Background
Ovarian cancer is a malignant tumor that seriously threatens women’s health; and of its many subtypes, high-grade serous ovarian carcinoma (HGSOC) is the most common and lethal one. The American Cancer Society (ACS) statistics reported that the new incidence of ovarian cancer accounted for 3% of tumors among women in 2019, and its mortality rate ranking fifth [
1]. Despite continuously improving living standards, surgical and chemotherapeutic techniques in the recent several decades, the overall survival rate of ovarian cancer has not improved significantly, and the 5-year survival rate is only about 30% [
2,
3]. At present, the most effective adjuvant therapy for HGSOC is the platinum and paclitaxel combination-based chemotherapy. However, 70% of patients relapse after primary therapy as a result of chemotherapy insensitivity or resistance [
2,
3].
DNA damage commonly occurs in cell due to exogenous and endogenous stressors, while cells also have consequently evolved a rounded DNA damage response (DDR) system which encompasses complex signal pathways to remove or tolerate DNA damage [
4] such as mismatch repair (MMR), non-homologous end joining (NHEJ) and homologous recombination (HR), etc. DDR defect creates vulnerability in specific cancer cells which provides us with potentially effective anticancer therapies [
5]. It is well known that homologous recombination repair (HRR) pathway alteration is important characteristics in HGSOC. In particular, approximately 35% of HGSOCs carry BRCA1 or BRCA2 (BRCA1/2) mutation and other 6–10% of cases harbor germline or somatic mutations of at least ten other genes in the HR pathway, such as ATM, CHEK1, CHEK2, etc. [
6‐
8]. The poly (ADPribose) polymerase inhibitors (PARPi), a compensatory DDR pathway inhibition, have been the most promising target therapy for HGSOC based on a conception of synthetic lethality [
9,
10]. The SOLO1 study showed that Olaparib, the most commonly used PARPi, could significantly benefit the patients with newly diagnosed advanced ovarian cancer and BRCA1/2 mutation by reducing the 70% risk of disease progression or death than placebo group (hazard ratio, 0.30; 95% CI 0.23 to 0.41;
P < 0.001) [
11]. Based on this result, in December 2018, the Food and Drug Administration (FDA) of USA has approved Olaparib as the first-line maintenance therapy in adult patients with advanced epithelial ovarian cancer, fallopian tube cancer, or primary peritoneal cancer who exhibited complete or partial clinical response after platinum-based chemotherapy and harbored deleterious or suspected deleterious germline or somatic BRCA mutations. Although platinum response was used to be a feasible indication for PARPi application, HR-related gene mutations (mainly BRCA1/2 mutation) remain the most convincing genetic indicators for response to PARPi. However, there were still 40–70% patients with BRCA1/2 mutations who failed to respond to PARPi [
12‐
14]. Furthermore, HR deficiency (HRD) genomic scar detection was developed and used in guiding PARPi therapy. But the NOVA and Arial 3 study revealed that HRD negative patients could also benefit from PARPi [
15,
16]. More specific and accurate methods to predict the response to PARPi and overcome the resistance warrant further investigation.
In this study, we firstly identified miRNA-509-3 as a prognostic and platinum-sensitive factor through bioinformatic analysis of miRNA data on TCGA database. Meanwhile, we established a patient-derived-xenograft (PDX) model to reveal that miR-509-3 plays an essential role in HR pathway regulation and tumor suppression and therefore increases the sensitivity to PARPi therapy in HGSOC. Furthermore, we detected the HR-related gene mutation and RAD51 expression level of each PDX parent tissue through whole-exome sequencing (WES) in order to investigate their correlation with Olaparib response.
Materials and methods
Raw data was downloaded from TCGA website including the miRNA expression profile of each case and the corresponding clinical data. According to NCCN Guidelines for Ovarian Cancer (2019, V1), the platinum response status were judged, which divided the cohort into platinum-sensitive (P-sen) group or platinum-resistant (P-res) group. The differential analysis on miRNA expression between the two groups was performed using the DEGseq R package [
17].
P value was adjusted using
q value [
18].
Q value < 0.05 and log2 fold change absolute value > 1 were set as the threshold for significantly differential expression by default.
Tissue samples and clinical data
A total of 126 HGSOC FFPE (formalin-fixed, paraffin-embedded) samples with detailed clinical data were collected from pathology department of Qilu Hospital, Shandong University. All patients were followed up for at least 5 years. The patients were staged by FIGO Staging System (8th ed., 2017) and distinguished into P-sen and P-res groups. The complete clinical characteristic of these enrolled patients is reported in Additional file
6: Table S1. All samples were used based on the patients’ or their guardians’ informed consent. Ethical approval was obtained from the Ethics Committee of Shandong University.
RNA extraction and real-time quantitative PCR
AllPrep DNA/RNA FFPE Kit (QIAGEN) was used to extract the total RNA (including small RNAs) from the FFPE tissue sections. As for the cultured cells, total RNA was extracted by TRIzol reagent (Ambion) following the manufacturer’s protocol. The miRNA and mRNA were reverse-transcribed using One Step PrimeScript miRNA cDNA Synthesis Kit (Takara) and PrimeScript RT Reagent Kit (Takara) respectively. The cDNA were used as templates for real-time quantitative PCR (RT-qPCR), which was performed using SYBR Green qPCR master mix (Takara).
Cell lines and cell culture
Human ovarian cancer cell line UWB1.289 (BRCA1-null) was purchased from American Type Culture Collection (ATCC). A2780, HEY, and HEK293T cell line was obtained from the Chinese Academy of Sciences (Shanghai, China). UWB1.289 and A2780 were cultured by RPMI 1640 medium (GIBCO) with 10% fetal bovine serum (FBS). HEY and HEK293T were cultured in DMEM (GIBCO) containing 10% FBS (BIOIND). All the cell lines were maintained at 37 °C with 5% CO2 in a humidified incubator.
Stable and transient transfection
Lentivirus expressing premiR-509-3 and corresponding negative control (NC) were purchased from Genechem (Shanghai, China). Further, 1 × 105 cells were plated into 6-well plates 24 h before stable transfection. The lentivirus was added into the culture medium with the multiplicity of infection (MOI) value ranging from 20 to 40. After 24 h, previous medium was replaced by fresh culture medium containing 2 μg/mL puromycin (Sigma-Aldrich) once every 2 days to obtain the stably transfected multiple colonies.
The specific small interfering RNA (siRNA) and negative control siRNA were synthesized by GenePharma (Shanghai, China) with the following sequences: HMGA2-si1 5′-CGCCAACGUUCGAUUUCUATT-3′, HMGA2-si2 5′-GGAAGAACGCGGUGUGUAATT-3′. The HMGA2 cDNA (in pEnter), RAD51 cDNA, and blank pEnter vector were purchased from Vigenebio (Shandong, China). Cells were transfected with Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer’s protocol and harvested after 24–48 h for the following assays.
Cell migration and invasion assays
The cells’ ability of migration and invasion was evaluated using the transwell technique which was performed in Cell Culture Insert (24-wells, 8.0 μm pore size, FALCON) with and without Matrigel Matrix (CORNING) respectively. Then, 1–1.5 × 105 cells resuspended in 200 μL FBS-free medium were seeded into the upper chambers of culture inserts and 700 μL culture medium containing 20% FBS was injected into the lower chambers as chemoattractants. The chambers were incubated at the 37 °C incubator for an appropriate time (6–24 h). Cells in the lower surface of chambers were fixed in methanol for 15 min, stained with 0.1% crystal violet for 20 min, and counted under a light microscope.
Cell viability and clonogenic assays
To examine the proliferation ability, each kind of cell was seeded in 96-well plates at densities of (0.8–1) × 103 cells per well in quintuplicate for 0–5 days. At specific time points, 20 μL 5 mg/ml MTT (3-(4, 5)-dimethylthiahiazo (-z-y1)-3,5-diphenytetrazoliumromide, Sigma-Aldrich) was added into each well and the plate was incubated in the 37 °C incubator. After 5 h, the supernatants in wells were carefully discarded and 100 μL DMSO (Sangon Biotech) was added into each well. The absorbance value at 490 nm was evaluated by Varioskan Flash microplate reader (Thermo Scientific).
For the colony formation assays, 600–800 single cells in 10%-FBS culture medium were seeded onto 6-well plates and incubated at 37 °C for 10–14 days. The resulting colonies were fixed with methanol and stained with 0.1% crystal violet. Colony formation ability was quantified by counting the surviving colonies containing more than 50 cells.
Cell cycle assay
The transiently transfected cells with miRNA mimics or negative control (both synthesized by GenePharma) were harvested and stained with propidium iodide according to the manufacturer’s protocol (MultiSciences, China). The treated cells were analyzed using a flow cytometer (FACSCalibur, BD, USA). The results were analyzed through Modifit LT software.
Drug resistance assay
Drug resistance assay was measured with MTT method as mentioned above. Cells were seeded onto 96-well plates (2000–3000 cells/well) and exposed to cisplatin (S1166, Selleckchem, Houston, TX, USA) or Olaparib (AZD2281, Selleckchem, Houston, TX, USA) at various final concentrations for 24–36 h (cisplatin) or 48–60 h (Olaparib). The final viability was estimated using the MTT reagent and surviving fractions were calculated.
Luciferase reporter assay
The 3′ untranslated regions (3′UTR) of HMGA2 or RAD51, potential target genes of miR-509-3, were synthesized by GenePharma (Shanghai, China) and inserted into pmirGLO vector (Promega) at the SacI and XhoI sites as the wild-type constructs. Mutant constructs were generated by overlap extension PCR. HEK293T cells were seeded in 96-well plate (2–3 × 104/well) and co-transfected with 50 ng constructs and 0.5 pmol miR-509-3 mimics or negative control with Lipofectamine 2000 reagent. After 36 h of transfection, luciferase activity was measured using the Dual-Glo Luciferase Assay System (Promega).
The HEY and UWB1.289 cell lines overexpressing miR-509-3 or corresponding negative control (NC) were used in these assays. For tumor metastasis assay, 2 × 106 cells resuspended in 200 μL PBS were injected into the peritoneal cavity of 4–5-week-old female athymic BALB/c nude mice (NBRI of Nanjing University, China). After 2–3 weeks, these mice were sacrificed and examined for the growth and metastasis of the tumors in peritoneal cavity.
For drug resistance assay in vivo, paired tumor cells (5 × 106 cells in 150 μL PBS per site) were injected subcutaneously into opposite sides of the axilla of ten 4-week-old nude mice. Two weeks after tumor cell injection, mice were randomly separated into treated group (n = 5) which were treated with olaparib (50 mg/kg) intraperitoneally and untreated group (n = 5) with DMSO dilution injection. After treatment once a day for 2 weeks, mice were sacrificed and their tumors were harvested and photographed. Tumors were weighed to assess the growth of each group. All the animal experiments were performed with the approval of Shandong University Animal Care and Use Committee.
Immunofluorescence assay
Cells (8 × 104) were seeded onto 24-well glass bottom plate (Cellivis) and treated with irradiation (Precision X-ray Inc.) of 4 Gy to induce DNA damage. Cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.1% TritonX-100, blocked with normal goat serum for 30 min at room temperature, and incubated with the RAD51 primary antibody overnight at 4 °C. The next day, the cells were incubated with secondary antibody, followed by counterstaining with DAPI. The fluorescence images were captured using the Opera Phenix™ High Content Screening System (PerkinElmer).
Western blot analysis
Cultured cells were lysed in RIPA Lysis Buffer (Beyotin, China) with 1% PMSF and 1% NaF by incubating on ice for half an hour. The supernatant were obtained by centrifugation and the protein concentration was determined by BCA Assay Kit (Thermo Scientific). A total of 30 μg protein samples per well were separated by SDS-PAGE (5.5% stacking gel and 11% separation gel), transferred to PVDF membranes (Millipore, ISEQ00010) by BIO-RAD Trans-blot (15 V, 90 min), and blocked in 5% non-fat milk solution at 25 °C for 2 h. The membrane was incubated overnight at 4 °C in the diluted primary antibody and then rinsed with TBST before incubating with the appropriate horseradish peroxidase-linked secondary antibodies for 1.5 h at 25 °C. The bands signal was detected using enhanced chemiluminescence (ECL) detection kit (PerkinElmer) by Image Quant LAS 4000 (GE Healthcare Life Sciences). β-Actin was detected as the endogenous control.
Antibodies and reagents
Antibodies for EMT, cell cycle, and DNA damage repair pathway were purchased as follows: ZEB-1 (CST:3396), N-CAD (CST:13116), E-CAD (CST:3195), Slug (CST:9585), Snail (CST:3879), Vimentin (CST:5741), β-catenin (CST:8480), CCND1 (CST:92G2), p21 (CST:2947 s), CDK4 (Abcam: Ab108357), CDK6 (Abcam:Ab124821), total ATM (CST:2873), phospho-ATM (Ser1981) (Abcam:5883), phospho-Chk2 (Thr68) (CST:2197), phospho-H2AX (Ser139) (CST:9718), RAD51 (Abcam:ab133534), and β-actin (CST:3700). All the western blot primary antibodies were 1:1000 diluted except RAD51 (1:10000 diluted). The antibody of HMGA2 (Proteintech, 20795-1-AP) was used in Immunohistochemistry (IHC).
PDX model establishment and therapy assay
Four- to six-week-old male NCG (NOD-Prkdcem26Cd52IL2rgem26Cd22/Gpt) mice were purchased from NBRI (Nanjing Biomedical Research Institute of Nanjing University, China). All NCG mice were housed in SPF room. The fresh primary ovarian carcinoma tissues were obtained from Department of Obstetrics and Gynecology, Qilu Hospital, Shandong University after acquiring the patients’ written informed consent. The collection of human tissue specimens and the PDX experimental procedures were approved by the Shandong University Animal Care and Use Committee.
Fresh patient tissues labeled as passage 0 (P0) were obtained from debulking surgeries in Qilu hospital within half an hour. Small tumor pieces (~ 10 × 10 × 10 mm) were washed by PBS solution and cut into homogenate. The homogenate was suspended with isometric PBS solution and mixed with Matrigel Matrix (CORNING). Then the homogenate was injected subcutaneously into the lower dorsal flank or axilla of the NCG mice. When the tumor size reached to 10 × 10 × 10 mm (often about 1–2 months later) which was labeled as P1, they were cut out and xenografted subcutaneously into another recipient mice as P2.
P2 were harvest from the sacrificed mice when tumors reached to 10 × 10 × 10 mm and made into homogenate according to aforementioned protocol. Isometric homogenate was intraperitoneally injected into new recipient mice. After 6 weeks, mice were randomly separated into three groups: NC + DMSO, NC + Olaparib, and miR-509-3 + Olaparib. The miR-509-3 groups were treated via intraperitoneal administration with adeno-associated virus (AAV, synthesized by Genechem, Shanghai) overexpressing premiR-509-3 and the other two groups accepted the AAV-NC (Genechem, Shanghai) injection. One week later, the Olaparib groups were exposed to Olaparib by intraperitoneal injection (50 mg/kg/mouse) once a day for 2 weeks while the DMSO group accept DMSO dilution injection. When treatment was completed, mice were sacrificed and their tumors harvested. Peritoneal metastasis’ weights (represent the tumor burden) and locations were recorded and compared among different groups.
Co-immunoprecipitation
Cells were lysed in cell lysis buffer (P0013; Beyotime, Shanghai, China). For co-immunoprecipitation with cell extracts, rabbit anti-human HMGA2 polyclonal antibody (Proteintech, 20,795-1-AP), normal rabbit IgG (CST, Cell Signaling Technology) was added and incubated for 12 h at 4 °C. A mixture of equal amounts of protein G Agarose (Fast Flow, Beyotime Biotechnology) was added and incubated for another 2 h at 4 °C. Immunoprecipitates were washed five times with lysis buffer and then denatured by boiling in 1% SDS.
Immunohistochemistry
The fresh tumor tissues were fixed by formalin for at least 24 h and cut into 4-μm-thick sections. Xylene and ethanol were used to deparaffinize and rehydrate. Antigen retrieval was achieved in 98 °C EDTA buffer (pH = 8.0) for 15 min. Endogenous peroxidase and nonspecific binding were blocked with 3% hydrogen peroxide and goat serum respectively. The primary antibody anti-HMGA2 (1:1000 diluted) and anti-RAD51 (1:500 diluted) were incubated in humid chamber overnight at 4 °C. Expression was detected by I-View 3,3′-diaminobenzidine (DAB) staining detection.
Whole-exome sequencing and mutation annotation
Nine pairs of genomic DNA (from FFPE HGSOC tissues and corresponding normal myometrium as control) were extracted using a salting-out procedure. Whole-exome capture was performed by Sure Select Human All Exon 60 Mb kit (Agilent Technologies, Santa Clara, CA) according to the standard protocol. The captured template DNA fragments of the constructed libraries were sequenced through the Illumina HiSeqX sequencing system to generate 150-bp paired-end reads. The paired-end reads from whole-exome sequencing (WES) were first aligned to human reference genome (hg19) using BWA (version 0.7). PCR duplicates were then marked in the alignments using Picard tools (version 1.1). The somatic mutation calling was carried out by Var Scan (version v2.3.9). For each candidate somatic mutation site, the common variants in the 1000 genomes (MAF > 5%) and the intergenic or introgenic region variants were also excluded in the following analysis. The latent effects of the somatic SNP and insertion-deletion were annotated using ANNOVAR.
Statics analysis
GraphPad Prism 6 (GraphPad Software, USA) was mainly used in data analyzing. Chi-squared test was used to analyze the differences in clinical characteristics. Survival analysis was performed by Kaplan-Meier and Log-rank test. Multivariate survival analysis was achieved by Cox regression in SPSS 17.0 (SPSS Inc., Chicago, IL). Student’s t test and one-way ANOVA analysis were applied to determine the statistical differences among different groups. Data are represented by means ± standard deviation of at least three independent experiments. P < 0.05 was considered to be statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).
Discussion
Although BRCA1 and BRCA2 (BRCA1/2) mutation status were regarded as an important indicator in the application of PARPi in ovarian cancer patients [
10], there were still 40–70% of patients not responding expectedly to Olaparib as a result of drug resistance through multiple mechanisms such as restoration of BRCA1 or BRCA2 protein functionality by secondary mutations, BRCA1/2 promoter methylation reversion, BRCA1/2 hypomorph overexpression, PARP1 expression deficiency, drug efflux, or acquisition of new mutations in other DDR genes [
30‐
33]. That is to say simply that BRCA1/2 mutation detection cannot be an ideal indicator for predicting PARPi sensitivity. Aiming at this present condition, scholars have developed better HRD detection methods to predict the sensitivity to PARPi or prognosis of ovarian cancer patient, in which the HRR Gene Panel and HRD Genomic Scar are the most well-known. For example, a 30-gene panel was proved to have a positive impact on overall survival and platinum response [
7]. A 61-gene panel was developed to distinguish BRCA-like tumors from non-BRCA-like tumors. Meanwhile, “BRCAness profile” correlates with sensitivity to platinum and PARPi and identifies a group of sporadic patients with good prognosis [
34]. In view of the increasing complexity of mutation analysis of all HR pathway genes, the genomic scar analysis emerged. Genomic scar means the gain or loss of large chromosomal regions or even whole chromosomes and can be observed as copy number variations resulting from DNA damage repair failure, which is focusing on DDR pathways instead of DDR genes [
35]. Genomic scar analysis, which mainly consist of “My Choice” test from Myriad Genetics or “Foundation Focus” test from Foundation Medicine, is believed to be more efficient in identifying “BRCA-like” tumors and has thus been undergoing clinical trials to discriminate between HR-proficient and HR-deficient tumors [
35‐
37]. However, the HRD negative patients could also benefit from PARPi. A randomized, double-blind, phase 3 trial containing 553 enrolled patients revealed that Niraparib could both improved progression-free survival in the HRD-negative subgroup and HRD-positive subgroup in non-gBRCA cohort [
15]. In another phase 3 clinical trial, the investigators classified that in patients with BRCA wild-type carcinomas, a benefit was also seen with rucaparib in patients with both high-LOH and low-LOH carcinomas [
16]. In sight of the limitation of genomic analysis, currently, scholars have devised a platform for functionally profiling DNA repair in short-term patient-derived HGSOC organoids. Their research results indicated that a combination of genomic analysis and functional testing of DDR pathway allows for the identification of DNA repair inhibitor including Olaparib [
38]. In a pan-cancer research, the somatic alterations of 33 cancer types was systematically analyzed to provide a comprehensive view of DDR deficiency, which defined a “core DDR” gene set of 80 genes from 276 genes encompassing all major DNA repair pathways including HR core gene [
29]. Our DDR mutation analysis of nine PDX cases was based on this study. In the HR positive patients, 75% (3/4) of them (PDX1, PDX2, and PDX3) harboring BRCA1/2 or other HR core gene mutation could effectively response to Olaparib treatment. Additionally, the only case (PDX4) who carried LIG1 mutation (a non-HR core gene) exhibited resistance to Olaparib. Therefore, we proposed that HR core gene analysis could be a potentially effective means for Olaparib response prediction in clinical application. However, this proposal requires a larger size preclinical study to be confirmed.
RAD51, a kind of DNA recombinase, is a crucial downstream component in the HR repair pathway. When cells are exposed to stressors (especially irradiation), RAD51 is relocalized within the nucleus to form distinct foci, which are thought to represent assemblies of HR repair proteins at these sites [
39]. Therefore, targeting RAD51 could confer cells to be defective in HR repair and the sensitivity to DNA-damaging drugs including platinum and PARPi [
32,
40]. More importantly, several recent studies have revealed that functional assay of RAD51 nuclear foci can be a surrogate marker for HR repair functionality and thus be used to predict the response to PARPi in vitro [
41]. Cruz et al. found that elevated RAD51 nuclear foci were the only common feature in PDX and patient samples with primary or acquired PARPi resistance [
42]. Meanwhile, the RAD51 score could more predictively discriminate PARPi sensitivity between PARPi resistance in breast cancer PDXs than the conventional genomic test [
43]. However, in our research, 20% (1/5) PDX cases with low primary RAD51 positive rate were not sensitive to Olaparib while 25% (1/4) with high rate actually responded well to Olaparib, and the Olaparib-sensitive PDX cases all showed low RAD51 expression rate in P3 (after treatment) group. This exception may be rationalized by the fact that RAD51 status would be better represented when cells were challenged with stressors such as irradiation. Similarly, in another study using patient-derived organoids, researchers similarly demonstrated that Olaparib sensitivity correlated with the absence of RAD51 foci formation after irradiation exposure [
38]. Therefore, we were convinced that functional detection of RAD51 (often treated with irradiation, etc.) is still a promising method in predicting PARPi response.
MiRNAs have already been hopeful tools and targets for novel therapeutic approaches and showed application prospects in preclinical study. For example, the mimic of tumor suppressor miRNA miR-34 has been undergoing phase 1 clinical trials (NCT01829971) in cancer therapy and the antimiRs targeting miR-122 in phase 2 clinical trials (NCT01872936, NCT02031133, NCT02508090) for hepatitis [
44,
45]. Furthermore, miRNAs have been proven to play a definite role in the PARPi response by regulating the HR pathway, such as miR-9, miR-506, miR-223, miR-182, miR-96, miR-622, miR-493, etc. [
19,
20,
46‐
49]. For instance, Sun et al. reported that miR-9 mediated the downregulation of BRCA1 and impeded DNA damage repair to improve chemotherapeutic and PARPi efficacy [
19]. Additionally, Srinivasan et al. found that miR-223 was a negative regulator of the NHEJ DNA repair and represented a therapeutic pathway in BRCA1-deficient cancers [
49]. In this study, we explored the TCGA dataset and uncovered miRNA-509-3 as a novel miRNA which predicted a better response to platinum-based chemotherapy and longer survival in ovarian cancer patients. We determined that miR-509-3 was a definite tumor suppressor that could significantly impact ovarian cancer cell metastasis, proliferation, and cell cycle. A series of cellular functional assay illustrated that miR-509-3 enhanced sensitivity to Olaparib by directly regulating RAD51 and HMGA2-ATM axis and performing synthetic lethal effect. Consistent with the result of in vitro experiment, the PDX model tumor burden of three treatment groups illustrated that when combined with miR-509-3, Olaparib was more effective in reducing tumor burden in two PDX cases (PDX1 and PDX9) along with the decline in RAD51 positive rate. Surprisingly, in one case (PDX8), miR-509-3 could reverse the Olaparib insensitivity by downregulating RAD51 expression. What is more, we consider that miR-509-3 level is positively correlated to the sensitivity to Olaparib, which needs more samples in the future. Consequently, these provided us with a potential promising target in Olaparib treatment.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.