Novel common target genes for breast cancer and colorectal cancer: A mendelian randomization and spatial transcriptomics study

doi:10.21203/rs.3.rs-4992169/v1

Download PDF

Research Article

Novel common target genes for breast cancer and colorectal cancer: A mendelian randomization and spatial transcriptomics study

https://doi.org/10.21203/rs.3.rs-4992169/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Introduction: Breast and colorectal cancer are a major global public health problem. Breast cancer is one of the most common cancers worldwide. Colorectal cancer is the third most common cancer and the second most common cause of tumor death worldwide. Central memory T (TCM) cells are closely related to the development of tumors and important targets for immunotherapy. Therefore, identifying the common signaling molecules of these two diseases in TCM cells can improve our understanding of these diseases and lead to the development of therapies that can be effective for treating both.

Methods:Single-cell RNA (scRNA) data of breast cancer (GSE161529) and colorectal cancer (GSE222300) patients was downloaded from the GEO database. The data were normalized and dimension reduced, then different T cell subsets were identified and differential gene expression analysis of central memory CD 8+ T cells was conducted. Mendelian randomization analysis, reverse causality detection, and co-localization analysis was performed to explore the relationship between differentially-expressed genes and the disease. Quasi-temporal analysis and metabolic analysis was done using scRNA sequencing technology and further analysis of gene expression and metabolism in spatial transcriptomes. Finally, the degree of association between drug target genes was analyzed by protein-protein interaction (PPI) analysis.

Results: Our analysis identified four genes (ZFP36L2,CKS1B, PTTG1, and ITGAE) that were associated with risk of both breast and colorectal cancer. In the pseudotime analysis, we found that the expression levels of CKS1B and PTTG1 decreased over time (P <0.05) while ZFP36L2 and ITGAE increased over time (P <0.05). In the metabolic analysis, these four genes were closely associated with the cysteine and methionine metabolism pathways, which was corroborated in the spatial transcription analysis. Finally, the PPI analysis among the drug target genes identified an interaction between PTTG1 and CKS1Bgenes.

Conclusion:This study reports that the ZFP36L2, CKS1B, PTTG1, and ITGAE genes could potentially influence breast cancer and colorectal cancer development via TCM CD8+ T cells. These four genes are putative common markers for diagnosis, treatment, and monitoring tumor response to therapies.

Mendelian randomization

Breast cancer

Colorectal cancer

Single-Cell RNA Sequencing

Spatial transcriptomics

Breast cancer is one of the most common cancers[1], with 2.3 million new cases diagnosed in 2020, accounting for 12.5% of all cancer diagnoses, and causing 685,000 deaths. Projections indicate that this will grow to 3 million new annual cases and more than 1 million yearly deaths by 2040[2]. Genetic predisposition is a key risk factor for breast cancer and families have been identified with high risk of developing the disease[3]. Fluctuations in the levels of hormones that affect sexual development and reproduction such as estrogen, ages at first menstruation and menopause, and aspects of pregnancy and breastfeeding have also been identified to contribute to risk levels[4–8]. Lifestyle and environmental factors, including physical activity, obesity, diet, alcohol intake, and smoking, further influence breast cancer incidence and progression[9–13].

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer death globally[1]. In 2020, it accounted for nearly 1.93 million new cases and 940,000 deaths[1]. The incidence of CRC varies by region, with the highest numbers reported in China and the United States. While rates have stabilized or declined in developed countries, they are expected to rise in lower-income regions[14, 15]. Lifestyle factors such as diet, alcohol, meat consumption, smoking, obesity, and low intake of dietary fiber, vitamin D, and calcium are linked to CRC incidence and mortality[16]. Aging and genetics also play significant roles in CRC risk. Advances in CRC treatment, including endoscopic therapy, surgery, radiotherapy, ablation, chemotherapy, and immunotherapy, have improved patient survival. Colonoscopy has been instrumental in early detection and treatment, reducing mortality rates[17]. However, the growing incidence of CRC presents economic and public health challenges, making early prevention and diagnosis essential.

Breast cancer and CRC are major public health issues, with a shared risk profile, including family history[18]. Estrogen, while protective against CRC, can increase breast cancer risk[19]. The BRCA2 gene is overexpressed in CRC cells that are resistant to olaparib, which suggests that targeting it could improve treatment efficacy[20]. The AIB1 gene is associated with breast and other cancers, and has been found to promote CRC progression via the Notch signaling pathway[21]. These shared risks underscore the need for early diagnostic markers and new therapeutic targets for both cancers. T cells, particularly central memory T (TCM) cells, are pivotal in tumor suppression and immune therapy, with CD8 + T cells playing a crucial role in the anti-tumor response[22]. Enhancing early memory T cells can improve the efficacy of adoptive cancer immunotherapy, with CD8 + T cells exhibiting potent anti-tumor immunity[23].

We sought to identify shared druggable targets for breast cancer and CRC. We analyzed single-cell RNA (scRNA) sequencing data from cancerous and normal breast and CRC tissues from the GEO database. After isolating T cells and performing dimensionality reduction and clustering, we integrated the data to identify differentially expressed genes between CD8 + T cells and other cell types. These genes were linked to expression quantitative trait loci (eQTLs) and genome-wide association study (GWAS) data helped identify potential therapeutic targets. We validated these targets through reverse causation tests and Bayesian colocalization analysis, followed by single-cell time-series and metabolic analyses, and spatial transcriptomics to confirm metabolic profiles and gene distribution. Finally, protein-protein interaction (PPI) analyses and drug evaluations were performed to identify shared therapeutic targets for both cancers.

2.1 single-cell RNA sequencing

Single-cell RNA sequencing (scRNA) data from breast (GSE161529) and colorectal (GSE222300) cancer patients were sourced from the GEO database. Using the Seurat V4.0 R package, cells with gene counts between 200 and 5,000 and mitochondrial gene counts less than 2,000 were selected. The data were normalized and merged to minimize batch effects. Cellular clusters were identified with "FindNeighbors" and "FindClusters," and visualized using UMAP. The celldex package was used for cell annotation. Integrated scRNA data from cancer and normal tissues were analyzed for cell-cell communication using "cell-cell chat," projecting ligand-receptor pairs onto a PPI network. Differential expression analysis between CD8 + T cells and other cells was conducted, with results considered significant at P < 0.05.

2.2 Data source

The data are divided into exposure and outcome groups. For the exposure group, we used the eQTL derived from the differentially expressed genes between CD8 + T and other cells in the integrated single-cell data. Only eQTLs meeting the following criteria were included: (1) demonstrate genome-wide significant associations(P < 5× 10^− 8); (2) linkage disequilibrium (LD) clumping r² < 0.001; (3) F-statistics > 10.

The outcome data for breast cancer are derived from large-scale case-control GWAS statistics involving 46,785 breast cancer cases and 42,892 control cases of European descent. The outcome data for colorectal cancer come from the Finnish database, which includes 3,022 cases of colorectal cancer and 174,006 control cases of European ancestry. These data are publicly available at https://gwas.mrcieu.ac.uk/ and https://www.finngen.fi/en/access_results.

2.3 Mendelian randomization

If only one eQTL was available for a given gene, then the Wald ratio was used. When two or more genetic instruments were available, then inverse variance weighted Mendelian Randomization (MR-IVW) was applied. Heterogeneity was assessed using the Cochran Q test to identify genes associated with the risk of breast cancer and colorectal cancer.

A bidirectional Mendelian randomization analysis was conducted by treating the outcome group as the exposure group and vice versa. The effect estimation was performed using MR-IVW, MR-Egger, Weighted Median, Simple Patterns, and Weighted Patterns. Results were considered statistically significant at P < 0.05.

2.4 Bayesian colocalization analysis

Bayesian colocalization analysis is used to assess the probability that two traits share the same causal variant, employing the "coloc" package with default parameters to evaluate whether GWAS loci and eQTLs share the same causal variation. Bayesian colocalization provides the posterior probability for the five hypotheses of whether a single variant is shared between the two traits. In this study, we tested the posterior probability of hypothesis 3 (PPH3) where both protein and MS are associated to this region via different variants and hypothesis 4 (PPH 4) where both protein and MS are associated to this region via shared variants. Due to the limited effectiveness of colocalization analysis, the analysis is constrained to cases where the sum of PPH3 (the posterior probability that the GWAS significant signal is related to the eQTL expression, but not at the same locus) and PPH4 exceeds 0.8[24, 25], thus identifying shared therapeutic targets for breast and colorectal cancer.

2.5 Downstream analysis of single-cell RNA sequencing technology

Using single-cell RNA sequencing, we analyzed therapeutic target genes for breast and colorectal cancer identified through Mendelian randomization, filtering out genes expressed in fewer than five cells. Log-normalized single-cell data with pseudotime information were visualized. For positively and negatively expressed CD8_CM target genes, cell-cell communication analysis was conducted using Cell-Cell Chat. Ligand-receptor pairs were identified using subsetData parameters and projected onto the PPI network.

2.6 Spatial transcriptomics processing and annotation

Breast cancer data was obtained from a publically available database (https://www.10xgenomics.com/datasets/human-breast-cancer-targeted-immunology-panel-1-standard-1-2–0) and colorectal cancer data was sourced from a published database (https://aacrjournals.org/cancerdiscovery/article/12/1/134/675646/Spatiotemporal-Immune-Landscape-of-Colorectal). The Load10X_Spatial function was used to read and analyze the data for both breast cancer and CRC. Principal Component Analysis (PCA) was utilized to reduce the dimensionality of the integrated data, facilitating further downstream analysis in R (version 4.2.2), with the “RunUMAP” function applied to perform UMAP on these components.

Expression of therapeutic target genes in breast cancer and CRC across cell clusters was visualized using DotPlot, and cellular metabolism was analyzed using scMetabolism. Furthermore, parametric analysis was conducted in R to explore the correlations between metabolic scoring data at each spatial location and gene features and expression patterns in T cells[26]. The results from the deconvolution of cell types were further submitted to SPOTlight for visualization[27].

2.7 Protein-protein interaction analysis

The PPI Core Network (PPICN) was used to link compounds to disease-related protein molecules. Additionally, DrugBank Online (https://go.drugbank.com/) was used to evaluate the drugs that could target these genes.

3.1 Single-cell RNA sequencing results

The distribution of cells after normalization from colorectal cancer, breast cancer, and normal tissues is shown in Fig. 1A, where the cells are relatively dispersed. After filtering and dimensionality reduction, the distribution of cells became more concentrated, as shown in Fig. 1B. Cells from breast cancer, colorectal cancer, and normal tissues were then merged and aggregated, and visualized using the "UMAP" method to cluster each cell with similar expression together, resulting in 22 cell clusters (Fig. 1C).

Each cluster was then annotated using reference datasets from the celldex package, which includes human primary cell atlas data, identifying cell identities such as T cells, epithelial cells, endothelial cells, macrophages, fibroblasts, B cells, monocytes, tissue stem cells, and neutrophils (Fig. 1D). As T cells are a major component of the Tumor Microenvironment (TME) and play a crucial role in tumor suppression, further clustering analysis of T cells was performed, resulting in 14 distinct clusters (Figs. 2A and B). Based on the characteristic gene expression of the cells, manual annotations were made (Fig. 2C), categorizing T cells into CD8_EM, CD4_REG, CD8_CM, CD8_exhau, CD4_Naiv, TH17, and CD4_EM (Fig. 2D). The analysis then focused on cell-cell communication studies involving CD8_CM.

3.2 Interactions of CD8_CM with other cell types

Cell communication analysis found that the interaction pathways between CD8_CM and other T cell clusters (CD8_EM, CD4_REG, CD4_EM, CD4_Naive, TH17, and CD8_exhau) as well as epithelial cells, endothelial cells, macrophages, fibroblasts, B cells, monocytes, tissue stem cells, and neutrophils were similar between breast and colorectal cancers. However, in colorectal cancer, CD8_CM have a greater number of interaction pathways with these cell types compared to breast cancer. In breast cancer, CD8_CM primarily interacts with these cells through pathways such as MIF-(CD74 + CXCR4), CXCL13-CXCR3, and MIF-(CD74 + CD44) as shown in Figures S1C and D. In colorectal cancer, CD8_CM mainly engages through pathways including CCL5-CCR1, ANXA1-FRR1, MIF-(CD74 + CD44), and MIF-(CD74 + CXCR4), impacting the same classes of cells, as illustrated in Figures S1A and B.

3.3 Screening for risk genes in breast and colorectal cancers

We identified genes that were differentially expressed (P < 0.05) between CD8_CM and the above-mentioned cell types (Table S1). These genes were then converted into eQTLs, and significant eQTLs (P < 5×10^− 8) were selected (Table S2). A Mendelian randomization (MR) analysis was conducted for breast and colorectal cancers with a significance threshold of P < 0.05. The analysis identified 64 genes (NUCKS1, FOSL2, ITGAE, BIRC5, CENPM, CDKN3, UBE2S, CLNK, SMC4, CDC20, FASLG, STMN1, CENPF, TMPO ZWINT, CENPK, DLGAP5, MAP2K2, DUT, NASP, CCNB1, PTPN22, AHI1, SMC2, ANP32B, NUSAP1, KIF20B, RASGEF1B, CENPE, VPS37B, SAE1, NUF2, ASXL2, CCNA2, MKI67, PDE3B, ZFP36L2, NR4A2, SGO2, HMGB2, PTTG1, PCLAF, TMIGD2, TK1, LDLRAD4, ZEB2, ARL6IP1, CKS1B, UBE2C, AURKB, KPNA2, HMGB1, HMGN2, CD3E, CENPW, LAYN, MXD3, DNAJC9, TRGC2, TUBA1B, PCNA, ANP32E, SRSF7, and DEK) associated with the risk of breast cancer and seven genes (CKS1B, PTTG1, LCP1, HSP90AA1, ITGAE, PRF1, and ZFP36L2) associated with the risk of CRC (Fig. 3A and B). The genes associated with breast cancer risk are shown in Table 1 and Fig. 4A. The genes associated with the risk of colorectal cancer are shown in Table 2 and Fig. 4B. The genes analyzed in this study did not exhibit heterogeneity (Tables S3 and S4).

Table 1

Genes Associated with Breast Cancer Risk
Gene	OR	95% CI	P-value	Risk Association
NUCKS1	0.0892	0.0234–0.3400	3.99462×10^-4	Reduced Risk
ITGAE	0.0447	0.0403–0.0495	Approaches 0	Reduced Risk
BIRC5	0.0252	0.0233–0.0272	Approaches 0	Reduced Risk
CENPM	0.0112	0.0100–0.0124	Approaches 0	Reduced Risk
UBE2S	0.0099	0.0019–0.0517	4.307373×10^-8	Reduced Risk
CDC20	0.0001	0.0001–0.0001	Approaches 0	Reduced Risk
FASLG	0.0653	0.0591–0.0721	Approaches 0	Reduced Risk
STMN1	0.0367	0.0334–0.0402	Approaches 0	Reduced Risk
ZWINT	0.2073	0.1611–0.2668	2.172205×10^-34	Reduced Risk
CENPK	0.3936	0.3741–0.4141	2.08499×10^-283	Reduced Risk
DEK	0.0984	0.0629–0.1540	3.541778×10^-24	Reduced Risk
PCNA	0.0325	0.0288–0.0367	Approaches 0	Reduced Risk
CCNB1	0.0065	0.0058–0.0071	Approaches 0	Reduced Risk
AHI1	0.2741	0.2595–0.2896	2.228502×10^-5	Reduced Risk
ANP32B	0.0058	0.0052–0.0066	Approaches 0	Reduced Risk
RASGEF1B	0.0043	0.0030–0.0062	7.349116×10^189	Reduced Risk
VPS37B	0.4877	0.4502–0.5282	1.814115×10^-69	Reduced Risk
SAE1	0.0225	0.0169–0.0298	1.474272×10^-151	Reduced Risk
NUF2	0.0058	0.0047–0.0072	Approaches 0	Reduced Risk
CCNA2	0.001	0.0008–0.0011	Approaches 0	Reduced Risk
MKI67	0	0.00–0.00	Approaches 0	Reduced Risk
NR4A2	1464.3163	1250.5872-1714.5724	Approaches 0	Increased Risk
LDLRAD4	13.61	12.5485–14.7611	Approaches 0	Increased Risk
ZEB2	39610.9575	165.9823–9452984.1047	1.506391×10^-4	Increased Risk
ARL6IP1	0.0753	0.0716–0.0792	Approaches 0	Reduced Risk
AURKB	0.0159	0.0146–0.0174	Approaches 0	Reduced Risk
KPNA2	0.0006	0.0004–0.0007	Approaches 0	Reduced Risk
HMGB1	0.0106	0.0078–0.0145	4.381093×10^-183	Reduced Risk
HMGN2	0.0778	0.0682–0.0887	Approaches 0	Reduced Risk
CENPW	0.0002	0.0002–0.0003	Approaches 0	Reduced Risk
DNAJC9	0.0532	0.0161–0.1765	1.603169×10^-6	Reduced Risk
TRGC2	0.3183	0.2965–0.3413	1.452474×10^-223	Reduced Risk
FOSL2	5519.5871	4269.0160–7136.5022	Approaches 0	Increased Risk
CDKN3	272.5049	241.0733–308.0345	Approaches 0	Increased Risk
CLNK	23541.1161	18746.8445–29561.4629	Approaches 0	Increased Risk
SMC4	56.6641	52.3585–61.3237	Approaches 0	Increased Risk
SRSF7	979126.666	739095.3724–1297111.3929	Approaches 0	Increased Risk
CENPF	2.2175	1.7736–2.7724	2.775729×10^-12	Increased Risk
TMPO	797.1278	683.0948–930.1971	Approaches 0	Increased Risk
TUBA1B	3473.9359	2954.3501–4084.9019	Approaches 0	Increased Risk
DLGAP5	24349.4693	19882.9579–29819.3387	Approaches 0	Increased Risk
MAP2K2	23.9293	22.1831–25.8130	Approaches 0	Increased Risk
DUT	5.0387	4.2692–5.9470	1.576729×10^-81	Increased Risk
PTPN22	28.8905	27.0316–30.8772	Approaches 0	Increased Risk
SMC2	27.6446	25.9311–29.4713	Approaches 0	Increased Risk
ANP32B	0.0058	0.0052–0.0066	Approaches 0	Reduced Risk
NUSAP1	8.1568	4.3107–15.4344	1.115594×10^-10	Increased Risk
KIF20B	18.3976	17.2445–19.6278	Approaches 0	Increased Risk
CENPE	5609.514	4561.0927–6898.9273	Approaches 0	Increased Risk
ANP32E	14.6967	13.4874–16.0143	Approaches 0	Increased Risk
ASXL2	103.2294	87.7300–121.4671	Approaches 0	Increased Risk
ZFP36L2	10992.8283	9201.7136–13132.5837	Approaches 0	Increased Risk
SGO2	850.0735	722.9885–999.4972	Approaches 0	Increased Risk
HMGB2	422.228	325.3194–548.0046	Approaches 0	Increased Risk
PTTG1	83.366	69.6113–99.8385	Approaches 0	Increased Risk
PCLAF	6.208	5.4655–7.0514	1.187580×10^-173	Increased Risk
TMIGD2	10.0123	8.5656–11.7034	4.39975×10^-184	Increased Risk
MXD3	293.2167	11.8219–7272.5849	5.249580×10^-4	Increased Risk
CKS1B	17308.2749	13304.8564–22516.3182	Approaches 0	Increased Risk
UBE2C	6136.5452	5158.1841–7300.4738	Approaches 0	Increased Risk
LAYN	9.5281	7.6405–11.8821	4.356807×10^-89	Increased Risk
CD3E	90.1386	82.4799–98.5086	Approaches 0	Increased Risk
TK1	336.804	242.1629–468.4322	5.838693×10^-262	Increased Risk
NR4A2	1464.3163	1250.5872-1714.5724	Approaches 0	Increased Risk

Table 2

Genes Associated with Colorectal Cancer Risk
Gene	OR	95% CI	P-value	Risk Association
CKS1B	0.3363	0.1537–0.7360	0.006386	Reduced Risk
PTTG1	0.6654	0.4515–0.9806	0.039511	Reduced Risk
LCP1	0.8625	0.6975–0.9937	0.042347	Reduced Risk
HSP90AA1	1.2731	1.0782–1.5033	0.004406	Increased Risk
ITGAE	1.305	1.0275–1.6575	0.029074	Increased Risk
PRF1	1.368	1.0177–1.8389	0.037864	Increased Risk
ZFP36L2	2.2596	1.3801–3.6997	0.001192	Increased Risk

3.4 Sensitivity analysis of pathogenic genes

Reverse Mendelian randomization analysis revealed that SRSF7, TUBA1B, DEK, PCNA, and ANP32E have inverse causal relationships with breast cancer, while the remaining genes show no reverse causal relationship with breast cancer (Figure S2). We did not find any of the genes that we analyzed to exhibit a reverse causal relationship with CRC (Figure S3). Bayesian colocalization analysis indicated that 59 genes (NUCKS1, FOSL2, ITGAE, BIRC5, CENPM, CDKN3, UBE2S, CLNK, SMC4, CDC20, FASLG, STMN1, CENPF, TMPO, ZWINT, CENPK, DLGAP5, MAP2K2, DUT, NASP, CCNB1, PTPN22, AHI1, SMC2, ANP32B, NUSAP1, KIF20B, RASGEF1B, CENPE, VPS37B, SAE1, NUF2, ASXL2, CCNA2, MKI67, PDE3B, ZFP36L2, NR4A2, SGO2, HMGB2, PTTG1, PCLAF, TMIGD2, TK1, LDLRAD4, ZEB2, ARL6IP1, CKS1B, UBE2C, AURKB, KPNA2, HMGB1, HMGN2, CD3E, CENPW, LAYN, MXD3, DNAJC9, and TRGC2) had PPH3 + PPH4 > 0.8, suggesting a strong causal relationship with breast cancer (P < 0.05) (Figures S4–8). However, in CRC, only the ZFP36L2 gene showed PPH3 + PPH4 > 0.8, indicating a strong causal relationship with the disease (P < 0.05) (Table S5), making this gene a priority candidate for a common therapeutic target for both diseases. Although the colocalization results for ITGAE, PTTG1, and CKS1B were not significant, MR analysis showed that these three genes still have a causal relationship with both breast cancer and CRC (P < 0.05), suggesting that they may also serve as common therapeutic targets (Figure S9).

3.5 Downstream analysis

We identified the ZFP36L2 gene as a common therapeutic target for both breast cancer and colorectal cancer that should be prioritized. The ITGAE, PTTG1 and CKS1B genes are also potential common therapeutic targets for these conditions. We therefore analyzed these four therapeutic target genes to validate them as targets at the single-cell level. Additionally, we performed simulated time-series analyses of these as drug targets associated with the two diseases. As depicted in Figs. 5A and 6A, the genes marked in purple represent risk genes. Genes positioned above the horizontal line in the diagrams indicate an increase in gene expression over time, whereas those below the line exhibit a decrease in gene expression as time progresses. ZFP36L2 and ITGAE are above the horizontal line, indicating an increase in their expression levels over time. Conversely, CKS1B and PTTG1 are below the line, indicating that their expression levels are decreasing over time (P < 0.05). Specifically, the expression of CKS1B (Figs. 5B and 6B ), PTTG1 (Figs. 5E and 6E ) decreases over time in both diseases(P < 0.05). Conversely, the expression levels of ZFP36L2 (Figs. 5C and 6C) and ITGAE (Fig. 5D and 6D) increased over time (P < 0.05).

We next explored the interactions between these target genes within CD8_CM and other clustered cells. In breast cancer, we found that CKS1B + CD8_CM, ZFP36L2 + CD8_CM, ITGAE + CD8_CM, PTTG1 + CD8_CM, CKS1B-CD8_CM, ZFP36L2-CD8_CM, ITGAE- CD8_CM PTTG1- CD8_CM mainly interact with other cell types through the CXCL13-CXCR3 pathways (Figure S10 A-H).. In CRC, we found that CKS1B + CD8_CM, ZFP36L2 + CD8_CM, ITGAE + CD8_CM, PTTG1 + CD8_CM, CKS1B-CD8_CM, ITGAE-CD8_CM, PTTG1-CD8_CM mainly interact with other cell types through the MIF- (CD74 + CXCR4), MIF- (CD74 + CD44) pathways (Figure S11 A-H). Additionally, the metabolic roles of cells were explored, revealing associations between these four drug-target genes (ZFP36L2, ITGAE, PTTG1, and CKS1B) and pathways involving cysteine and methionine metabolism (Figure S10 I–L and Figure S11 I–L).

3.6 Spatial transcriptomics

We identified 12 distinct cell clusters in breast cancer samples (Fig. 7A) and 13 distinct cell clusters in CRC (Fig. 7B) after unsupervised clustering and UMAP visualization of spatial barcode patches. DotPlot analysis was employed to assess expression levels of common drug target (Fig. 7C, D).

Subsequent analysis identified ZFP36L2 as having the highest expression level in cell cluster 0 of breast cancer and in cell cluster 2 of CRC. Additionally, CKS1B, ITGAE, and PPTG1 exhibited the highest expression levels in breast cancer cell cluster 12 and CRC cell cluster 4. Metabolic assessment using scMetabolism showed that in clusters with expression of CKS1B, ITGAE, and PPTG1, the average scores for cysteine and methionine metabolism pathways were 0.8 higher compared to other cell clusters, which is consistent with single-cell sequencing results (Figure S13).

Positive correlations in expression and spatial distribution of ZFP36L2, CKS1B, ITGAE, and PPTG1 were observed (Figs. 8 and 9), indicating that regulation of these drug target genes is altered in both breast cancer and CRC.

3.7 Pharmacological evaluation of protein-protein interactions and potential drug targets

PPI analysis among these four genes only predicted interactions between PTTG1 and CKS1B (Figure S13). Among these four genes, only CKS1B has been targeted for drug development, and fluoxetine has been investigated as a treatment to inhibit its expression to arrest the growth of breast cancer[28] and enhance the sensitivity of bladder cancer to cisplatin[29].

This study employed scRNA sequencing to identify distinct subpopulations of T cells and analyze differentially expressed genes between CD8_CM and other cell types using R programming tools. We further performed cellular communication analysis to reveal genetic interactions and associations among cells. The screened genes were transformed into eqtLs to identify common drug targets for breast and colorectal cancer. We employed a combination of MR and colocalization analyses to evaluate potential therapeutic targets for breast cancer and CRC as a clinical translation of GWAS findings[30]. "Causal relationships" identified by MR could represent reverse causation, horizontal pleiotropy, or confounding due to linkage disequilibrium (LD)[31]. By conducting primary MR analyses on identified differentially-expressed genes, we were able to exclude risk genes associated with reverse causation from further analysis. To mitigate bias due to horizontal pleiotropy, eQTLs were used as exposures, given their direct role in the transcription and/or translation of genes associated with eQTLs[32]. Moreover, Bayesian colocalization set a critical threshold for posterior probabilities at 0.8, identifying NUCKS1, FOSL2, ITGAE, BIRC5, CENPM, CDKN3, UBE2S, CLNK, SMC4, CDC20, FASLG, STMN1, CENPF, TMPO, ZWINT, CENPK, DLGAP5, MAP2K2, DUT, NASP, CCNB1, PTPN22, AHI1, SMC2, ANP32B, NUSAP1, KIF20B, RASGEF1B, CENPE, VPS37B, SAE1, NUF2, ASXL2, CCNA2, MKI67, PDE3B, ZFP36L2, NR4A2, SGO2, HMGB2, PTTG1, PCLAF, TMIGD2, TK1, LDLRAD4, ZEB2, ARL6IP1, CKS1B, UBE2C, AURKB, KPNA2, HMGB1, HMGN2, CD3E, CENPW, LAYN, MXD3, DNAJC9, and TRGC2 as likely harboring the same variants. In the colocalization analysis for CRC, ZFP36L2 was identified by MR as being suggested to share the same variants. Because this gene was identified in analyses of both breast cancer and CRC, it could potentially be a common therapeutic target for both diseases. Although the colocalization between CKS1B, PTTG1, and ITGAE is not strong in CRC, Mendelian analysis indicates a causal relationship between these genes and both diseases, suggesting that they might represent common targets for colorectal cancer and breast cancer (P < 0.05).

ZFP36L2 dysregulation is one of the key drivers of colorectal cancer. Mutations in this gene reduce levels of the encoded RNA-binding protein and CRISPR/Cas9-mediated knockout of this gene results in an enhancement of the migration and invasion capabilities of cells to promote carcinogenesis[33]. This study is the first to find an association between ZFP36L2 and the risk of breast cancer, with increased expression levels of ZFP36L2 raising the risk of developing the disease (P < 0.05).CKS1B is typically upregulated in CRC tissues and knockout of the gene can inhibit the proliferation and migration of CRC cells, providing a new potential strategy for treating colorectal cancer[34]. Specifically in breast cancer, elevated expression of CKS1B is associated with increased turnover of the Kip1 gene and cell renewal, leading to increased cell proliferation and poor prognosis[34]. Overall, CKS1B may become a potential therapeutic target and prognostic marker for various cancers including CRC, breast cancer, pancreatic cancer, and retinoblastoma. In breast cancer patients, PTTG1 mRNA levels increase with elevated estrogen levels, which known risk factor for breast cancer, suggesting that estrogen may influence the progression of breast cancer by regulating PTTG1[35]. This study is the first to find an association between PTTG1 and the risk of CRC with increased expression levels of ZFP36L2 raising the risk of developing the disease (P < 0.05). ITGAE, also known as CD103 or CD4T[36]. Mature differentiated CD103-specific cytotoxic T lymphocytes can self-regulate by producing activated TGFβ1, increasing T cell receptor antigen sensitivity and enhancing the rapid recognition and clearance of cancer cells[37]. In CRC patients, the degree of ITGAE + lymphocyte invasion correlates with patient survival, which is potentially linked to interferon-responsive chemokine and epithelial-mesenchymal transition signaling pathways, suggesting that ITGAE is a possible biomarker for CRC[38]. The number of CD103 + T cells is higher in breast cancer tissues than in normal tissues because of increased lymphocyte migration and retention, which impacts anti-tumor immune functions[39].

Through single-cell analysis and simulated time-series, we found that the expression level of ZFP36L2, ITGAE gradually increases over time, whereas the expression of CKS1B, PTTG1 decreases. We examined the relationships between drug target genes that are differentially expressed between different cells and identified ligands which could facilitate communication between different cell types. These include ANXA1-FPR1, ANXA1-FPR2, MIF-(CD74 + CD44), LGALS9-CD45, MIF-(CD74 + CXCR4), CCL5-CCR5 for + CD8_CM, and CCL5-CCR1, MIF-(CD74 + CD44), CCL5-CCR5, MIF-(CD74 + CXCR4) for -CD8_CM. Notably, the LGALS9-CD45 signaling pathway, in addition to impacting breast and colorectal cancers, also affects gastric cancer by influencing monocytes within the tumor microenvironment, thereby significantly impacting T cells and endothelial cells[40]. We performed metabolomic studies and found that the drug target genes were linked to the metabolism of cysteine and methionine. Previous reports suggest that cysteine metabolism is associated with a variety of diseases, such as cardiovascular diseases (CVD), ischemic stroke, neurological disorders, diabetes, lung and colorectal cancer, renal failure-related diseases, and vitiligo[41], whereas the relationship of methionine metabolism with diseases has not been extensively studied yet. In the spatial transcriptomics metabolic analysis, we found that the highest scoring pathways related to drug targets were cysteine and methionine metabolism. Analysis of the distribution and expression of drug target genes revealed that most are in a mid-differentiation state. PPI analysis showed very limited interactions among potential drug target genes, with a significant interaction only observed between PTTG1 and CKS1B. Fluoxetine is a drug that has been developed to reduce expression of CKS1B which has been shown to inhibit the growth of breast cancer[26] and enhance the sensitivity of bladder cancer to cisplatin[27].

The database used in this study is limited to a European population, and so whether the conclusions that we have made are specific to this population or globally applicable requires further investigation. Additionally, the statistical analyses employed and the strict significance thresholds may have filtered out some marker genes which could potentially be common drug targets for breast and colorectal cancer. Therefore, further extensive database analyses and human case studies are necessary to assess drug target genes associated with the development and progression of breast cancer and CRC. While every cancer has context-specific biology, previous studies have reported that some risk genes, such as CDC20, play a common role in both breast and colorectal cancers[42]. We did not find that CDC20 is a common risk for both breast cancer and CRC, but this may be due to sample size and population-specific factors. We identified several drug targets and their interactions with common signaling pathways in the PPI analysis. However, PPI analysis is only suggestive, not conclusive for clinical research, and has its limitations. Although we have reported the roles of pathways as well as the expression and distribution of genes in single-cell and spatial transcriptomics, further experimental validation is required. Additionally, studies need to include more diverse, non-European populations to evaluate the applicability of these findings for clinical use.

In conclusion, the genes ZFP36L2, CKS1B, PTTG1, and ITGAE may serve as common drug targets for breast and colorectal cancers. They might influence the pathogenesis and pathophysiology of these cancers through functioning in CD8_CM in addition to potentially affecting the cancer cells through altering cysteine and methionine metabolic pathways. The four genes could be potential therapeutic targets as well as biomarkers for screening for the presence of cancer and monitoring the effectiveness of therapies.

CRC: Colorectal cancer; BRCA2: Breast Cancer Gene 2; TME: Tumor Microenvironment; T_CM: central memory T cells; CD8_CM: central memory CD8+ T cells; PPI: protein-protein interaction; scRNA: single-cell RNA; GEO: Gene Expression Omnibus; eQTL: expression quantitative trait loci; MR-IVW: inverse variance weighted Mendelian Randomization; PCA: Principal Component Analysis; PPICN: Protein-Protein Interaction Core Network; MR: Mendelian randomization; ECM: extracellular matrix; LD: linkage disequilibrium; CVD: cardiovascular diseases.

Ethics approval

The data used is from public dataset, and the original data has obtained ethical approval.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and materials

The data used and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests

All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

Funding

Financial support came from Shanghai Jiao Tong University School of Medicine.

Authors' contributions

Rui Tang: Conceptualization, Formal analysis, Methodology, Writing-original draft; Hongquan Cui: Validation, Visualization, Writing-original draft; Pengyu Miao: Resources, Writing-review & editing; Zhengrui Li: Investigation, Validation, Resources, Writing-review & editing; Keliang Wang: Supervision, Resources, Writing-review & editing.

Acknowledgements

We would also like to thank all laboratory members for helpful discussions.

Sung, H.; Ferlay, J.; Siegel, R.L.; Laversanne, M.; Soerjomataram, I.; Jemal, A.; Bray, F. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: a cancer journal for clinicians 2021, 71, 209-249, doi:10.3322/caac.21660.
Arnold, M.; Morgan, E.; Rumgay, H.; Mafra, A.; Singh, D.; Laversanne, M.; Vignat, J.; Gralow, J.R.; Cardoso, F.; Siesling, S.; et al. Current and future burden of breast cancer: Global statistics for 2020 and 2040. Breast (Edinburgh, Scotland) 2022, 66, 15-23, doi:10.1016/j.breast.2022.08.010.
Ellisen, L.W.; Haber, D.A.J.A.r.o.m. Hereditary breast cancer. 1998, 49, 425-436.
Breast cancer and breastfeeding: collaborative reanalysis of individual data from 47 epidemiological studies in 30 countries, including 50302 women with breast cancer and 96973 women without the disease. Lancet (London, England) 2002, 360, 187-195, doi:10.1016/s0140-6736(02)09454–0.
Yager, J.D.; Davidson, N.E. Estrogen carcinogenesis in breast cancer. The New England journal of medicine 2006, 354, 270-282, doi:10.1056/NEJMra050776.
Butler, L.M.; Potischman, N.A.; Newman, B.; Millikan, R.C.; Brogan, D.; Gammon, M.D.; Swanson, C.A.; Brinton, L.A. Menstrual risk factors and early-onset breast cancer. Cancer causes & control : CCC 2000, 11, 451-458, doi:10.1023/a:1008956524669.
Clavel-Chapelon, F.; Gerber, M. Reproductive factors and breast cancer risk. Do they differ according to age at diagnosis? Breast cancer research and treatment 2002, 72, 107-115, doi:10.1023/a:1014891216621.
Olsson, H.L.; Olsson, M.L. The Menstrual Cycle and Risk of Breast Cancer: A Review. Frontiers in oncology 2020, 10, 21, doi:10.3389/fonc.2020.00021.
Macacu, A.; Autier, P.; Boniol, M.; Boyle, P. Active and passive smoking and risk of breast cancer: a meta-analysis. Breast cancer research and treatment 2015, 154, 213-224, doi:10.1007/s10549–015-3628-4.
McDonald, J.A.; Goyal, A.; Terry, M.B. Alcohol Intake and Breast Cancer Risk: Weighing the Overall Evidence. Current breast cancer reports 2013, 5, doi:10.1007/s12609–013–0114-z.
Moore, S.C.; Lee, I.M.; Weiderpass, E.; Campbell, P.T.; Sampson, J.N.; Kitahara, C.M.; Keadle, S.K.; Arem, H.; Berrington de Gonzalez, A.; Hartge, P.; et al. Association of Leisure-Time Physical Activity With Risk of 26 Types of Cancer in 1.44 Million Adults. JAMA internal medicine 2016, 176, 816-825, doi:10.1001/jamainternmed.2016.1548.
Tsai, H.H.; Yu, J.C.; Hsu, H.M.; Chu, C.H.; Chang, T.M.; Hong, Z.J.; Feng, A.C.; Fu, C.Y.; Hsu, K.F.; Dai, M.S.; et al. The Risk of Breast Cancer between Western and Mediterranean Dietary Patterns. Nutrients 2023, 15, doi:10.3390/nu15092057.
Playdon, M.C.; Matthews, S.B.; Thompson, H.J. Weight change patterns and breast cancer risk: a brief review and analysis. Critical reviews in eukaryotic gene expression 2013, 23, 159-169, doi:10.1615/critreveukaryotgeneexpr.2013007047.
Fidler, M.M.; Bray, F.; Vaccarella, S.; Soerjomataram, I. Assessing global transitions in human development and colorectal cancer incidence. International journal of cancer 2017, 140, 2709-2715, doi:10.1002/ijc.30686.
Arnold, M.; Sierra, M.S.; Laversanne, M.; Soerjomataram, I.; Jemal, A.; Bray, F. Global patterns and trends in colorectal cancer incidence and mortality. Gut 2017, 66, 683-691, doi:10.1136/gutjnl-2015-310912.
Murphy, N.; Moreno, V.; Hughes, D.J.; Vodicka, L.; Vodicka, P.; Aglago, E.K.; Gunter, M.J.; Jenab, M. Lifestyle and dietary environmental factors in colorectal cancer susceptibility. Molecular aspects of medicine 2019, 69, 2-9, doi:10.1016/j.mam.2019.06.005.
Dekker, E.; Tanis, P.J.; Vleugels, J.L.A.; Kasi, P.M.; Wallace, M.B. Colorectal cancer. Lancet (London, England) 2019, 394, 1467-1480, doi:10.1016/s0140-6736(19)32319–0.
Lai, J.H.; Park, G.; Gerson, L.B. Association between breast cancer and the risk of colorectal cancer. Gastrointestinal endoscopy 2017, 86, 429-441.e421, doi:10.1016/j.gie.2017.04.008.
Barzi, A.; Lenz, A.M.; Labonte, M.J.; Lenz, H.J. Molecular pathways: Estrogen pathway in colorectal cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 2013, 19, 5842-5848, doi:10.1158/1078–0432.Ccr-13–0325.
Chen, S.; Allgayer, H. Epigenetically Downregulated Breast Cancer Gene 2 through Acetyltransferase Lysine Acetyltransferase 2B Increases the Sensitivity of Colorectal Cancer to Olaparib. Cancers 2023, 15, doi:10.3390/cancers15235580.
Mo, P.; Zhou, Q.; Guan, L.; Wang, Y.; Wang, W.; Miao, M.; Tong, Z.; Li, M.; Majaz, S.; Liu, Y.; et al. Amplified in breast cancer 1 promotes colorectal cancer progression through enhancing notch signaling. Oncogene 2015, 34, 3935-3945, doi:10.1038/onc.2014.324.
Yang, Z.; Liu, L.; Zhu, Z.; Hu, Z.; Liu, B.; Gong, J.; Jin, Y.; Luo, J.; Deng, Y.; Jin, Y.; et al. Tumor-Associated Monocytes Reprogram CD8(+) T Cells into Central Memory-Like Cells with Potent Antitumor Effects. Advanced science (Weinheim, Baden-Wurttemberg, Germany) 2024, 11, e2304501, doi:10.1002/advs.202304501.
Klebanoff, C.A.; Gattinoni, L.; Torabi-Parizi, P.; Kerstann, K.; Cardones, A.R.; Finkelstein, S.E.; Palmer, D.C.; Antony, P.A.; Hwang, S.T.; Rosenberg, S.A.; et al. Central memory self/tumor-reactive CD8+ T cells confer superior antitumor immunity compared with effector memory T cells. Proceedings of the National Academy of Sciences of the United States of America 2005, 102, 9571-9576, doi:10.1073/pnas.0503726102.
Su, W.M.; Gu, X.J.; Dou, M.; Duan, Q.Q.; Jiang, Z.; Yin, K.F.; Cai, W.C.; Cao, B.; Wang, Y.; Chen, Y.P. Systematic druggable genome-wide Mendelian randomisation identifies therapeutic targets for Alzheimer's disease. Journal of neurology, neurosurgery, and psychiatry 2023, 94, 954-961, doi:10.1136/jnnp-2023-331142.
Giambartolomei, C.; Vukcevic, D.; Schadt, E.E.; Franke, L.; Hingorani, A.D.; Wallace, C.; Plagnol, V. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS genetics 2014, 10, e1004383, doi:10.1371/journal.pgen.1004383.
Dries, R.; Zhu, Q.; Dong, R.; Eng, C.L.; Li, H.; Liu, K.; Fu, Y.; Zhao, T.; Sarkar, A.; Bao, F.; et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome biology 2021, 22, 78, doi:10.1186/s13059–021–02286-2.
Elosua-Bayes, M.; Nieto, P.; Mereu, E.; Gut, I.; Heyn, H. SPOTlight: seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes. Nucleic acids research 2021, 49, e50, doi:10.1093/nar/gkab043.
Lei, B.; Xu, L.; Zhang, X.; Peng, W.; Tang, Q.; Feng, C. The proliferation effects of fluoxetine and amitriptyline on human breast cancer cells and the underlying molecular mechanisms. Environmental toxicology and pharmacology 2021, 83, 103586, doi:10.1016/j.etap.2021.103586.
Yang, C.J.; Tan, Z.L.; Yang, J.D.; Hsu, F.T.; Chiang, C.H. Fluoxetine inactivates STAT3/NF-κB signaling and promotes sensitivity to cisplatin in bladder cancer. Biomedicine & pharmacotherapy = Biomedecine & pharmacotherapie 2023, 164, 114962, doi:10.1016/j.biopha.2023.114962.
McGowan, L.M.; Davey Smith, G.; Gaunt, T.R.; Richardson, T.G. Integrating Mendelian randomization and multiple-trait colocalization to uncover cell-specific inflammatory drivers of autoimmune and atopic disease. Human molecular genetics 2019, 28, 3293-3300, doi:10.1093/hmg/ddz155.
Zheng, J.; Haberland, V.; Baird, D.; Walker, V.; Haycock, P.C.; Hurle, M.R.; Gutteridge, A.; Erola, P.; Liu, Y.; Luo, S.; et al. Phenome-wide Mendelian randomization mapping the influence of the plasma proteome on complex diseases. Nature genetics 2020, 52, 1122-1131, doi:10.1038/s41588–020–0682-6.
Montgomery, S.B.; Dermitzakis, E.T. From expression QTLs to personalized transcriptomics. Nature reviews. Genetics 2011, 12, 277-282, doi:10.1038/nrg2969.
Chen, H.N.; Shu, Y.; Liao, F.; Liao, X.; Zhang, H.; Qin, Y.; Wang, Z.; Luo, M.; Liu, Q.; Xue, Z.; et al. Genomic evolution and diverse models of systemic metastases in colorectal cancer. Gut 2022, 71, 322-332, doi:10.1136/gutjnl-2020-323703.
Hwang, J.S.; Jeong, E.J.; Choi, J.; Lee, Y.J.; Jung, E.; Kim, S.K.; Min, J.K.; Han, T.S.; Kim, J.S. MicroRNA-1258 Inhibits the Proliferation and Migration of Human Colorectal Cancer Cells through Suppressing CKS1B Expression. Genes 2019, 10, doi:10.3390/genes10110912.
Meng, C.; Zou, Y.; Hong, W.; Bao, C.; Jia, X. Estrogen-regulated PTTG1 promotes breast cancer progression by regulating cyclin kinase expression. Molecular medicine (Cambridge, Mass.) 2020, 26, 33, doi:10.1186/s10020–020–00161-7.
Gu, Y.; Chen, Y.; Jin, K.; Cao, Y.; Liu, X.; Lv, K.; He, X.; Lin, C.; Liu, H.; Li, H.; et al. Intratumoral CD103(+)CD4(+) T cell infiltration defines immunoevasive contexture and poor clinical outcomes in gastric cancer patients. Oncoimmunology 2020, 9, 1844402, doi:10.1080/2162402x.2020.1844402.
Abd Hamid, M.; Colin-York, H.; Khalid-Alham, N.; Browne, M.; Cerundolo, L.; Chen, J.L.; Yao, X.; Rosendo-Machado, S.; Waugh, C.; Maldonado-Perez, D.; et al. Self-Maintaining CD103(+) Cancer-Specific T Cells Are Highly Energetic with Rapid Cytotoxic and Effector Responses. Cancer immunology research 2020, 8, 203-216, doi:10.1158/2326-6066.Cir-19–0554.
Hu, X.; Li, Y.Q.; Li, Q.G.; Ma, Y.L.; Peng, J.J.; Cai, S.J. ITGAE Defines CD8+ Tumor-Infiltrating Lymphocytes Predicting a better Prognostic Survival in Colorectal Cancer. EBioMedicine 2018, 35, 178-188, doi:10.1016/j.ebiom.2018.08.003.
Seo, E.H.; Song, G.Y.; Oh, C.S.; Kim, S.H.; Kim, W.S.; Lee, S.H. CD103(+) Cells and Chemokine Receptor Expression in Breast Cancer. Immune network 2023, 23, e25, doi:10.4110/in.2023.23.e25.
Wei, C.; Ma, Y.; Wang, F.; Chen, Y.; Liao, Y.; Zhao, B.; Zhao, Q.; Tang, D. Machine learning and single-cell sequencing reveal the potential regulatory factors of mitochondrial autophagy in the progression of gastric cancer. Journal of cancer research and clinical oncology 2023, 149, 15561-15572, doi:10.1007/s00432–023–05287-9.
Rehman, T.; Shabbir, M.A.; Inam-Ur-Raheem, M.; Manzoor, M.F.; Ahmad, N.; Liu, Z.W.; Ahmad, M.H.; Siddeeg, A.; Abid, M.; Aadil, R.M. Cysteine and homocysteine as biomarker of various diseases. Food science & nutrition 2020, 8, 4696-4707, doi:10.1002/fsn3.1818.
Wang, L.; Zhang, J.; Wan, L.; Zhou, X.; Wang, Z.; Wei, W. Targeting Cdc20 as a novel cancer therapeutic strategy. Pharmacology & therapeutics 2015, 151, 141-151, doi:10.1016/j.pharmthera.2015.04.002.

No competing interests reported.

originaldata.rar

Download PDF

Editorial decision: Revision requested
30 Aug, 2024
Editor assigned by journal
29 Aug, 2024
Submission checks completed at journal
29 Aug, 2024
First submitted to journal
28 Aug, 2024

You are reading this latest preprint version

Novel common target genes for breast cancer and colorectal cancer: A mendelian randomization and spatial transcriptomics study

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials and Methods

2.1 single-cell RNA sequencing

2.2 Data source

2.3 Mendelian randomization

2.4 Bayesian colocalization analysis

2.5 Downstream analysis of single-cell RNA sequencing technology

2.6 Spatial transcriptomics processing and annotation

2.7 Protein-protein interaction analysis

3. Results

3.1 Single-cell RNA sequencing results

3.2 Interactions of CD8_CM with other cell types

3.3 Screening for risk genes in breast and colorectal cancers

3.4 Sensitivity analysis of pathogenic genes

3.5 Downstream analysis

3.6 Spatial transcriptomics

3.7 Pharmacological evaluation of protein-protein interactions and potential drug targets

4. Discussion

5. Conclusion

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1