrence, approximately four out of every five patients with breast cancer receive adjuvant therapies such as chemotherapy, hormone therapy, and/or radiotherapy following tumor resection [
2]. Even with state-of-the-art adjuvant treatments, the 5-year recurrence rates for stage I, II, and III breast cancer are 7%, 11%, and 13%, respectively [
3]. After 10 years, the overall breast cancer recurrence rate increases to 20% [
3]. Furthermore, side effects associated with current adjuvant therapies can be life-altering and even life-threatening [
4]. Thus, strategies capable of more effectively and more safely preventing progressive breast cancer recurrences, particularly after standard-of-care tumor resection, are urgently needed.
Adjuvant breast cancer vaccines are of interest due to their potential to educate a patient’s immune system to recognize and eliminate occult tumor cells before a recurrence can develop. In particular, autologous tumor cell vaccines (ATCVs) comprise a promising class of vaccines capable of inducing personalized, polyclonal anti-tumor immune responses [
5‐
14]. Patient/tumor-specific polyclonal immune responses are especially relevant for breast cancer with high intra-patient and inter-patient molecular heterogeneity that facilitates resistance to targeted therapies [
15‐
19]. Because ATCVs are generated from a patient’s own malignant cells, they present a complete and personalized library of tumor-associated antigens (TAAs). In contrast, peptide-based vaccines deliver one or a couple different peptides and are prone to tumor escape through downregulation of the targeted epitope(s). Furthermore, since ATCVs are “antigen agnostic,” they could be used in the management of any subtype of breast cancer including triple-negative breast cancers (TNBCs), which lack hormone and human epidermal growth factor receptor 2 (HER2) receptors, the usual targets for breast cancer therapies.
While ATCVs have been shown to be safe and active in numerous clinical studies, a major barrier to their widespread clinical use is inconsistent, if not limited, immunogenicity. Patient-derived cancer cells, which form the basis of the vaccine, have undergone extensive immunoediting to avoid elimination by the host’s immune system [
20]. Common mechanisms that cancer cells use during immune escape include (1) downregulation of major histocompatibility complex (MHC I/II) molecules and development of defects in antigen presentation; (2) downregulation of costimulatory molecules, such as B7–1 and B7–2; (3) upregulation of immunoinhibitory molecules, such as programmed death-ligand 1 (PD-L1); (4) loss or modification of tumor-associated antigen(s); and (5) increased production of immunosuppressive factors such as indoleamine 2,3-dioxygenase (IDO), IL-10 and tumor growth factor (TGF)β [
21]. As a result, nearly all ATCVs currently under development utilize strategies to boost tumor cell immunogenicity through one or more of the following: transfection of autologous tumor cells with costimulatory molecules [
22‐
26], conjugation of immunostimulatory moieties to autologous tumor cells [
10,
27]; co-formulation with immunostimulatory molecules [
6,
8,
27‐
30]; or engineering autologous tumor cells to secrete adjuvant cytokines [
9,
31‐
41]. Employing these strategies has demonstrated significant increases in antitumor immunity against various malignancies in clinical studies [
8‐
10,
26,
27,
33,
36,
37,
39,
41‐
43].
For breast cancer, ATCV clinical studies have been limited to three completed [
44‐
46] and two active trials [
47,
48]. All three completed studies show promise in generating antitumor responses [
49]. Despite the relatively small number of clinical studies, breast cancer remains an ideal indication for ATCV deployment as (1) 62% of breast cancer cases are diagnosed at stage I, where the tumor is still localized in the breast with minimal impact on the patient’s immune status [
50]; (2) nearly all patients with breast cancer undergo tumor resection, thus ensuring a source of tumor cells for ATCV production; and (3) the vast majority of patients with breast cancer have minimal, if any, detectable disease after resection so the tumor burden is low.
Because of the aforementioned heterogeneity in breast cancer, it is expected that breast ATCVs will display varying degrees of immunogenicity. Thus, the goal of this study was to begin to define the primary determinants of ATCV immunogenicity by comparing two murine models of breast adenocarcinoma, 4T1 and EMT6: 4T1 is a poorly immunogenic murine breast cancer cell line that shares many features with human stage IV breast cancer [
51‐
53]. EMT6 on the other hand, is a highly aggressive, yet immunogenic cell line [
54‐
56]. By understanding the key drivers of breast cancer immunogenicity, we may be able to directly and more efficiently enhance ATCVs during ex vivo modifications. At the very least, data gathered could be used to identify which patients are better candidates for adjuvant ATCV therapy. During the study, we observed that myeloid-derived suppressor cells (MDSCs) played a dominant role in influencing breast ATCV immunogenicity. The immunosuppressive role of MDSCs in breast cancer progression and metastasis is well-documented [
57‐
60]. In particular, the levels of circulating MDSCs were found to correlate with clinical stage and metastatic tumor burden [
61]. However, to the best of our knowledge, the influence of MDSCs on ATCV efficacy has not been explored. Thus, the focus of the latter stages of this study shifted towards identifying and blocking the origin of breast-cancer-related MDSCs as a strategy to enhance ATCV immunogenicity.
Methods
Cell culture
Murine breast adenocarcinoma cells 4T1 and EMT6 were purchased from American Type Culture Collection (Manassas, VA, USA). The rest of the breast cancer cells, namely 4T07, 67NR, 66Cl4, 168FARN were a generous gift from Dr Fred Miller (Karmanos Cancer Institute, Detroit, MI, USA). All cell lines except EMT6 cells were maintained in Dulbecco’s modified eagle medium (DMEM), supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). EMT6 cells were maintained in Roswell Park Memorial Institute-1640 (RPMI-1640) medium, supplemented with 15% FBS and 1% P/S. All cells were cultured at 37 °C in a humidified incubator with 5% CO2.
Mice
All experimental procedures were approved by the Institutional Animal Care and Use Committee at University of Arkansas. Female Balb/cByJ mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and were housed in microisolator cages. Mice were utilized for experiments at 8–12 weeks of age and animal care followed The Guide for Care and Use of Laboratory Animals (National Research Council).
In vitro proliferation assay
The 4T1 and EMT6 cells were irradiated at 0, 20, 40, 60, 80, or 100 Gy using a Gammacell 1000 cesium irradiator. Cells were then plated in triplicate on a 96-well plate and incubated at 37 °C for 24, 48, 72, or 96 h. After incubation, 20 μl of CellTiter 96 Aqueous One Solution Reagent from Promega (Madison, WI, USA) was added to each well and incubated for another hour. Using a Biotek Synergy 2 plate reader from Biotek Instruments Inc. (Winooski, VT, USA), absorbance was measured at 490 nm and compared to the absorbance of similarly treated known numbers of irradiated 4T1/EMT6 cells to determine the number of viable cells in the sample wells.
Expression of MHC and costimulatory molecules
Irradiated (100 Gy) and non-irradiated 4T1 and EMT6 cells (5 × 105) were stained with fluorochrome-conjugated anti-CD80 (clone 16-10A1), anti-CD86 (clone GL1), anti-H-2Kb (MHC I) (clone AF6–88.5), anti-I-Ad/I-Ed (MHC II) (clone M5/114.15.2), anti-CD54 (ICAM-1) (clone 3E2), and anti-CD95 (FasR) (clone Jo2) (BD Biosciences). Cells were analyzed on a FACSCantoII and differences in median fluorescence intensities (ΔMFI) between unstained and stained cells were determined using FlowJo software (Tree Star, San Carlos, CA, USA).
In vitro cytokine analysis
The cells (5 × 105 4T1 or EMT6 cells, untouched or irradiated, and 5 × 105 untouched 4T07, 67NR, 168FARN or 66Cl4 cells) were seeded in separate T25 flasks and cultured for 48 h. Cell culture supernatants were collected and centrifuged to remove any non-adherent cells and stored at − 80 °C until analysis. From the untouched and irradiated 4T1 or EMT6 cells, levels of monocyte-colony stimulating factor (M-CSF), vascular endothelial growth factor (VEGF), transforming growth factor-β (TGF-β), interleukin-6 (IL-6), monocyte chemotactic protein (MCP-1), GM-CSF and G-CSF in cell culture supernatants were quantified. On the other hand, the cell culture supernatants from untouched 4T07, 67NR, 168FARN and 66Cl4 were only evaluated for G-CSF. Levels of M-CSF, VEGF and TGF-β were analyzed using ELISA kits from R&D systems Inc. (Minneapolis, MN, USA) and Biolegend (San Diego, CA, USA). Levels of IL-6, MCP-1, GM-CSF, and G-CSF were analyzed using a cytometric bead array (CBA) on a FACSCantoII from BD Biosciences.
CRISPR/Cas9 genomic deletion of G-CSF
Using the CRISPR design tool provided by the Zhang laboraoty at Massachussetts Institute of Technology (MIT) (http://crispr.mit.edu/), a 20-bp guide sequence targeting the G-CSF gene in 4T1 cells was identified. Guide sequences were cloned into separate pCas-Guide-EF1a-green fluorescent protein (GFP) plasmid via Origene’s cloning service. Plasmids were amplified in Escherichia coli and isolated via QIAGEN Plasmid Maxi Kit. For transfection, plasmid encoding guide RNA (gRNA) (10 μg) was mixed with Lipofectamine™ 3000 reagent (ThermoFisher) and added to 1 × 106 4T1 cells pre-seeded in a 6-well plate. After 48 h, cells expressing GFP were sorted using a FACSAriaIII system (BD Biosciences). Sorted cells were subsequently cloned by limiting dilution. G-CSF expression was quantified by enzyme-linked immunosorbent assay (ELISA) from R&D systems Inc. (Minneapolis, MN, USA). A mixture of clones producing lower than detectable levels of G-CSF were identified and denoted as 4T1.G-CSF−.
Prophylactic vaccination studies
Tumor cell vaccines were generated by irradiating 4T1 or EMT6 cells at 100 Gy using a Gammacell 1000 cesium irradiator. Mice were subcutaneously vaccinated with a primary and booster vaccine 10 days apart, which comprised 1 × 106 irradiated 4T1 cells (4T1 vaccine) or 5 × 105 irradiated EMT6 cells (EMT6 vaccine). For mice in the ipsilateral and contralateral hybrid vaccine groups, 1 × 106 irradiated 4T1 cells and 5 × 105 irradiated EMT6 cells were subcutaneously injected on the same and opposite flanks, respectively. In some instances, where the effect of G-CSF on overall survival was investigated, mice received 4T1.G-CSF− cells in place of 4T1 cells. Further, all vaccinated mice were challenged with 1 × 106 live 4T1, 5 × 105 live EMT6 cells or 1 × 106 live 4T1.G-CSF− cells, 10 days after the booster vaccine. Tumor volumes were recorded 2–3 times per week using the formula:
V = (w × w × l)/2,
where V is tumor volume, w is tumor width and l is tumor length.
G-CSF in serum from mice
When tumor volumes in mice bearing 4T1, 4T1.GCSF−, 4T07, 67NR, 168FARN and 66Cl4 reached about 500 mm3, about 400–500 μl of blood was collected in microcentrifuge tubes by submandibular bleeding. After allowing the blood to clot for 30 min at room temperature, samples were centrifuged at 2000 × g for 10 min at 4 °C. The serum was carefully collected from each sample and the levels of G-CSF were determined by ELISA (R&D systems Inc.; Minneapolis, MN, USA).
Tissue collection and analysis of immune cell subsets
Spleens and draining lymph nodes (DLNs) from 4T1 and 4T1.GCSF− tumor-bearing mice were isolated when tumors reached 500–700 mm3. Single cell suspensions were prepared by mechanically dissociating spleen and DLNs with a syringe plunger and passing samples through a 40-μm nylon mesh cell strainer. Splenocytes were additionally treated with ammonium-chloride-potassium buffer (Lonza, Allendale, NJ, USA) for 10 min to lyse red blood cells. Single cell suspensions were then blocked with purified rat anti-mouse CD16/CD32 monoclonal antibody (BD Biosciences) and stained with fluorochrome-conjugated anti-CD11b (clone M1/70), anti-CD19 (clone 1D3), anti-Ly6G and Ly6C (clone RB6-8C5), anti-CD25 (clone PC61), anti-CD4 (clone GK1.5), and anti-CD3ε (clone 145-2C11) (BD Biosciences).
Cells were then rinsed, fixed and permeabilized with 1× Perm/Wash buffer from BD Biosciences. The permeabilized cells were further stained with fluorochrome-conjugated anti-FoxP3 and read on a BD FACSCanto II flow cytometer. Frequencies of MDSCs, T cells, B cells, and regulatory T cells (Tregs) in the single cell suspensions were determined using FlowJo software (Tree Star, San Carlos, CA, USA). For mice bearing 4T07, 67NR, 66Cl4 and 168FARN tumors, only the spleens were isolated and stained for MDSCs.
Statistical analysis
All data were analyzed using GraphPad Prism software, version 7 (GraphPad Software, Inc., San Diego, CA, USA). For all in vivo vaccine studies, Kaplan-Meier tumor-free survival curves were plotted and statistical comparisons made using the log rank test. For all other studies, data are represented as mean ± standard deviation. For the experiments that compare cytokine release and expression of MHC and costimulatory molecules by 4T1 and EMT6 before and after irradiation, statistical comparisons were made using two-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison post-hoc test. For experiments where different immune cell subsets in spleen and DLN of mice bearing 4T1 or 4T1.G-CSF− tumors are compared to subsets in naïve mice, statistical comparisons were made using the Kruskal-Wallis test followed by Dunn’s post-hoc test. For all other experiments, statistical comparisons were made using one-way ANOVA followed by Tukey’s post-hoc analysis.
Discussion
A major barrier to the widespread development of ATCVs as an adjuvant to breast tumor resection is the sporadic, often poor, immunogenicity of resected breast cancer cells. Thus, in this study, we set out to determine the key factors influencing the immunogenicity of breast cancer ATCVs by comparing two murine breast cancer cells, 4T1 and EMT6. This study was not meant to provide a comprehensive account of immunogenic and immunosuppressive elements in breast cancer, but rather a springboard for further exploration of key factors in other tumor models and clinical samples.
Though the use of irradiated tumor cells to develop autologous tumor cell vaccines is reported extensively in the literature, we first wanted to demonstrate that we could effectively inactivate 4T1 and EMT6 cells using this approach. We tested for the effect of different doses of irradiation (20, 40, 60, 80, and 100 Gy) on tumor cell proliferation using a proliferation assay as described in “Methods”. We confirmed that at all five doses, both 4T1 and EMT6 cells failed to proliferate over the observed period of 4 days (Fig.
1). Thus, for experiments throughout this study, we chose 100 Gy, the highest dose tested, as a standard to develop our tumor cell vaccines.
The 4T1 cell line is often referred to as poorly immunogenic, while EMT6 is considered as a highly immunogenic cell line. Since the immunogenicity of these two cell lines has never been directly compared in any study, we first confirmed the differences in their immunogenicity in the ATCV setting. We vaccinated mice with irradiated 4T1 or EMT6 cells and subsequently challenged with live 4T1 or EMT6 cells. We found that mice immunized with EMT6 cells developed partial-to-complete protective immunity against live EMT6 challenge. On the other hand, 4T1 vaccine failed to provide any measurable protective immunity (Fig.
2). These data were consistent with previous publications [
54,
74].
To explore potential causes of immunogenic differences, we first looked at the expression of the immunologically relevant surface molecules MHC I, MHC II, B7–1, B7–2, ICAM-1, and FasR. We found that irradiated EMT6 cells express significantly higher levels of MHC I, B7–1, B7–2, and FasR, which could be responsible for the enhanced immune response to EMT6 vaccine (Table
1). We next analyzed cytokines released by each cell line and found that irradiated 4T1 cells released very high levels of GM-CSF, M-CSF, and in particular G-CSF (Fig.
3). Each of these cytokines can be immune-activating or immune-suppressive depending on their concentrations and context [
75‐
78].
The hybrid vaccine studies (Fig.
4) were designed to help tease out the relative contributions of cell surface markers versus cytokines on ATCV responses. The 4T1 vaccine was unlikely to improve the activity of the EMT6 vaccine as the former was found to be non-immunogenic. However, if the 4T1 vaccine had no effect on the EMT6 vaccine, or if the 4T1 vaccine reduced the efficacy of the EMT6 vaccine only locally, i.e. in the ipsilateral setting, then decreased costimulation (signal 2) on the part of the 4T1 vaccine could have induced a localized tolerogenic effect. One also could have argued that local immunosuppressive cytokines may have inhibited the EMT6 vaccine as well. Conversely, if the 4T1 vaccine inhibited the EMT6 vaccine both locally and systemically, as was observed, then a soluble factor secreted at high levels by 4T1 cells must be responsible for the inhibited EMT6 vaccine response.
Of the different cytokines released by 4T1 cells, G-CSF was produced at exceptional levels (Fig.
3). At such high levels, G-CSF and other colony stimulating factors have been associated with MDSC expansion and immune impairment [
71,
79‐
82]. Not surprisingly, we found that MDSC levels more or less correlated with G-CSF concentrations produced by five sister breast cancer cell lines: 4T1, 4T07, 67NR, 66Cl4 and 168FARN (Fig.
8). Although they did not measure cytokine production, Talmadge et al., also found significant differences in MDSC frequency among different breast tumor models, with 4T1 inducing the highest levels [
83]. It should also be noted that, although other publications have suggested a link between G-CSF and/or MDSCs and tumor metastasis [
64,
80,
84,
85], mice bearing 4T07 tumors, which do not induce visible metastatic lesions [
73], produced high levels of G-CSF and more splenic MDSCs than mice bearing any other tumor including highly metastatic 4T1 or 66CL4 tumors. These data suggest that other factors, such as IL-6 [
86], may be involved in breast cancer metastasis. At the very least, the relationship between G-CSF/MDSCs and metastasis may be model-dependent.
The correlation between G-CSF and MDSC accumulation was further solidified by functional deletion of G-CSF. Mice with 4T1 tumors, when compared to mice with 4T1.G-CSF
− tumors, had increased amounts of MDSCs in both spleens and DLNs (Figs.
6,
7). It should be noted that relatively high levels of CD11b
+Ly6G
+Ly6C
+ cells were found in the DLNs of mice bearing 4T1.G-CSF tumors (Fig.
7). We did not evaluate the immunosuppressive activity of these cells, but it is unlikely that they were highly suppressive, if at all, given that mice vaccinated with 4T1.G-CSF
− cells were protected from a 4T1.G-CSF
− tumor challenge. This is a reminder that MDSCs are a diverse family of cells and that not all CD11b
+Ly6G
+Ly6C
+ can be classified as MDSC [
87]. While there were other differences in immune subset populations between 4T1 and 4T1.G-CSF
− tumor-bearing mice (Figs.
6,
7), similar to differences in costimulatory molecules, these were overshadowed by the enormous differences in the MDSC populations. To verify that tumor-derived G-CSF was responsible for abrogating vaccine efficacy, we repeated the hybrid vaccine study utilizing 4T1.G-CSF
− cells in the contralateral vaccine group. We found that the hybrid vaccine containing 4T1.G-CSF
− cells was far less immunosuppressive than the hybrid vaccine containing parental 4T1 cells (Fig.
9). Overall, the findings from this study establish a causal link between tumor-derived G-CSF and a loss of responsiveness to breast ATCVs.
As mentioned previously, the finding that 4T1-derived G-CSF leads to MDSC accumulation and immune suppression is not novel. However, that this immune impairment can be eliminated by knocking out a single, non-essential protein, G-CSF, despite an otherwise aggressive phenotype was somewhat surprising. In fact, it should be noted that after G-CSF deletion, the 4T1.G-CSF
− vaccine was more immunogenic and more protective than the EMT6 vaccine (Fig.
9b versus Figs.
2,
4). These data imply that tumor cell surface phenotype is not as important as tumor-derived secreted factors when establishing breast ATCV immunogenicity.
Although this study has provided useful insight into the effect of tumor-derived factors on ATCV efficacy, we acknowledge that it has a few limitations and opportunities for additional exploration. First, through the entirety of the study, tumor cell vaccines were only used in a prophylactic setting. To truly recapitulate the effect of the tumor-derived factors, future studies ought to focus on vaccinating the mice post tumor resection. Second, this study only used the 4T1 cell line to establish the causal link between G-CSF secretion and ATCV efficacy. Future studies that involve either knocking in or knocking out G-CSF in the other cancer cell lines that intrinsically secrete low or high levels of G-CSF will further strengthen the findings of this study. Third, there is no question that G-CSF-induced MDSCs are responsible for vaccine impairment in our models. However, G-CSF could also be causing immune suppression through additional pathways. In a recent clinical study, G-CSF was highly expressed in tumors in patients with breast cancer with more aggressive disease and was correlated with poorer overall survival [
88]. As this study illustrates, tumor-associated macrophages (TAMs) are another key immunosuppressive subset that is strongly influenced by G-CSF. Assessing differences in TAM number and function between 4T1 and 4T1.G-CSF
− tumors is the subject of ongoing research. Likewise, while G-CSF is clearly an important target in breast cancer, it is important to note that our findings do not eliminate the possibility of other mechanisms that could be involved in MDSC expansion. For instance, knocking out other colony stimulating factors such as GM-CSF may have a similar effect on vaccine efficacy. Last, given that G-CSF and GM-CSF are routinely administered to prevent neutropenia in patients with breast cancer undergoing chemotherapy, a closer look at the immunosuppressive impacts of these cytokines, particularly in a setting of minimal residual disease, is warranted.