Transcriptomic analysis identified new lipid biomarkers for patient stratification in breast cancer

doi:10.21203/rs.3.rs-1837155/v1

Background

Breast cancer is a common malignancy among women and a big challenge that causes an extensive burden on public health. Breast cancer is a heterogeneous malignancy, and the exact mechanism of development is not clear. Lipid biomarkers can play an important role in predicting the outcome of breast cancer patients. Previously different biomarkers have been identified against survival, metastasis, and prognosis. However, lipid associated biomarkers are not readily used for predicting prognosis and devising new treatment options. Therefore, we aimed to analyze lipid metabolism genes to identify the most significant lipid metabolism biomarkers in BC.

Methods

RNA-seq data of 1787 breast cancer patients were downloaded from Gene Expression Omnibus and The Cancer Genome Atlas databases. The mRNA expression of ABAT, SRD5A1, and PSAT1 was analyzed in SK-BR3, MDA-MB-231, BT-549 BC cell lines and normal breast cells. We investigated the clinicopathological parameters of lipid metabolism genes using the Kruskal-Wallis test, post-hoc Mann-Whitney-U-test, Chi-square/Fisher-exact test, Kaplan–Meier, and cox regression survival analysis.

Results

The expression values of 21 lipid metabolism genes showed significant association with different clinical parameters including grade, node, tumor size, stage, race, age, and molecular subtypes of BC patients. Furthermore, survival analysis revealed that ABAT, SRD5A1, and PSAT1 were significantly associated with RFS, DFS, and OS in BC. Overexpression of ABAT was significantly associated with better RFS, DFS, and OS in all independent cohorts. In contrast, overexpression of SRD5A1 and PSAT1 was found associated with poor RFS, DFS, and OS. According to in vitro validation results, mRNA expression of ABAT was downregulated, while, SRD5A1 and PSAT1 were upregulated in all tested BC cell lines compared to normal cells.

Conclusion

We concluded that ABAT can be used as a better prognostic marker, while SRD5A1 and PSAT1 can be used as poor prognostic markers in breast cancer patients. These findings can lead us to devise new strategies classifying BC into different prognosis groups which may lead to novel treatment approaches.

Lipid biomarkers

Breast cancer

Overall survival

Prognosis

Breast cancer is the most common malignancy among women, arises from connective tissues, lobules, or ducts (Bray, Ferlay et al. 2018). Despite a better decline in its prevalence rate, breast cancer is still a big challenge and causes an extensive burden on public health. In 2022, about 43,250 breast cancer deaths and 287,850 new cases are expected to occur in the United States (Siegel, Miller et al. 2022). Biological heterogeneity in breast cancer, due to clinical behavior, risk factors, different molecular subtypes, and responses to therapy represent the main problem to treatment success and survival. Investigations based on global transcriptome and gene expression analysis found four major molecular subtypes of breast tumors defined as Luminal-A, Luminal-B, HER2E, Basal-like, and triple-negative breast cancer (TNBC). All these molecular subtypes of breast cancer have also been categorized based on baseline prognosis, risk factors, differences in incidence rate, response to treatment, and age at diagnosis (Curtis, Shah et al. 2012). Only TNBC accounts for 11%-17% of all breast tumor cases (Bianchini, Balko et al. 2016), and it has a very poor prognosis due to fewer therapeutic choices and frequent early relapse (Dignam, Dukic et al. 2009, Denkert, Liedtke et al. 2017). T Studies found that 50% of breast cancer patients who possess nodes are cured of localized treatment while 30% of untreated patients with non-nodal metastasis developed into recurrent or metastatic cancer after several years (Paoletti and Hayes 2014). Other major prognostic biomarkers which have been commonly used are tumor grade and tumor size (Cianfrocca and Goldstein 2004).

Reprogramming of lipid metabolism is a well-known hallmark of cancer disease (Koundouros and Poulogiannis 2020), normally characterized by high lipid storage and lipogenesis, required for metastasis and increased tumor growth (Song, Liu et al. 2020). In the case of breast cancer, the cells fulfill energy demand from their microenvironment, and through de novo biosynthesis which are critical for cancer cells survival and metabolic flexibility (Glatz, Luiken et al. 2010). In breast cancer cells it has been shown that metabolic reprogramming supports cell survival and proliferation. Alteration in the metabolism of lipids has been identified as an important metabolic rewiring process in breast cancers (Du, Sikora et al. 2018). Increase in the biosynthesis and uptake of lipid that accumulates fats in different cancers induces tumor growth (Hultsch, Kankainen et al. 2018). In TNBC the most aggressive subtype of breast cancer, overexpression of HER2 directly confers lipogenic phenotype and is strongly associated with tumor suppressor P53 mutation. Overexpression of HER2 activates PI3K/AKT/mTOR pathway to enhance proliferation, cell growth, and survival. PI3K/AKT/mTOR pathway activation alters phospholipid composition such as phosphatidylinositol (PI) and phosphatidylserine in breast cancer cells (Baumann, Sevinsky et al. 2013). Several phospholipids are found to be upregulated estrogen receptor-negative breast cancer (Hilvo, Denkert et al. 2011). To date, several lipid biomarkers have been identified in breast cancer cells. For instance, ACLY, SREBPs, CPT1, Acetyl-CoA, PPARγ, ACC, CPT1B, PGC1α, FASN, SCD, and CDCP1, are the key lipid metabolism enzymes that regulate metabolism in breast cancer. FASN is highly upregulated in breast cancer tissues as compared to normal breast tissues with the highest expression in TNBC (Menendez and Lupu 2017, Ward, Anderson et al. 2021). However, despite of extensive efforts, no lipid metabolism associated biomarkers are used as breast cancer treatment options.

Currently, the regulatory association between transcriptome and lipid metabolism has not been fully explored. Therefore, understanding the relationship between lipid metabolism and transcriptomic data may identify new prognostic subtypes related to BC development. In the current study, a new set of lipid metabolism genes were identified by integrating large scale breast cancer data extracted from Gene Expression Omnibus (GEO) and TCGA. Moreover, findings were also validated in SK-BR3, MDA-MB-231, and BT-549 breast cancer cell lines.

Data collection and screening

The overall study design is summarized in a flow chart, Fig. 1. Clinical and transcriptomic profiling of lipid metabolic genes in 1787 (156 + 278 + 167 + 104 + 1082) breast cancer patients was performed. The discovery cohort 1 GSE50948 (D1), discovery cohort GSE20194 (D2), and discovery cohort 3 GSE53031 (D3) were downloaded from the Gene Expression Omnibus (GEO) database. LogR mRNA expression values were generated by Robust multiarray average (RMA) normalization. Based on their median values, the expression data from discovery cohorts were divided into high and low expression groups. Clinical parameters of discovery cohort 1, included ER, HER2, PR status, age, grade, tumor size, residual and menopausal status, detailed information of all discovery cohorts can be found in Table 1. The clinical features of discovery cohort 2 included ER, HER2, PR status, node, T classification, and grade. Clinical features of discovery cohort 3 contain pam50, ER, HER2, PR status, tumor size, grade, node, and relapse-free survival information. For validation of survival analysis, validation cohort 1 and validation cohort 2 was processed. Validation cohort 1 (GSE42568) data was retrieved from the GEO database and RMA normalization was performed to generate mRNA expression. The clinicopathological parameters of validation cohort 1included age, ER status, grade, size, lymph node, relapse-free survival, and overall survival. RNA sequencing data of 1082 breast cancer patients (validation cohort 2) RPKM and z-score values of 1082 breast cancer patients were downloaded from cBioPortal database (http://www.cbioportal.org/public-portal/). Clinical information includes N stage, tumor size, M stage, T stage, disease-free survival, and overall survival. The detailed information of validation cohort 1 and validation cohort 2 is summarized in Table 1. Additionally, a list of 1710 lipid metabolism genes was downloaded through text mining, database search, and from MSigDB–GSEA (http://www.gsea-msigdb.org/gsea/msigdb/) and identified unique genes list of 1035 genes in KEGG, PID, and Reactome lipid metabolism pathways (Additional file 1, table S1). BRB array tools (4.6.2 - Beta_1) and R-script (4.1.2) were used to pre-process and modify all the cohorts under study.

Cell Culture

Human breast carcinoma cell lines SK-BR3 (HTB-30), MDA-MB-231 (HTB-26), BT-549 (HTB-122), and normal breast cells MCF10A purchased from American Type Culture Collection (ATCC), were respectively cultured in McCoy’s 5A Medium (16600082, Gibco), Leibovitz’s L-15 Medium(11415056, Gibco), RPMI-1640 medium (11876093, Gibco) supplemented with 0.023 IU/mL insulin (Sigma, 91077C) and MEMB medium, with 0.1% penicillin streptomycin (Thermo, 15140122), 10% FBS (fetal bovine serum) (Gibco, 10100147), and anti-mycoplasma reagent Save-It (HB-SV1000, Hanbio). All cell lines were maintained in a humidified 5% CO2 atmosphere at 37°C.

Rna Extraction And Rt-qpcr

RNA extraction and RT-qPCR

To find expression of ABAT, SRD5A1, and PSAT1 in three different breast cancer cell lines, and normal MCF-10A cells. Total RNA was extracted using RNAex pro reagent (Accurate biotechnology Hunan, China, AG21101) according to the manufacturer’s protocol. The extracted RNA was then used to synthesize cDNA using Easy-Script, one-step gDNA removal, and cDNA synthesis super-mix (AE31102, Trans-Gen Biotech) for qPCR. RT-qPCR was performed using a universal SYBR qPCR master mix (Vazyme Biotech, Q71102), processed on a real-time PCR system (Agilent-technologies, Stratagene, MX3005P). ACTB was kept as an internal control to normalize the expression of RNA. The primers used for RT-qPCR are shown in Additional file 1 Table S2.

Statistical analysis

IBM SPSS Statistics for Windows, Version 22.0. (Armonk, NY, USA), Graph pad prism and R software version 4.1.2 was used for all the statistical analysis. Statistical analysis of gene expression was analyzed against the clinical features present in each cohort. The association of genes was analyzed using Fisher Exact, Kruskal Wallis, Pearson correlation, and Chi-square tests. Survival analysis was performed using Kaplan Meier plots to determine the association of genes with relapse-free survival, disease-free survival, and overall survival. To evaluate the association between gene expression values and survival Cox proportional hazard regression, univariate and multivariate analysis was used. P-value < 0.05 was considered statistically significant. The hazard ratio and 95% CI (confidence interval) were also analyzed for each clinical factor.

Table 1

clinicopathological factors of discovery and validation cohorts of BC patients
Factors	D1		D2		D3		V1		V2
	N	%	N	%	N	%	N	%	N	%
ER-negative ER-positive NA	102 52 2	65.38 33.33 1.28	108 161 9	38.84 57.91 3.23	43 124	25.74 74.25	34 67 3	32.69 64.42 2.88
HER2 negative HER2 positive NA	41 113 2	26.28 72.43 1.28	219 59	78.77 21.22	136 31	81.43 18.56
PR negative PR positive	121 35	77.56 22.43	157 121	56.47 43.52	53 114	31.73 68.26
Age ≤ 51 > 51	76 78	48.71 50	≤ 52 161 > 52 117	57.91 42.08	≤ 36 89 > 36 78	53.29 46.70	≤ 58 57 > 58 47	54.80 45.19	≤ 58 546 > 58 536	50.46 49.53
Grade G1 G2 G3	0 67 86	0 43.79 56.20	13 105 150	4.67 37.76 53.95	G1-2 68 G3 92 NA 7	40.71 55.08 4.19	G1- G2 51 G3 53	49.03 50.96
Residual pCR RD	53 103	33.97 66.02	56 222	20.14 79.85
T status T1-T2 T3-T4 NA			172 104 2	61.87 37.41 0.71	71 90 7	42.51 53.89 4.19	≤ 2.78 55 > 2.78 49	52.88 47.11	T1-≤T2 904 >T2 175 3	83.54 16.17 0.27
Node N0- N1 N2-N3 NA			204 72 2	73.38 25.89 0.71	N0 78 N1 53 N2-N3 36	46.70 31.73 21.55	N0 45 N1 59	43.26 56.73	N0- N1 870 N2-N3 192 20	80.40 17.74 1.84
Race White Black₊ Asian Hispanic			176 50 42	63.30 17.98 15.10					White 749 African 243 NA 90	69.22 22.45 8.31
RFS event Alive Dead NA					125 42	74.85 25.14	56 48	53.84 46.15	DFS 856 84 142	79.11 7.76 13.12
OS event Alive Dead							69 35	66.34 33.65	931 151	86.04 13.95
Pregnancy 0 1					113 54	67.66 32.33
Metastasis M0 M1 NA									900 27 161	83.17 2.49 14.87
Trimester 1–2 3 NA					41 12 114	24.55 7.18 68.26
Tumor stage I-II III-IV NA									795 268 21	73.47 24.76 1.94

Footnotes: D, Discovery cohort; N, Node; T, tumor; RD, Residual; pCR, complete pathological response; RFS, Relapse free survival; M, Metastasis; OS, Overall survival.

Clinical analysis of D1 (GSE50948)

The differential expression analysis of breast cancer subtypes (ER-negative vs positive, PR negative vs positive, and HER2 negative vs positive) was performed, and a total of 2547 differentially expressed genes (p = < 0.05) were obtained. The differentially expressed genes were matched with lipid metabolism genes (n = 1035) and screened 273 common genes (Figure 2a). After processing, the mRNA expression values of 273 lipid metabolism genes were evaluated against all the available clinical features. Among these 273 genes, 114 lipid metabolic genes were significantly associated with most of the clinical parameters. Out of 114 genes, 53 genes showed overexpression in ER-negative samples, while 51 showed downregulation. In HER2 negative samples, 18 genes were upregulated, and 16 genes showed downregulation. Similarly, 39 lipid metabolic genes were overexpressed, and 15 genes showed downregulation in PR negative samples. Furthermore, most genes were significantly associated with grade, tumor size, age, residual, and menopausal status, detailed in additional file 2, Table S1. On subsequent analysis, 26 genes (8 downregulated, 18 upregulated) were significantly associated with tumor grade and other clinical factors, (Additional file 1, table S3).

Clinical analysis of D2 (GSE20194)

Like cohort 1 the differential expression analysis of breast cancer subtypes (ER-negative vs positive, PR negative vs positive, and HER2 negative vs positive) was performed, and a total of 3801 differentially expressed genes (p = < 0.05) were obtained. The differentially expressed genes were matched with lipid metabolism genes (n = 1035) and screened 510 common genes Figure 2a. Among 510 lipid metabolism genes, 114 genes were common with cohort 1 and were significantly associated with different clinical parameters. Genes significantly associated with breast cancer subtypes ER, HER2, and PR were 112, 54, and 89, respectively. In ER-negative sample, 62 genes were downregulated, and 51 genes showed upregulation, while in HER2 negative class 33 lipid genes were downregulated and 20 genes were upregulated. Moreover, 48 genes were downregulated, and 41 lipid genes showed upregulation in PR-negative patients. On subsequent analysis, 96 genes were associated with tumor grade whereas 21 genes were common with cohort 1. According to the clinical analysis results, high expression of 15 genes (ALDH3A2, PTDSS1, CERS6, ITPR1, IRS1, ABAT, ADRA2A, FOXA1, AREG, CCND1, IGF1R, MYB, CSAD, ESR1, ERLIN2), and low expression of 6 genes (SRD5A1, PSAT1, CXCL9, RARRES1, PDXK, UGP2) were significantly (P > 0.05) associated with early tumor grade (Additional file 1, table S4). Detailed association with other clinicopathological parameters is available in Additional file 2, Table S2.

Clinical analysis of D3 (GSE53031)

After differential expression analysis, 4100 differentially expressed genes were identified and matched with 1035 lipid metabolism genes. After processing, 476 lipid metabolism genes were identified and subjected to clinical analysis. Moreover, the genes signature of 114 lipid metabolism genes, common in cohorts 1 and 2 were also significantly correlated with different clinicopathological parameters in discovery cohort 3. Briefly, genes significantly (p = <0.05) correlated with ER, HER2, and PR subtypes are 110, 20, and 106 respectively. Out of 114 genes, 55 lipid metabolic genes showed overexpression, and 55 genes were downregulated in ER-negative samples. In HER2 negative samples, 4 genes were downregulated, and 16 genes were overexpressed. Similarly, in PR negative samples 59 genes were downregulated and 47 genes showed overexpression. Moreover, 45 genes were downregulated, and 48 genes showed overexpression in early tumor grade (Additional file 2, Table S3). Furthermore, genes commonly significant in discovery cohorts 1 and 2 also showed significant association with tumor grade and other clinicopathological factors in discovery cohort 3 (Additional file 1 table S5. The genes commonly associated with clinical features in all discovery cohorts were subjected to Kaplan Meier survival analysis. Survival analysis results revealed that overexpression of only ABAT (P = 0.0089), was associated with better relapse-free survival (RFS) figure 2b.

Cox regression survival analysis of discovery cohort 3

Furthermore, univariate, and multivariate cox regression survival analysis was conducted to analyze the relevant hazard ratio, 95% confidence interval of HR, and p-value. According to the univariate analysis of discovery cohort 3, relapse-free survival was positively associated with ABAT (HR 0.420, CI 0.215-0.822, p-value=0.011, while RFS was negatively associated with grade (HR 2.357, CI 1.148-4.840, p-value= 0.020), and node classification (HR 1.974, CI 1.047-3.723, p-value= 0.036), shown in table 2. Moreover, the corresponding multivariable analysis showed that FRS was also significantly associated with tumor grade (HR 5.082, CI 1.089-23.709, p-value = 0.039), and node classification (HR 3.721, CI 1.266-10.936, P-value = 0.017).

Table 2: univariate and multivariate analysis of discovery cohort 3

Factors	Univariate analysis				Multivariate analysis
	HR	95% CI		P	HR	95% CI		P
	HR	Lower	Upper	P	HR	Lower	Upper	P
ABAT	0.420	0.215	0.822	0.011	0.313	0.072	1.349	0.119
Grade	2.357	1.148	4.840	0.020	5.082	1.089	23.709	0.039
N classification	1.974	1.047	3.723	0.036	3.721	1.266	10.936	0.017
Trimester	.962	0.562	1.646	.887	0.819	0.431	1.556	0.531
Age	.902	0.492	1.654	.739	2.043	0.683	6.107	0.683
Tumor size	1.170	0.624	2.193	0.625	1.459	1.459	0..496	0.496

Survival analysis of validation cohort 1 (GSE42568)

Kaplan-Meier survival analysis of validation cohort 1 (GSE42568), revealed several genes of 21 genes signature were associated with RFS and OS (Additional file 1, figure S1a, S1b). Expression of ABAT, PSAT1, and SRD5A1 was evaluated against both RFS and OS. Kaplan Meier survival analysis revealed that high expression of ABAT (P = 0.0001) (figure 2c) was associated with better RFS while overexpression of SRD5A1 (P = 0.023) (Figure 2d) and PSAT1 (P = 0.0034) (Figure 2e) was associated with poor RFS. Moreover, overexpression of ABAT (P = 0.00015) (figure 2f), was significantly associated with good OS whereas overexpression of SRD5A1 (P = 0.015) (figure 2g) PSAT1 (P = 0.037) was associated with poor OS (figure 2h).

Cox regression survival analysis of validation cohort 1

Univariable survival analysis of validation cohort 1 showed that overall survival was positively associated with ABAT (HR 0.529, CI 0.373-0.752, P = 0.000), and negative association with SRD5A1 (HR 2.277, CI 1.155-4.492, P = 0.018), PSAT1 (HR 1.417, CI 1.015-1.979, P = 0.041), grade (HR 3.960, CI 1.849-8.485, P = 0.000), node classification (HR 3.987, CI 1.739-9.143, P = 0.001), tumor size (HR 2.028, CI 1.030-3.996, P = 0.041) with 95 CI and hazard ratio presented in table 3. Furthermore, multivariable analysis also showed that ABAT (HR 0.574, CI 0.371-0.888, p = 0.013), grade (HR 2.440, CI 1.064-5.596, P= 0.035), and node classification (HR 3.036, CI 1.272-7.244, P = 0.012) was significantly correlated with overall survival.

Table 3: univariate and multivariate analysis of validation cohort 1

Factors	Univariate analysis				Multivariate analysis
	HR	95% CI		P	HR	95% CI		P
	HR	Lower	Upper	P	HR	Lower	Upper	P
ABAT	0.529	0.373	0.752	0.000	0.574	0.371	0.888	0.013
SRD5A1	2.277	1.155	4.492	0.018	1.334	0.656	2.714	0.426
PSAT1	1.417	1.015	1.979	0.041	0.854	0.564	1.304	0.472
Grade	3.960	1.849	8.485	0.000	2.440	1.064	5.596	0.035
N classification	3.987	1.739	9.143	0.001	3.036	1.272	7.244	0.012
Tumor size	2.028	1.030	3.996	0.041	1.521	0.750	3.083	0.245

Survival analysis of validation cohort 2 (TCGA-PAN Cancer)

Kaplan Meier survival analysis of TCGA data showed that several genes were associated with disease-free survival and overall survival (Additional file 1, figure S2a, S2b). Consistent with discovery and validation cohort 1, overexpression of ABAT (P = 0.0066) was also significantly associated with better disease-free survival (DFS) in validation cohort 2 (Figure 2i). Furthermore, overexpression of SRD5A1 (P = 0.019) was correlated with poor DFS (Figure 2j). Similarly, overexpression of PSAT1 (P = 0.039) were associated with poor OS (Figure 2k).

Cox regression survival analysis of validation cohort 2

According to the univariate analysis of validation cohort 2, ABAT (HR 0.744, CI 0.599-0.923, P= 0.007), SRD5A1 (HR 1.663, CI 1.082-2.556, P = 0.020), PSAT1 (HR 1.182, CI 1.008-1.387, P= 0.040), tumor classification (HR 0.760, CI 0.633-0.912, P= 0.003), node classification (HR 0.679, CI 0.560-0.823, P= 0.000), metastasis (HR 0.473, CI 0.366-0.612, P= 0.000), age (HR 1.794, CI 1.299-2.479, P= 0.000), and tumor stage (HR 0.609, CI 0.516-0.720, P= 0.000) was significantly associated with overall survival, shown in (Table 4). Moreover, multivariable analysis revealed that PSAT (HR 1.743, CI 1.142-2.658, P= 0.010), metastasis (HR 0.646, CI 0.481-0.867, P= 0.004), age (HR 2.075, CI 1.454-2.962, P= 0.000), and tumor stage (HR 0.325, CI 0.174-0.610, P= 0.000), are the independent prognostic factors associated with overall survival in the validation cohort 2.

Table 4: univariate and multivariate analysis of validation cohort 2 (TCGA data)

Factors	Univariate analysis				Multivariate analysis
	HR	95% CI		P	HR	95% CI		P
	HR	Lower	Upper	P	HR	Lower	Upper	P
ABAT	0.744	0.599	0.923	0.007	0.979	0.637	1.504	0.921
SRD5A1	1.663	1.082	2.556	0.020	1.210	0.814	1.798	0.346
PSAT1	1.182	1.008	1.387	0.040	1.743	1.142	2.658	0.010
T classification	0.760	0.633	0.912	0.003	1.108	0.820	1.872	0.700
N classification	0.679	0.560	0.823	0.000	1.138	0.634	2.043	0.665
Metastasis	0.473	0.366	0.612	0.000	0.646	0.481	0.867	0.004
Age	1.794	1.299	2.479	0.000	2.075	1.454	2.962	0.000
Tumor stage	0.609	0.516	0.720	0.000	0.325	0.174	0.610	0.000

In-vitro validation of ABAT, SRD5A1, and PSAT1

ABAT, SRD5A1, and PSAT1's relative mRNA expression was validated in MCF10A (normal breast epithelial cells) and SK-BR3, MDA-MB-231, as well as BT549 breast cancer cell lines by quantitative polymerase chain reaction. The mRNA expression of ABAT was significantly downregulated, while PSAT1 and SRD5A1 mRNA expression was significantly upregulated in all tested BC cells compared to control (MCF10A) (Figure 3). This data further suggested the role of ABAT as a good prognostic marker, whereas SRD5A1 and PSAT1 overexpression suggest their key role as poor prognostic markers.

Functional enrichment and pathway annotation

GO enrichment and pathway enrichment analysis were performed to better understand the function of genes common in all cohorts. In the GO biological process, lipid metabolic process, response to oxygen, cellular response to lipid, lipid biosynthetic process, response to inorganic compounds, and lipid metabolic process were the most enriched terms (figure 4a). KEGG pathway analysis showed that multiple genes were enriched in pathways in cancer, metabolic pathways, microRNAs in cancer, HIF-1 signaling, prostate, and breast cancer, (figure 4b). In Reactome pathway analysis, the metabolism of lipids and several oncogenic pathways were enriched, (figure 4c). Furthermore, in molecular function analysis, the top enriched terms were kinase binding, protein kinase, transcription factor, insulin-like growth factor-1, and cholesterol binding were the top enriched terms (figure 4d).

Previously several studies have revealed that lipid metabolism biomarkers are involved in breast cancer progression, however, a comprehensive evaluation of lipid prognostic biomarkers has not been analyzed to identify their clinical importance. This is the first prognostic study based on clinicopathological parameters and lipid metabolic biomarkers expression profiles in breast cancer patients. In this study, we have screened 114 lipid metabolic genes out of 1035. In these 114 lipid metabolism genes, 21 gene signatures were significantly associated with tumor grade and other clinical factors in all discoveries and validation cohorts. GO analysis showed a significant association of identified genes with lipid metabolism, cancer cell signaling, cell proliferation, lipid biosynthetic process, and response to estradiol. Similarly, KEGG pathway analysis showed that all these genes were enriched in different cancers including breast cancer, and metabolic as well as other signaling pathways. Our analysis suggested that lipid metabolism genes including ABAT, SRD5A1, and PSAT1 as novel prognostic markers and predictors of breast cancer survival.

ABAT (4-aminobutyrate transaminase) is a protein-coding gene involved in the metabolism of pyruvate and catabolism of leucine, isoleucine, and valine (Chen, Jiao et al. 2018). Clinical analysis of all discovery cohorts indicated that ABAT was downregulated in ER and PR (P < 0.05) negative samples whereas overexpressed in early grade < 2, tumor size < 2, early node, residual, pregnancy status, and HER2-negative (P < 0.05) samples (Additional file 1, figure S3). Furthermore, Kaplan Meier survival revealed that high expression of ABAT is associated with a positive prognosis. Consistent with our results, Budczies, J. et al found that ABAT catalyzes degradation of β-Alanine to malonate semialdehyde enzyme. Then, malonate semialdehyde is converted to acetyl-CoA by simultaneous reduction of NAD + to NADH. These enzymatic reactions are downregulated in ER- breast cancer and are responsible for the induced level of β-Alanine ER-negative breast cancer subtype. Moreover, Kaplan–Meier analysis of large datasets showed that ABAT transcript is a positive prognostic factor for breast cancer. Low expression of ABAT transcript was associated with high β-Alanine concentration and low ABAT protein expression (Budczies, Brockmöller et al. 2013). Another recent study stated that high expression of ABAT was observed with good survival in ccRCC patients. Overexpression of ABAT or ALDH6A1decreased lactate production, whereas ABAT upregulation also decreases the ratio of NADPH/NADP which impairs cells oncologic metabolism (Lu, Chen et al. 2020). Invitro mRNA expression by RT-qPCR showed downregulation of ABAT mRNA in SK-BR3, MDA-MB-231, and BT-549 breast cancer cell line compared to normal BC cells. Similarly, Chen, X., et al, reported downregulation of ABAT mRNA and protein expression in triple-negative BC cells. These findings suggested that the expression of ABAT inhibits metastasis and tumorigenicity both in vivo and in vitro. Hence, loss of ABAT expression is contributed to the aggressiveness of BLBC model through activation of the Ca2 + NFAT1 axis. Loss of ABAT expression was strongly associated with metastasis, high grade, large tumor size, chemotherapy resistance, and poor survival in breast cancer patients (Chen, Cao et al. 2019). A recent study conducted on LUAD patients showed that high expression of ABAT was associated with longer overall survival (Wang, Pei et al. 2021). Another Study of twelve datasets showed that low expression of ABAT revealed a negative association with metastasis and the aggressive nature of breast cancer as well as correlated with a good prognosis of breast cancer patients (Kim 2021). So far, in this study, we identified the ABAT gene as a novel lipid metabolism biomarker that is downregulated in BC and associated with a good prognosis.

High expression of SRD5A1 was associated with ER, PR-negative samples, and early age < 52 (P < 0.05), while downregulated in early grade, early node, non-pregnant status, and residual (RD) status (P < 0.05) (Additional file 1, figure S4). Both SRD5A1 and SRD5A2 induce the inactivation of cortisol to 5-dihydrocortisol, which is further converted into 5-tetrahydrocortisol a non-rate limiting step by 3-hydroxysteroid dehydrogenase. SRD5A1 is highly expressed in human adipocytes and hepatocytes to regulate lipid metabolism (Thigpen, Silver et al. 1993, Wake, Strand et al. 2007). The important male androgens are testosterone, produced by Leydig cells, and DHT derived from SRD5A-induced catabolism of testosterone in peripheral tissues. The expression of SRD5A1 in several available datasets is upregulated in primary and metastatic PCa compared to non-tumor prostate tissues (Audet-Walsh, Yee et al. 2017). In this model, the mRNA expression of SRD5A1 was found upregulated in three independent BC cell lines as compared to normal cells. Furthermore, high expression of SRD5A1 was associated with BC subtypes, poor survival, and downregulated in early tumor grade. Consistent with our results, the differential expression of SRD5A1 in different PCa grades patients is suggestive of a deregulated androgenic signal-mediated oncogenicity in PCa patients. TCGA data analysis revealed that high expression of SRD5A1, HSD17B2, and SHBG of metabolic tetrad were associated with poor overall survival (Bamodu, Tzou et al. 2021). Statistical analysis of SRD5A1 showed a significant association with non-pregnant status, interestingly, a case-control study based on a large population indicated that genetic variations in SRD5A1 affect postmenopausal-breast cancer risk patients using combined therapy (Hein, Abbas et al. 2012). Recently, another study of gene expression profiling of multiple myeloma indicated that SRD5A1 aberrant expression was associated with poor prognosis and disease progression. This oncogenic role of SRD5A1 was investigated in ARP1 and H929 myeloma cell lines as well as multiple myeloma xenograft models. Inhibition of SRD5A1 regulated the autophagy through PI3K/Akt/ mTOR signaling pathway and Bcl-2 family protein-induced apoptosis in MM cells (Dou, Qian et al. 2021). Furthermore, Serum and tissue concentration of androgen was measured in two datasets of patients, indicating that SRD5A1 was an independent prognostic factor in endometrioid endometrial adenocarcinoma (Tanaka, Miki et al. 2015). Nasiri et al revealed that SRD5A1expression in mouse liver causes the pathogenesis of the metabolic disease. Moreover, genetic alteration in SRD5A1 increases NAFLD and lipid accumulation (Nasiri, Nikolaou et al. 2015).

PSAT1 was overexpressed in ER negative, PR negative samples, while downregulated in early tumor grade as well as tumor size < 2, HER2 subtypes, and residual (RD) status (Additional file 1, figure S5). Serine is a type of non-essential amino acid required for cancer cell proliferation not only for the synthesis of protein but also important for the synthesis of nucleotides, antioxidants, lipids, and other amino acids (Mattaini, Sullivan et al. 2016). PSAT1 is a member of the serine synthesis pathway which is mostly depleted in luminal breast cancer compared to basal tumors. low expression of the PSAT1 gene inhibits de novo serine biosynthesis and sensitizes luminal breast tumor cells to glycine and serine starvation both in vivo and in vitro (Choi, Rawat et al. 2022). A recent study indicated that ADH1C upregulation inhibited migration, growth, colony formation, and migration of CRC cell lines and inhibited the growth of xenograft tumors in nude mice. This inhibition of tumorigenesis by ADH1C is due to its effect of reducing the expression of PSAT1/PHGDH and serine level (Li, Yang et al. 2022). Results of our model suggested that PSAT1 was upregulated both in BC patients and cell lines and downregulated in early tumor grade. Overexpression of PSAT1 was associated with poor OS, RFS, and DFS. Like our findings, whole-genome expression profiling of isogenic cell line, MDA-MB-231(SA), and parental MDA-MB-231 revealed that PSAT1 was upregulated in metastatic variant, while overexpression of PSAT1 in primary breast cancer was significantly correlated with the malignant phenotypic characteristic of breast tumors and poor relapse-free and overall survival of breast cancer patients (Pollari, Käkönen et al. 2011). Similarly, Liao, et al identified that high expression of the PSAT1 gene was associated with advanced tumor stage and poor prognosis in a dataset of 124 nasopharyngeal carcinoma patients (Liao, Chao et al. 2016). Global test analysis revealed that the expression of PSAT1 is associated with JAK-STAT and cytokine signaling pathways (De Marchi, Timmermans et al. 2017). Tissue microarray of 297 ER-negative breast cancer samples was analyzed by immunohistochemistry assay which demonstrated that overexpression of the PSAT1 gene predicted poor prognosis (Gao, Ge et al. 2017). In another study the role of PSAT1 was investigated in the invasion and migration of TNBC and suggested it as a potential biomarker, also PSAT1 expression increases with the clinical grade of TNBC (Metcalf, Dougherty et al. 2020). Lipid-induced stress enhances the expression of ATF4 which in turn control PSAT1, PHGDH, and SPH expression to influence epigenetics through the production of S-adenosyl methionine. This mechanism of lipid-induced stress is associated with the transformation of non-transformed breast cells into malignant cells (Eduardo, Yadav et al. 2022). Yang et al suggested that PSAT1 induces non-small cell lung cancer cell proliferation by regulating mTOR pathway which is an important molecular sensor for nutrient availability and links with serine biosynthesis (Yang, Wu et al. 2015).

In conclusion, we identified ABAT, PSAT1, and SRD5A1 genes as the most significant BC biomarkers based on lipid metabolism. The expression of these three genes can be used to classify BC into different prognostic subtypes. Regardless of different shortcomings in the current study, these lipid biomarkers may not only play an important role in predicting prognosis but also provide an insight into the fundamental mechanism of lipid metabolism in tumorigenesis and investigation of more biomarkers for BC treatment.

DEGs

Differentially expressed genes

BC

Breast cancer

PCa

Prostate cancer

ABAT

4-aminobutyrate transaminase

SRD5A1

Steroid 5 Alpha-Reductase 1

PSAT1

Phosphoserine Aminotransferase 1

GO

Gene ontology

RFS

Relapse free survival

DFS

Disease free survival

OS

Overall survival

GEO

Gene expression omnibus

TCGA

The cancer genome atlas.

Ethical approval and consent to participate

Not Applicable.

Consent for publication

Not applicable.

Authors contributions

ZD, BC performed formal analysis and interpretation; FH, FP and ZD designed the project and analyzed the data; CW and HG helped with data analysis; FH wrote some part of codes in R language; FP, FH, and QL, made contributions to the final revision; FP, FH, and QL guided the research, revised the whole manuscript and final approval of the manuscript; ZD wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding:

This research work was supported by the National Natural Science Foundation of China (No. 81820108024 to QL, No. 82003141 to FP, No. 82002960 to BC), National Key R&D Program of China (2019YFA0110300 to QL), Program for Changjiang Scholars and Innovative Research Team in University of Ministry of Education of China (No. IRT_17R15), the Science and Technology Innovation Foundation of Dalian (No.2020JJ25CY008 to QL), the Free Exploration Basic Research Program for the Central Government Guiding Local Funding of Scientific and Technological Development (2021 to FP), Dalian High-level Talents Innovation Support Program-Young Science and Technology Star (2021RQ004 to BC), the "Seedling cultivation" Program for Young Scientific and Technological Talents of Liaoning (No. LZ2020044 to FP).

Data Availability Statement: Multiple publicly available datasets are used in the study. The details are provided in the methodology section where necessary.

Competing interests: The authors declare no competing interests.

Acknowledgements:

We are thankful to Muhammad Faraz Arshad Malik for his contribution to our manuscript. We are also thankful to International Education College of Dalian Medical University and National Natural Science Foundation of China for their support.

Audet-Walsh, É., T. Yee, I. S. Tam and V. Giguère (2017). "Inverse regulation of DHT synthesis enzymes 5α-reductase types 1 and 2 by the androgen receptor in prostate cancer." Endocrinology 158(4): 1015–1021.
Bamodu, O. A., K.-Y. Tzou, C.-D. Lin, S.-W. Hu, Y.-H. Wang, W.-L. Wu, K.-C. Chen and C.-C. Wu (2021). "Differential but concerted expression of HSD17B2, HSD17B3, SHBG and SRD5A1 testosterone tetrad modulate therapy response and susceptibility to disease relapse in patients with prostate cancer." Cancers 13(14): 3478.
Baumann, J., C. Sevinsky and D. S. Conklin (2013). "Lipid biology of breast cancer." Biochimica et Biophysica Acta (BBA)-Molecular and Cell Biology of Lipids 1831(10): 1509–1517.
Bianchini, G., J. M. Balko, I. A. Mayer, M. E. Sanders and L. Gianni (2016). "Triple-negative breast cancer: challenges and opportunities of a heterogeneous disease." Nature reviews Clinical oncology 13(11): 674–690.
Bray, F., J. Ferlay, I. Soerjomataram, R. L. Siegel, L. A. Torre and A. Jemal (2018). "Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries." CA: a cancer journal for clinicians 68(6): 394–424.
Budczies, J., S. F. Brockmöller, B. M. Müller, D. K. Barupal, C. Richter-Ehrenstein, A. Kleine-Tebbe, J. L. Griffin, M. Orešič, M. Dietel and C. Denkert (2013). "Comparative metabolomics of estrogen receptor positive and estrogen receptor negative breast cancer: alterations in glutamine and beta-alanine metabolism." Journal of proteomics 94: 279–288.
Chen, B., Z. Jiao, X. Yin, Z. Qian, J. Gu and H. Sun (2018). "Novel insights into biomarkers associated with renal cell carcinoma." Oncology Letters 16(1): 83–90.
Chen, X., Q. Cao, R. Liao, X. Wu, S. Xun, J. Huang and C. Dong (2019). "Loss of ABAT-mediated GABAergic system promotes basal-like breast cancer progression by activating Ca2+-NFAT1 axis." Theranostics 9(1): 34.
Choi, B.-H., V. Rawat, J. Högström, P. A. Burns, K. O. Conger, M. E. Ozgurses, J. M. Patel, T. S. Mehta, A. Warren and L. M. Selfors (2022). "Lineage-specific silencing of PSAT1 induces serine auxotrophy and sensitivity to dietary serine starvation in luminal breast tumors." Cell reports 38(3): 110278.
Cianfrocca, M. and L. J. Goldstein (2004). "Prognostic and predictive factors in early-stage breast cancer." The oncologist 9(6): 606–616.
Curtis, C., S. P. Shah, S.-F. Chin, G. Turashvili, O. M. Rueda, M. J. Dunning, D. Speed, A. G. Lynch, S. Samarajiwa and Y. Yuan (2012). "The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups." Nature 486(7403): 346–352.
De Marchi, T., M. A. Timmermans, A. M. Sieuwerts, M. Smid, M. P. Look, N. Grebenchtchikov, F. C. Sweep, J. G. Smits, V. Magdolen and C. H. van Deurzen (2017). "Phosphoserine aminotransferase 1 is associated to poor outcome on tamoxifen therapy in recurrent breast cancer." Scientific reports 7(1): 1–11.
Denkert, C., C. Liedtke, A. Tutt and G. von Minckwitz (2017). "Molecular alterations in triple-negative breast cancer—the road to new treatment strategies." The Lancet 389(10087): 2430–2442.
Dignam, J. J., V. Dukic, S. J. Anderson, E. P. Mamounas, D. L. Wickerham and N. Wolmark (2009). "Hazard of recurrence and adjuvant treatment effects over time in lymph node-negative breast cancer." Breast cancer research and treatment 116(3): 595–602.
Dou, R., J. Qian, W. Wu, Y. Zhang, Y. Yuan, M. Guo, R. Wei, S. Yang, A. Jurczyszyn and S. Janz (2021). "Suppression of steroid 5α-reductase type I promotes cellular apoptosis and autophagy via PI3K/Akt/mTOR pathway in multiple myeloma." Cell death & disease 12(2): 1–13.
Du, T., M. J. Sikora, K. M. Levine, N. Tasdemir, R. B. Riggins, S. G. Wendell, B. Van Houten and S. Oesterreich (2018). "Key regulators of lipid metabolism drive endocrine resistance in invasive lobular breast cancer." Breast Cancer Research 20(1): 1–15.
Eduardo, M. B., S. Yadav, S. Khan and S. Clare (2022). "Abstract P5-05-05: Unraveling the mechanism of Lipid-induced malignant transformation in non-transformed breast epithelial cells." Cancer Research 82(4_Supplement): P5-05-05-P05-05-05.
Gao, S., A. Ge, S. Xu, Z. You, S. Ning, Y. Zhao and D. Pang (2017). "PSAT1 is regulated by ATF4 and enhances cell proliferation via the GSK3β/β-catenin/cyclin D1 signaling pathway in ER-negative breast cancer." Journal of Experimental & Clinical Cancer Research 36(1): 1–13.
Glatz, J. F., J. J. Luiken and A. Bonen (2010). "Membrane fatty acid transporters as regulators of lipid metabolism: implications for metabolic disease." Physiological reviews.
Hein, R., S. Abbas, P. Seibold, R. Salazar, D. Flesch-Janys and J. Chang-Claude (2012). "Polymorphism Thr160Thr in SRD5A1, involved in the progesterone metabolism, modifies postmenopausal breast cancer risk associated with menopausal hormone therapy." Breast cancer research and treatment 131(2): 653–661.
Hilvo, M., C. Denkert, L. Lehtinen, B. Müller, S. Brockmöller, T. Seppänen-Laakso, J. Budczies, E. Bucher, L. Yetukuri and S. Castillo (2011). "Novel theranostic opportunities offered by characterization of altered membrane lipid metabolism in breast cancer progression." Cancer research 71(9): 3236–3245.
Hultsch, S., M. Kankainen, L. Paavolainen, R.-M. Kovanen, E. Ikonen, S. Kangaspeska, V. Pietiäinen and O. Kallioniemi (2018). "Association of tamoxifen resistance and lipid reprogramming in breast cancer." BMC cancer 18(1): 1–14.
Kim, J. (2021). "In silico analysis of differentially expressed genesets in metastatic breast cancer identifies potential prognostic biomarkers." World Journal of Surgical Oncology 19(1): 1–15.
Koundouros, N. and G. Poulogiannis (2020). "Reprogramming of fatty acid metabolism in cancer." British journal of cancer 122(1): 4–22.
Li, S., H. Yang, W. Li, J.-y. Liu, L.-w. Ren, Y.-h. Yang, B.-b. Ge, Y.-z. Zhang, W.-q. Fu and X.-j. Zheng (2022). "ADH1C inhibits progression of colorectal cancer through the ADH1C/PHGDH/PSAT1/serine metabolic pathway." Acta Pharmacologica Sinica: 1–14.
Liao, K.-M., T.-B. Chao, Y.-F. Tian, C.-Y. Lin, S.-W. Lee, H.-Y. Chuang, T.-C. Chan, T.-J. Chen, C.-H. Hsing and M.-J. Sheu (2016). "Overexpression of the PSAT1 gene in nasopharyngeal carcinoma is an indicator of poor prognosis." Journal of Cancer 7(9): 1088.
Lu, J., Z. Chen, H. Zhao, H. Dong, L. Zhu, Y. Zhang, J. Wang, H. Zhu, Q. Cui and C. Qi (2020). "ABAT and ALDH6A1, regulated by transcription factor HNF4A, suppress tumorigenic capability in clear cell renal cell carcinoma." Journal of translational medicine 18(1): 1–15.
Mattaini, K. R., M. R. Sullivan and M. G. Vander Heiden (2016). "The importance of serine metabolism in cancer." Journal of Cell Biology 214(3): 249–257.
Menendez, J. and R. Lupu (2017). "Fatty acid synthase regulates estrogen receptor-α signaling in breast cancer cells." Oncogenesis 6(2): e299-e299.
Metcalf, S., S. Dougherty, T. Kruer, N. Hasan, R. Biyik-Sit, L. Reynolds and B. F. Clem (2020). "Selective loss of phosphoserine aminotransferase 1 (PSAT1) suppresses migration, invasion, and experimental metastasis in triple negative breast cancer." Clinical & experimental metastasis 37(1): 187–197.
Nasiri, M., N. Nikolaou, S. Parajes, N. P. Krone, G. Valsamakis, G. Mastorakos, B. Hughes, A. Taylor, I. J. Bujalska and L. L. Gathercole (2015). "5α-reductase type 2 regulates glucocorticoid action and metabolic phenotype in human hepatocytes." Endocrinology 156(8): 2863–2871.
Paoletti, C. and D. F. Hayes (2014). "Molecular testing in breast cancer." Annual review of medicine 65: 95–110.
Pollari, S., S.-M. Käkönen, H. Edgren, M. Wolf, P. Kohonen, H. Sara, T. Guise, M. Nees and O. Kallioniemi (2011). "Enhanced serine production by bone metastatic breast cancer cells stimulates osteoclastogenesis." Breast cancer research and treatment 125(2): 421–430.
Siegel, R. L., K. D. Miller, H. E. Fuchs and A. Jemal (2022). "Cancer statistics, 2022." CA: a cancer journal for clinicians.
Song, L., Z. Liu, H.-H. Hu, Y. Yang, T. Y. Li, Z.-Z. Lin, J. Ye, J. Chen, X. Huang and D.-T. Liu (2020). "Proto-oncogene Src links lipogenesis via lipin-1 to breast cancer malignancy." Nature communications 11(1): 1–13.
Tanaka, S., Y. Miki, C. Hashimoto, K. Takagi, Z. Doe, B. Li, N. Yaegashi, T. Suzuki and K. Ito (2015). "The role of 5α-reductase type 1 associated with intratumoral dihydrotestosterone concentrations in human endometrial carcinoma." Molecular and Cellular Endocrinology 401: 56–64.
Thigpen, A. E., R. I. Silver, J. M. Guileyardo, M. L. Casey, J. McConnell and D. W. Russell (1993). "Tissue distribution and ontogeny of steroid 5 alpha-reductase isozyme expression." The Journal of clinical investigation 92(2): 903–910.
Wake, D. J., M. Strand, E. Rask, J. Westerbacka, D. E. Livingstone, S. Soderberg, R. Andrew, H. Yki-Jarvinen, T. Olsson and B. R. Walker (2007). "Intra‐adipose sex steroid metabolism and body fat distribution in idiopathic human obesity." Clinical endocrinology 66(3): 440–446.
Wang, Z., H. Pei, H. Liang, Q. Zhang, L. Wei, D. Shi, Y. Chen and J. Zhang (2021). "Construction and Analysis of a circRNA-Mediated ceRNA Network in Lung Adenocarcinoma." OncoTargets and therapy 14: 3659.
Ward, A. V., S. M. Anderson and C. A. Sartorius (2021). "Advances in Analyzing the Breast Cancer Lipidome and Its Relevance to Disease Progression and Treatment." Journal of mammary gland biology and neoplasia: 1–19.
Yang, Y., J. Wu, J. Cai, Z. He, J. Yuan, X. Zhu, Y. Li, M. Li and H. Guan (2015). "PSAT1 regulates cyclin D1 degradation and sustains proliferation of non-small cell lung cancer cells." International journal of cancer 136(4): E39-E50.

No competing interests reported.

Additionalfile1.docx
Additional file 1: Table S1: lipid metabolism genes and their pathways extracted from MSigDB.Additional file 1: Table S2: Primers sequence ABAT, SRD5A1, and PSAT1.Additional file 1: Table S3: Genes associated with tumor grade in discovery cohort 1.Additional file 1: Table S4: Genes associated with tumor grade in discovery cohort 2.Additional file 1: Table S5: Genes associated with tumor grade in discover cohort 3.Additional file 1: Figure S1. Kaplan Meier survival analysis of BC patients in validation cohort 1. (a) Relapse-free survival (b) Overall survival.Additional file 1: Figure S2. Kaplan Meier survival analysis of BC patients in validation cohort 1. (a) Disease-free survival (b) Overall survival.Additional file 1: Figure S3. The association of ABAT expression with clinicopathological parameters in all discovery cohorts.Additional file 1: Figure S4. The association of SRD5A1 expression with different clinicopathological features in all discovery cohorts.Additional file 1: Figure S5. The association of PSAT1 expression with clinicopathological parameters in all discovery cohorts.
Additionalfile2.xlsx
Additional file 2: Table S1: Genes associated with clinical features in discovery cohort 1 (GSE50948).Additional file 2: Table S2: Genes associated with clinical features in discovery cohort 2 (GSE20194)Additional file 2: Table S3: Genes associated with clinical features in discovery cohort 3 (GSE53031)

Transcriptomic analysis identified new lipid biomarkers for patient stratification in breast cancer

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Introduction

Materials And Methods

Data collection and screening

Cell Culture

Rna Extraction And Rt-qpcr

Statistical analysis

Results

Discussion

Conclusion

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1