Background
Breast cancer is the most commonly diagnosed cancer and the leading cause of cancer death among women worldwide [
1]. Breast cancer comprises different pathological subtypes and has infiltrating immune cells or therapeutic targets heterogeneously presented within and between patients [
2]. The heterogeneity of breast tumor immune microenvironment (TIME) is considered the major cause of treatment failure in cancer immunotherapy. The use of anti-programmed death-ligand 1 (anti-PD-L1), programmed death-1 (anti-PD-1) and cytotoxic T-lymphocyte-associated protein 4 (anti-CTLA-4) has been shown to increase life expectancy in patients with certain cancers. However, these molecules are in fact not valuable for all cancers [
3,
4]. The factors affecting immunotherapy mainly include PD-L1 expression levels, the content of tumor infiltrating T cells (TILs), tumor mutation burden, microsatellite instability, function of gene mismatch repair and the patient history of chemoradiotherapy [
5,
6]. Previous evidence has revealed the clinical efficacy of PD-1/PD-L1 antagonists in a small group of metastatic breast cancer patients. It has been shown that breast cancer patients with triple negative, PD-L1 positive and higher levels of TILs will have better clinical outcomes [
7]. Low immunogenicity and T cell infiltration with immunosuppressive tumor microenvironment impede the success of immunotherapy in breast cancer [
8]. Ipilimumab (anti-CTLA-4 antibody) has shown promising results in treating melanoma, but progress has been slow in other cancer types because of low response rate and immunotherapy-related adverse events. [
9,
10]. Therefore, it is critical to develop alternative immune therapeutic targets that might be functional in the non-responders.
Members of the B7 family of coregulatory molecules have been shown to play a crucial role in regulating tumor-specific response [
11]. B7-H4, also known as V-set domain containing T cell activation inhibitor 1 (
VTCN1/B7x/B7 homolog4/B7S1), is a type 1 transmembrane protein that belongs to co-inhibitory B7 family ligands, which has been reported to exert critical effects on the inhibition of T cell-mediated immune response. B7-H4 was found to be constitutively expressed in several cancer cells including ovarian cancer [
12], prostate cancer [
13], melanoma [
14] and invasive ductal and lobular breast cancers [
15‐
17]. Previous studies including ours have demonstrated that B7-H4 in breast tumor cells is a negative regulator of CD8 T cell activation, expansion and cytotoxicity, resulting in low T cell infiltration in TIME [
15]. Conversely, tumor growth inhibition was not observed in the B7-H4
−/− mouse model, and the absence of B7-H4 leads to a reduction in CTLs (Cytotoxic T-Lymphocyte) granzyme B levels and the inability of tumor-specific T cells in the breast TIME [
18]. These inconsistent findings might be due to the different tumor models and the heterogeneity of TIME.
Evidence shows that in the estrogen receptor-positive (ER+) breast cancer cell model, the transferring tumor cells result in lower metastasis, enhanced survival and decreased tumor infiltration of immunosuppressive cells in B7-H4 knockout compared to wildtype mice [
19]. These results suggest that the expression of B7-H4 within the breast TIME by infiltrating immune cells is hypothesized to promote immune evasion. Epithelial-to-mesenchymal transition (EMT) is a typical embryonic development process and has long been linked to increasing invasiveness, favoring escape from the primary tumor site and thus metastasis [
20]. In addition to immune invasion and chemoresistance, invasiveness and cancer cell stemness are the most critical properties of tumor development in EMT processes. Previous reports show that PD-L1 production by coordinating EMT processes modulates immunoresistance toward CTLs [
21,
22]. These findings suggest that EMT processes, including invasiveness and stemness, might manipulate switching PD-L1 expression and govern immunotherapeutic responses in human breast cancer cells.
Most studies have focused on B7-H4 expression and its regulatory mechanism in immune cells. In the present study, we surprisingly found a dynamic downregulation of B7-H4 expression when tumor cells escaped from cytotoxicity of B7-H4 CAR-T cells, which is similar to the tendency of B7-H4 expression from stage I to stage II in human breast cancer patients. We further demonstrate that the absence of B7-H4 expression in breast cancer cells promotes EMT and stem cell differentiation, resulting in tumor migration, metastasis and chemoresistance. Our study presents new insight into the precise strategy of adoptive T cell therapy targeting breast cancer cells at different stages of tumor progression.
Methods
Human cells and cell culture
Human breast cancer cell lines were purchased from the American Type Culture Collection (ATCC), including SK-BR-3(SKBR3), MCF-7 (MCF7), MDA-MB-231, MDA-MB-436, MDA-MB-468, T-47D and BT-474. Tumor cells were cultured in DMEM or RPMI-1640 medium (Corning, NY, USA) with 10% fetal bovine serum (FBS), penicillin (100 U/mL), streptomycin (100 μg/mL) (Gibco, NY, USA), and cell lines were regularly tested for mycoplasma negative. All cells were incubated at 37 °C in a humidified chamber containing 5% CO2.
The construction of tumor cell lines with B7-H4 gene knockout or overexpression
B7-H4 gene knockout cells (SKBR3-KO) were generated using the CRISPR/Cas9 system by infecting a custom-made lentivirus vector containing the Cas9 gene and sgRNA targeting the sequence of human B7-H4, and then, single cell clones were isolated by serial dilutions. B7-H4-overexpressing tumor cells (MDA-MB-231-OE and MCF7-OE) were prepared by transfection of the pcDNA-hB7-H4 plasmid encoding full-length human B7-H4 into MDA-MB-231 and MCF7 cells. G418 (Corning, NY, USA) was used in antibiotic selection to obtain stable clones.
Immunoblotting
Immunoblotting was performed using standard techniques: Cell lysates were produced in RIPA lysis buffer (Abcam, MA, USA) supplemented with protease/phosphatase inhibitor (Cell Signaling Technology, USA). The protein concentration was quantified using a BCA protein assay kit (Beyotime Biotechnology, Shanghai, China). Lysates were separated by SDS-PAGE and then transferred onto PVDF membranes (Millipore, MA, USA). After blocking with the 5% milk blocking buffer, the membranes were incubated overnight at 4 °C with the primary antibodies: rabbit anti-human B7-H4 (clone D1M8I, #14572S) and anti-β-actin (clone13E5, #4970S). After incubated with secondary anti-rabbit antibodies conjugated to horseradish peroxidase (HRP) (1:10,000), the membranes were detected using enhanced chemiluminescent (ECL) HRP substrate (Vazyme, Nanjing, China). All the antibodies were purchased from Cell Signaling Technology (CST, MA, USA).
Flow cytometry
Tumor cells were resuspended in 100 μl staining buffer after blocking Fc receptors, then incubated with APC-conjugated anti-B7-H4 (cloneMIH43, #358,108) or isotype control (BioLegend, CA, USA) for 30 min. After washing and centrifuging in staining buffer, cells were fixed and permeabilized, and then performed intracellular staining of PE-conjugated anti-Ki67 (BioLegend, CA, USA). After staining, cells were evaluated using BD FACS Verse and analyzed using FlowJo 10 (BD Biosciences, CA, USA).
Proliferation assay and cell cycle analysis
SKBR3-WT and -KO cells were plated in 96-well plates at 1000 cells/well. Cell growth rates were monitored using a Cell Counter (BioRad, CA, USA) for 7 days. For cell cycle analysis, cells were grown to 70–80% confluency, treated with RNase A, stained with propidium iodide (Sigma-Aldrich, MO, USA) and then subjected to flow cytometry with BD FACS Verse, and cell cycle distribution was calculated using the FlowJo 10. Immunoblotting was used to analyze the expression levels of cell cycle regulators in SKBR3-KO and SKBR3-WT. Rabbit anti-human primary antibodies were obtained from Cell Signaling Technology (CST, MA, USA), including anti-Cyclin D1(clone 92G2, #2978), anti-Cyclin E2(#4132), anti-CDK2(clone 78B2, #2546), anti-CDK4(clone D9G3E, #12,790), anti-p27(clone D69C12, #3686) and anti-GAPDH (clone D16H11, #5174) antibodies.
Preparation of B7-H4 CAR-T cells
High titers of the lentivirus vectors pMSCV-H4-CAR were generated by transient transfection and used to transduce HEK 293 T cells at MOI of 5. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation over Ficoll-Paque (GE Healthcare, Little Chalfont, UK) and activated with anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific, MA, USA) combined with 50 U/mL recombinant human IL-2 (Peprotech, NJ, USA) for 24 h, and then, PBMCs were cultured with the desired lentivirus at appropriate MOI for 24 h. The purity and phenotype of CAR-T cells were verified by flow cytometry.
T cell cytotoxicity assay
To analyze B7-H4 CAR-T cell cytotoxicity, 5 × 104 SKBR3 (WT and KO) cells were labeled with 3 μM carboxyfluorescein diacetate succinimidyl ester (CFSE, Thermo Fisher Scientific, MA, USA) as target cells and then incubated with B7-H4 CAR-T cells 24 h at various effector-to-target ratios. Killing effect was evaluated by a cell death marker (LIVE/DEAD® Fixable Dead Cell Stain Kits, Thermo Fisher Scientific, MA, USA) using flow cytometry. Live cell imaging and data analysis were performed using EVOS FL Auto microscope (Thermo Fisher Scientific, MA, USA).
Modified transwell migration assay induced by T cell cytotoxicity
The modified transwell migration assay was used to identify and characterize tumor cell escape from T cell cytotoxicity by migration. SKBR3 cells (5 × 104) were labeled with 3 μM CFSE as target cells and then incubated with B7-H4 CAR-T cells or activated T cells for 24 h in the upper chamber of an 8.0-μm transwell (Corning, NY, USA). Cell culture media containing 10% FBS was placed in the bottom well of the lower chamber in a 24-well plate. We took photographs of 5 fields randomly and counted the number of CFSE labeled tumor cells migrated to the lower chamber, and then, we collected cells and analyzed the all samples by flow cytometry. Tumor cells migrated from the upper to the lower were collected and analyzed B7-H4 expression.
Wound healing assay
Cells were seeded in 24-well plates and cultured in a complete medium overnight. When the confluence reached 90%, cells were scraped across the well gently with a micropipette tip, and floating cells were washed away with PBS. Cell migration was monitored at 0 h, 24 h and 48 h using a microscope, and distance between the two edges of the wound was calculated.
Transwell invasion assay
2 × 104 cells were seeded in the upper side chamber of an 8.0-μm transwell in a 24-well culture plate (Corning, NY, USA) with the upper membrane covered by Matrigel (BD Biosciences, CA, USA). The lower chamber comprised DMEM containing 10% FBS and then we took photographs of 5 fields randomly and counted the number of cells passing through the matrigel-coated membrane after the samples were fixed and stained with 0.4% trypan blue solution (Thermo Fisher Scientific, MA, USA) to evaluate the invasiveness of cells on the membrane.
Quantitative RT-PCR
Total RNA isolation from cells was performed using RNA Isolating Kit (Vazyme, Nanjing, China), and cDNA was synthesized using PrimeScript RT Master Mix (TaKaRa, Tokyo, Japan) following the manufacturer’s protocol. The following qPCR procedure was initiated by 3 min at 95 °C, and followed by 40 cycles of 15 s at 95 °C, 30 s at 55 °C, and 30 s at 72 °C. The expression levels of the target genes were calibrated by the internal control. The CT calculated quantification of all the samples by the software and relative fold changes were calculated using the 2−ΔΔCT.
RNA-Seq
Total RNA was extracted using TRIzol (Thermo Fisher Scientific, MA, USA) reagent. Expression values of EMT and stemness-associated genes and transcription factors were evaluated from the transcriptome RNA sequencing data of SKBR3 (WT and KO), MDA-MB-231(WT and OE) and MCF7 (WT and OE) cell samples. The unit used to establish the fold change in RNA-seq analysis is TPM. We compared the TPM value of each gene’s expression between two cell lines (KO and WT, or WT and OE) in the RNA-seq analysis. Then, the ratio of gene expression levels between the two samples was calculated to establish the fold change. Log2 scale is represented as the logarithm base 2 of the fold change.
Mouse xenograft model
8-week-old adult male BABL/c nude mice were purchased from Beijing Vital River (Beijing, China) and housed in pathogen-free facilities in the Experimental Animal Centre of Fujian Medical University. SKBR3 (WT and KO), MDA-MB-231 (WT and OE) and MCF7 (WT and OE) cells were embedded in matrigel (Corning, NY, USA) and then subcutaneously injected into the right flank at limiting dilutions. Tumor length and width were measured with a caliper every other day to calculate tumor diameter using the equation (length + width)/2. Mice were killed when tumors diameter reached the maximum allowed size (15 mm) or when signs of ulceration were evident. For the lung metastasis tumor model, the mice were randomized in groups, and tumor cells were resuspended in 100 μl of PBS and injected into the tail vein to generate metastasis in nude mice. The mice were euthanized 4 weeks after the injection of cancer cells, and lungs were collected. All procedures performed in studies involving animals were approved by the Fujian Medical University Institutional Animal Care and Use Committee (IACUC) in accordance with the ethical standards.
HE and IHC staining
HE staining of lung tissue was conducted to evaluate the metastasis of cancer cells within the lung. The lung tissues were fixed with 4% paraformaldehyde overnight, then embedded in paraffin and sliced into 8 μm thick sections. After dewaxed with xylene, the dehydrated with increasing ethanol concentration, slides were stained with HE solution. For Immunohistochemistry (IHC) staining, human breast cancer tissues were fixed and paraffin-embedded according to the manufacturer’s instructions. The sections were immunolabeled with mouse anti-B7-H4 antibodies (clone 6H3, in house), followed by anti-mouse peroxidase kit (Vector, CA, USA) labeling, respectively, according to the manufacturer’s instructions. The slides were photographed under an optical microscope. Expression levels for B7-H4 were scored semiquantitatively based on staining intensity (SI) and percentage of positive cells (PP). An immunoreactive score (IRS) was calculated using the following formula: IRS = PP × SI [
23]. All procedures performed in studies involving human tissues were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Immunofluorescence
Lung tissue sections or cells growing on the glass slide (20,000 cells/slide) were fixed with 4% paraformaldehyde, permeabilized by 0.1% Triton and stained with the indicated concentrations of rabbit anti-human primary antibodies, including anti-E-cad (clone 24E10, #3195), anti-N-cad (clone D4R1H, #13116 T), anti-H3K27me3 (clone C36B11, #9733 T), and anti-H3K9me3 (clone D4W1U, #13969 T), followed by Alexa Fluor-488-conjugated donkey anti-rabbit secondary antibody (1:1000). All the antibodies were obtained from Cell Signaling Technology (CST, MA, USA). The antibody-labeled cells on coverslips were stained with DAPI (1:1000, Sigma-Aldrich, MO, USA) for 2 min. Cells were observed under an inverted microscope using 405 and 488 nm lasers to visualize nuclei and target protein expression, respectively. Then, we randomly selected photographs of 4 fields for fluorescence quantitative analysis to evaluate the relative optical density by Image J.
MTT assay for cytotoxicity
Breast cancer cells were seeded into 96-well plates (10,000 cells/well) and cultured in complete medium for 24 h and then treated with different concentrations (0–500 μM) of chemotherapy agents, including Doxorubicin (Dox) (MCE, NJ, USA), Gefitinib (Gef) (Sigma-Aldrich, MO, USA), Oxaliplatin (Oxa) (Sigma-Aldrich, MO, USA) and Fluorouracil (5-FU) (Sigma-Aldrich, MO, USA). After another 20 h incubation, 10 μl MTT was added to each well for 4 h at 37 °C. The formazan produced by viable cells was dissolved with DMSO (Sigma-Aldrich, MO, USA) and then measured the absorbance at 570 nm.
500 cells were seeded in 96-well ultra-low attachment plates (Corning, NY, USA) in a serum-free tumorsphere culturing medium containing epidermal growth factor (20 ng/ml), basic fibroblast growth factor (20 ng/ml) and B27 supplements. Tumor sphere formation was monitored using a microscope. Total RNA was isolated from the spheres on the 14th day for gene expression analysis by real-time PCR.
Acquisition of gene expression profiles from TCGA datasets
The RNA-seq data and clinical information of breast cancer samples (TCGA BRCA) were collected and downloaded through UCSC Xena and TGCA real-time hub. The gene expression analysis in normal breast tissues was collected using the Genome Tissue Expression (GTEx) portal.
GSEA (gene set enrichment analysis)
GSEA was performed on various gene signatures by comparing gene sets from the Molecular Signature Database (MSigDB) database or published gene signatures. Gene sets with a false discovery rate (FDR) value < 0.25,
p value < 0.05, and ∣normalized enrichment score (NES) ∣ ≥ 1 were considered statistically significant [
24,
25].
Statistical analysis
All experiments were performed three or more times independently under similar conditions. Results are shown as the means, and the error bars represent the standard error of the mean (SEM), unless stated otherwise. GraphPad Prism 9 was used to generate graphs and perform statistical analysis. P values were calculated by unpaired two-tailed Student’s t test, one-way ANOVA and chi-squared test. Statistically significant results are represented as follows: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.
Discussion
Over the past few years, immune checkpoint inhibitor (ICI) targeting the PD-1/PD-L1 pathway has shown promising therapeutic effects in many cancers and has fundamentally changed the paradigm for the clinical management of cancer patients. Unfortunately, some patients fail to respond to PD-1/PD-L1 antibodies. Further study on the mechanism of other immune molecules is expected to develop new immunotherapy strategies. B7-H4, a member of the B7 family of immunoregulatory proteins, inhibits T cell proliferation and cytokine production. Notably, the B7-H4 protein is highly expressed in several tumors, including breast cancer, in combination with its low or absent protein expression in normal tissues, suggesting that B7-H4 is an attractive immunotherapeutic target for ICI’s [
35]. The expression levels of B7-H4 have been reported to be associated with immunosuppression in the tumor microenvironment, tumor progression and poor prognosis, including urinary tract urothelial carcinoma, cervical cancer, colorectal cancer, lung cancer, ovarian cancer and breast cancer [
16,
26,
36‐
39]. A mutually exclusive pattern of B7-H4 with PD-L1 expression has been demonstrated in some types of cancer, such as breast cancers, lung cancer and glioma [
40‐
42], suggesting that some types of cancers might be preferentially undergoing B7-related immune evasion pathway.
Although the receptor on the T cell of B7-H4 remains unknown and controversial, it is generally known that cell surface-associated B7-H4 inhibits T cell activation and leads to a decrease in the number of CD4 and CD8 T cells [
43,
44]. Our previous study also illustrated that high B7-H4 on breast cancer cells negatively correlated with CD8 T cell infiltration in tumor sites [
15]. However, it is still unclear whether CTL could affect B7-H4 expression in tumor cells, contributing to the inverse correlation between B7-H4 and CD8 T cells. Our study shows that B7-H4 is downregulated in breast cancer cells after co-culture with CAR-T and the active T cells (Additional file
7). In addition, the absence of B7-H4 leads to an increase in breast cancer cell proliferation, migration and metastasis by promoting the progression of the cell cycle. Furthermore, B7-H4 can intervene in human breast cancer stem cell differentiation, EMT and chemoresistance.
Protection from CTL-mediated lysis has been reported to be associated with breast cancer cells undergoing EMT, wherein silencing of the WNT pathway coincides with hyperactivity of TGF-β signaling, EMT and acquisition of stemness properties [
45]. Previous reports indicate that B7-H4 induces EMT and promotes cancer cell proliferation, invasion and stemness by using the temporary gene knockdown of cancer cell lines [
38,
46‐
48]. Nevertheless, our results demonstrate that T cells-mediated downregulation of B7-H4 and B7-H4-KO breast cancer cells exhibits increased cell proliferation, migration and invasion. Cancer cells undergoing EMT processes implicate the loss of cell polarity and lead to the transformation of the mesenchymal phenotype, accompanied by regulation of E-cadherin and the upregulation of vimentin and/or N-cadherin. Meanwhile, the critical transcription factors, including
TWIST1,
TWIST2,
ZEB1,
ZEB2,
SNAI1 and
SNAI2, participate in the EMT process, leading to cancer cell migration and invasion. Our results from RNA-seq analysis of these EMT- and CSC-related genes between different expression levels of B7-H4 breast cancer cells (KO vs. WT and WT vs. OE) further confirmed that B7-H4 deficiency promotes breast cancer cell growth, EMT and stemness.
Emerging evidence suggests a phenotypic and molecular signal pathway association between EMT and chemoresistance in several cancers, including breast cancer. In addition, tumor cells acquire stemness through the EMT process, secondary tumor-initiating and chemoresistance properties [
29,
49]. The low expression of
CD24 and high expression of aldehyde dehydrogenase 1 family member A1 (
ALDH1A1) may define subsets of breast cancer stem cells with enhanced migratory potential [
50]. Our results further confirm the low expression of
CD24, high expression of
ALDH1A1 and stem cell self-renew genes in B7-H4-KO breast cancer cells. In addition, the B7-H4 deficiency facilitates breast cancer cell growth, EMT and chemoresistance, with the opposite effects observed in the B7-H4 overexpression of breast cancer cells, including downregulation of migration and EMT critical genes such as vimentin and N-cadherin. Chemoresistance is a tremendous challenge in the progression and recurrence of breast cancer patients. We thus determined whether B7-H4 dysregulation may induce chemoresistance in breast cancer cells by evaluating the IC50 values for Doxorubicin [
30], Oxaliplatin [
31,
32], Fluorouracil [
33] and Gefitinib [
34], four commonly used chemotherapy in breast cancer patients. The significantly difference of IC50 was confirmed in SKBR3-KO or MDA-MB-231-OE cells, but not be seen in MCF7-OE cells (data not shown). These findings suggest that the loss of B7-H4 plays a crucial role in legitimizing breast cancer stem cell differentiation and EMT process and might at least in part mediate chemoresistance.
Epigenetic dysregulation is commonly observed in carcinogenesis and tumor progression and histone methylation is a critical epigenetic machinery in regulating cell gene transcription [
51]. We have noticed in our transcriptomics analysis that B7-H4 knockout leads to a profound and broad influence on global gene transcription in SKBR3-KO cell lines, which prompted us to investigate whether B7-H4 may affect gene expression in breast cancer cells by epigenetic reprogramming. The enrichment analysis of differentially expressed genes showed that genes related to the upregulation of H3K27me3 were overrepresented in the B7-H4 KO cell line. Thus, we determined several candidate genes involved in DNA methylation and histone methylation and found a significant change in the expression of EZH2, the enzyme directly responsible for H3K27me3 generation after B7-H4 knockout. We also confirmed the difference in H3K27me3 levels between MDA-MB-231 WT and OE cells. These results support a potential link between B7-H4 dysregulation and histone methylation-mediated epigenetic reprogramming, as histone methylation is closely related to the subtyping and grading of breast cancers [
52,
53]. The expression level of H3K27me3 has been shown to affect the expression of genes associated with EMT and stem cell potential to tolerate chemotherapy in breast cancer, leading to enhanced cancer cells metastasis and chemoresistance [
54‐
56]. Given the molecular evidence from experiments and literature, we think that EZH2-H3K27me3 could be the possible regulators in mediating B7-H4-induced transcriptional alteration. Our results herein demonstrate that B7-H4 deficiency promotes H3K27me3 but H3K9me3, resulting in enhanced breast cancer stem cells and the potential for epigenetic reprogramming.
To better elucidate our surprisingly novel findings into the role of B7-H4 in breast cancer cells, the tumor microenvironment involved with immune cells should certainly be considered in factual circumstances. We analyzed the expression levels of B7-H4 using the TCGA BRCA database profiling 932 breast tumor patients. The data demonstrate that the expression levels of B7-H4 are significantly downregulated from AJCC stage II to stage III. Our results further confirm the inverse expression levels of B7-H4 with EMT- and CSC-related genes in a group of breast cancer patients. The results strongly suggest that downregulation of B7-H4 leads to breast tumor growth and an invasion of the nearby lymph nodes undergoing the EMT process but also indicate that the breast tumor immune microenvironment plays a critical role in the impairment of immune system responses in developing tumor progression. Considering the widespread expression of B7-H4 in different cancer types and low expression in normal tissues, B7-H4 is a reasonable target for the blockades, in which monotherapy such as CD3 bispecific antibody (BsAb) has proven effective in mouse models [
57]. However, based on our findings, we should be carefully considered for application on patients of different breast cancer stages when using CAR-T or BsAb target to B7-H4 for destruction by interfering tumor immune microenvironment and incorporating with appropriate chemotherapy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.