SIRT5 downregulation is associated with poor prognosis in glioblastoma

Abstract

OBJECTIVE:

Sirtuins (SIRT) are NAD ${}^{+}$ -dependent protein deacetylases that are involved in the regulation of cancer-associated pathways. However, the biological role of these deacetylases remains elusive in glioblastoma (GBM). Here, we evaluated the effects of 7 sirtuins regarding their occurrence and prognostic value for GBM.

METHODS:

In this research, the effects of SIRT5 on the occurrence and prognosis of GBM were evaluated using integrative bioinformatics analyses.

RESULTS:

Based on comprehensive analyses of data obtained from web-based bioinformatics platforms, the data demonstrate that only SIRT5 expression is statistically decreased in GBM tissues. The clinical relevance analysis shows that downregulation of SIRT5 is significantly correlated with a shorter survival time. Moreover, the expression levels of SIRT5 were confirmed to be negatively associated with DNA methylation status. In addition, a protein-protein interaction network was constructed to determine the relationship of genes coexpressed with SIRT5. Functional enrichment analysis revealed that SIRT5 was potentially involved in epithelial-mesenchymal transition and in regulating cell communications.

CONCLUSIONS:

Collectively, our results indicate that SIRT5 acts as a potential suppresser during tumorigenesis, and suggest that SIRT5 may be a promising prognostic biomarker of GBM.

1. Introduction

Glioma is the most common primary malignant brain tumor, accounting for approximately 80% of all primary brain tumors. Among the histological subtypes of gliomas, glioblastoma (GBM) is the most common and most lethal form, with only 5.5% patients surviving beyond 5 years after the initial diagnosis [1]. Despite of the improvements in therapeutic strategies, such as surgery, external radiotherapy and chemotherapy, the prognosis of GBM patients remains undesirable [2, 3]. In consideration of the challenges for clinical treatment, novel prognostic biomarkers and therapeutic targets are urgently needed.

Table 1

The main bioinformatics tools used to analyze the functions of SIRT5 in glioblastoma

Database	Samples	URL	Refs
GEPIA	Tissues	http://gepia.cancer-pku.cn/	[11]
GlioVis	Tissues	http://gliovis.bioinfo.cnio.es/	[12]
UALCAN	Tissues	http://ualcan.path.uab.edu/index.html	[13]
Wanderer	Tissues	http://maplab.imppc.org/wanderer/	[14]
DiseaseMeth	Tissues	http://bio-bigdata.hrbmu.edu.cn/diseasemeth/	[15]
MethSurv	Tissues	https://biit.cs.ut.ee/methsurv/	[16]
STRING	–	http://string-db.org/	[17]
Cytoscape	–	https://cytoscape.org/	[18]
FunRich	–	http://www.funrich.org	[19]

Figure 1.

Analysis of the SIRT5 expression levels in GBM tissues. A–C. The mRNA expression of 7 sirtuins was examined using three different datasets from the GEPIA, GlioVis and UALCAN platforms, respectively.

Figure 2.

Relationship between SIRT5 expression and the clinical characteristics of GBM patients. A–C. Cancer genome atlas (TCGA) clinical data from the UALCAN web-based tool was used to categorize GBM patient characteristics based on their expression levels of SIRT5, such as gender, age, race, and so on. D. Kaplan-Meier analysis of overall survival (OS) according to SIRT5 expression.

Sirtuins (SIRT), homologs of the yeast SIR2 gene, that predominately function as nicotinamide adenine dinucleotide (NAD ${}^{+}$ )-dependent deacetylases, are known to regulate a wide range of biological process in ageing and age-associated diseases [4, 5]. Recently, an increasing number of studies have shown that seven mammalian sirtuins (SIRT1-7) are linked to tumor carcinogenesis. In GBM, the role of sirtuins is controversial as distinct sirtuin isoforms show opposing functions of either tumor suppression or promotion. Upregulation of SIRT1, -2, and -6 expression is associated with tumor suppression. Upregulated SIRT1 can facilitate cell senescence via FOXO1 [6] and induce autophagic GBM cell death via the AMPK-mTOR-ULK complex [7]. Downregulation of SIRT2 expression or pharmacological blockade of its enzymatic activity counteracts the cell cycle arrest induced by resveratrol [8]. Overexpression of SIRT6 suppresses U87-MG glioma cell growth and induces apoptosis via reduction of oxidative stress and inhibition of JAK2/STAT3 activation [9]. On the contrary, SIRT3 can act as tumor promoter since activation of SIRT3 deacetylation activity is induced after therapeutic irradiation, enhancing tumor resistance [10]. Although SIRT4, -5, and -7 have been explored in various cancers, there is no evidence of their biological role in GBM, which needs to be further investigated.

In the present study, we considered the putative role of sirtuins as therapeutic targets and prognostic biomarkers of GBM and evaluated the function and mechanism of 7 sirtuins using bioinformatics web-based resources. Interestingly, our data reveal that only the expression of SIRT5 was significantly downregulated in GBM, which was consistently shown in 3 different datasets. We also show that downregulation of SIRT5 in GBM is correlated with poor prognosis. DNA methylation status is negatively associated with the expression of SIRT5. Moreover, coexpressed network and functional enrichment analyses are performed to understand the potential biological pathways and functional roles of SIRT5.

2. Methods and materials

2.1 Data acquisition and reanalysis using different bioinformatics methods

The expression levels of SIRT5 and other sirtuins in GBM tissues were examined using several bioinformatics web-based resources. These tools are summarized in Table 1. Gene Expression Profiling Interactive Analysis (GEPIA) is a web-based tool that has fast and customizable functionalities, including differential expression analysis across tumor and normal samples [11]. In GEPIA, 163 tumor samples from The Cancer Genome Atlas (TCGA) and 207 GTEx normal samples were integrated for very comprehensive expression analysis of SIRT family. GlioVis, which contains a large collection of brain tumor entries, allows researchers to explore specific gene expression [12]. In GlioVis, a dataset (TCGA_GBM) with 528 tumor samples and 10 normal samples was used to analyze the expression level of SIRT family. UALCAN is a user friendly and interactive web-based application for analyzing the relative expression of given genes based on comparisons between normal and cancer tissues [13]. In UALCAN, the GBM transcriptome data (156 GBM samples and 5 TCGA normal samples from TCGA) were provided to evaluate the expression profiles of SIRT family. Thus, according to the data available in these open bioinformatics platforms, 7 sirtuins expression patterns were potently identified in GBM tissues.

The relevance between clinicopathological parameters and SIRT5 expression was evaluated by UALCAN and GlioVis. Using the 156 GBM transcriptome sequencing data from TCGA, UALCAN was used to estimate the associations between gene expression and various clinical parameters of GBM patients. Through GlioVis patterns, survival analysis was performed on data collected from the dataset (TCGA_GBM).

Analysis of the correlation between DNA methylation and SIRT5 expression was performed using Wanderer, an intuitive web-based tool that contains DNA methylation profiles from TCGA [14]. In Wanderer, the 155 GBM samples with methylation data (infinium HumanMethylation 450 BeadChip, HM450K) were downloaded from the dataset Glioblastoma (TCGA, Provisional). In addition, Mining of the differential DNA methylation status as investigated using DiseaseMeth [15], which provides a cross-data set analysis between GBM and normal tissues. In DiseaseMeth, the 288 GBM samples with methylation data (infinium HumanMethylation 27 BeadChip, HM27K) were downloaded from the dataset Glioblastoma (TCGA, Nature 2008). And MethSurv web portal [16] was also used to perform the survival analysis based on the DNA methylation data. In MethSurv, the 139 GBM samples with methylation data (HM450K) were downloaded from the dataset Glioblastoma (TCGA March 2017).

The data (TCGA_GBM) from GlioVis was used to investigate the genes that were coexpressed with SIRT5 and map the differential expression profile. The protein-protein association data of the SIRT5 interaction partners was collected and integrated from the STRING database [17] and Cytoscape [18]. Using the FunRich tool [19], functional enrichment analysis was conducted to identify genes coexpressed with SIRT5.

2.2 Statistical analyses

Differential mRNA expression of between cancerous and non-cancerous tissues was analyzed using Student’s $t$ -test with the statistical package for the social sciences (SPSS) 22.0 software (IBM Analytics, USA). Chi-square tests were performed to analyze the associations between SIRT5 expression and clinicopathologic characteristics. Kaplan-Meier analysis was used to analyze the survival of GBM patients. Correlations between genes were assessed by the Pearson correlation coefficient. Statistic significant was defined as $p<$ 0.05.

Table 2

The higher methylation values of CpG sites in GBM samples compared with normal tissues

Probe	Norm_mean	Norm_sd	Tum_mean	Tum_sd	Fold-change	$P$ _val
cg02206980	0.095018	0.019713	0.20232	0.131147	2.12928	0.145732
cg01064902	0.028836	0.012943	0.046472	0.049134	1.611585	0.424999
cg19366958	0.033466	0.001278	0.039504	0.019421	1.180428	0.659123
cg11824369	0.015073	0.001349	0.01776	0.004885	1.178242	0.315252
cg06139021	0.033666	0.014309	0.039063	0.009072	1.16033	0.566967
cg20150972	0.01707	0.001295	0.019534	0.005598	1.144342	0.632175
cg02712746	0.022937	0.001456	0.024985	0.006498	1.089324	0.829089
cg07909422	0.077107	0.003084	0.083126	0.015007	1.078069	0.632175
cg08619119	0.049397	0.001273	0.051729	0.012522	1.047204	0.858466
cg10907941	0.021131	0.001547	0.021865	0.003454	1.034705	0.659123
cg07534331	0.047851	0.017984	0.04831	0.020266	1.009599	0.873231
cg26753137	0.049065	0.001032	0.04909	0.009506	1.000507	0.902888

Table 3

The survival analysis of GBM patients with different methylation values of CpG sites

Probe	HR	CI	Wald_ $P$ value		Current_split	Mean	Median	Range
cg02206980	0.666	(0.409; 1.085)	0.	1	q75	0.2	0.16	(0.034; 0.697)
cg01064902	0.825	(0.52; 1.31)	0.	42	q25	0.05	0.04	(0.017; 0.387)
cg19366958	0.768	(0.488; 1.21)	0.	25	q25	0.04	0.04	(0.022; 0.208)
cg11824369	0.887	(0.564; 1.397)	0.	61	q25	0.02	0.02	(0.011; 0.052)
cg06139021	0.721	(0.441; 1.178)	0.	19	q75	0.04	0.04	(0.021; 0.087)
cg20150972	0.697	(0.415; 1.172)	0.	17	q75	0.02	0.02	(0.011; 0.039)
cg02712746	0.793	(0.523; 1.203)	0.	27	maxstat	0.03	0.02	(0.015; 0.053)
cg07909422	0.882	(0.553; 1.405)	0.	6	q75	0.08	0.08	(0.055; 0.129)
cg08619119	1.692	(1.121; 2.556)	0.	012 ${}^{*}$	maxstat	0.05	0.05	(0.024; 0.095)
cg10907941	0.833	(0.529; 1.311)	0.	43	q25	0.02	0.02	(0.014; 0.032)
cg07534331	1.516	(0.975; 2.358)	0.	065	maxstat	0.05	0.04	(0.024; 0.175)
cg26753137	0.764	(0.506; 1.153)	0.	2	median	0.05	0.05	(0.024; 0.077)
Note: ${}^{*}$ indicated the statistic significance.

Figure 3.

Associations between SIRT5 expression and DNA methylation in GBM patients. A. Difference of the DNA methylation status in GBM and normal tumors based on DiseaseMeth. B. The Kaplan-Meier Plotter was used to quickly confirm overall survival in the patients with high or low level of methylation values at a specific CpG site (cg08619119). C. The DNA methylation level of one CgG site (cg08619119) is negatively associated with SIRT5 transcriptional expression level according to the dataset from Wanderer.

Figure 4.

Analysis of genes that coexpressed with SIRT5 in GBM. A–B. Heatmap and volcano plot showing the differentially genes that coexpressed with SIRT5 in GBM via GlioVis. C. The PPI network of the SIRT5 interaction partners according to the STRING database and Cytoscape. D–E. FunRich software identified the biological roles of the coexpressed genes in GBM.

3. Results

3.1 SIRT5 is identified to be downregulated in glioblastoma tissues

To estimate the different expression levels of 7 sirtuins members between GBM and normal tissues, we used three independent bioinformatics databases to identify the sirtuins expression profiles. First, the transcript levels of 7 sirtuins were determined from the GEPIA platform: SIRT1, -2, -6, and -7 were upregulated while SIRT3, -4, and -5 were downregulated (Fig. 1A). The next validation was carried out via the GlioVis and UALCAN tools to verify whether the expression trends were in accordance with data from other platforms. The results from the GlioVis database showed that the mRNA expression levels for SIRT3 and SIRT5 were significantly downregulated in GBM tissue (Fig. 1B) ( $p<$ 0.001; $p<$ 0.001, respectively), whereas SIRT6 and SIRT7 were significantly upregulated in GBM tissue ( $p=$ 0.006; $p=$ 0.001, respectively), which are consistent with the trends shown in Fig. 1A. No apparent significant differences were observed for SIRT1, -2, and -4. However, we only confirmed the downregulation of SIRT5 transcripts from UALCAN ( $p<$ 0.001) (Fig. 1C), while there was no significant difference between GBM and normal tissues in other sirtuins. Thus, based on analysis of the data from three different databases, we identified decreased expression of SIRT5 in GBM tissue, which indicated its potential role in tumor suppression.

3.2 SIRT5 as a potential prognostic factor in GBM

Although the association of SIRT5 with clinicopathological parameters and prognoses have been demonstrated in several types of cancers, such as hepatocellular, lung, and colon cancers [20, 21, 22], there has been no exploration of these correlations with GBM. We downloaded datasets from UALCAN to investigate the relationship between the expression of SIRT5 and clinical characteristics. The level of SIRT5 expression was significantly increased in male patients, whereas SIRT5 expression did not show significant associations with age and race (Fig. 2A–C). The correlations of SIRT5 expression with prognosis was studied using an available dataset from the GlioVis platform. The survival curves suggested that high expression of SIRT5 mRNA was significantly favorable for overall survival time (OS) (HR $=$ 1.36, 95% CI $=$ 1.09–1.69, $p=$ 0.0053) (Fig. 2D). All of these findings suggested that SIRT5 could be a prognostic biomarker for GBM.

3.3 Analysis of the SIRT5 methylated status in GBM

DNA methylation has been demonstrated to alter the differentiation properties of glioblastoma and regulate its occurrence and progression [23, 24]. To investigate whether DNA methylation is involved in the expression of SIRT5 in GBM, the first stage was to assess the global DNA methylation status of SIRT5 between GBM and normal tissues. Using the HM27K sequencing data from DiseaseMeth, we found that the global methylation level in GBM tissues was significantly higher than that in normal tissues ( $p<$ 0.001) (Fig. 3A). Meanwhile, the data from the Wanderer database also confirmed the higher methylation values of 12 CpG sites in GBM cases compared with normal brain tissues, though there are no statistic differences (Table 2). The survival analysis suggested that only the cases with lower methylation values of the CpG site (cg08619119) had a significantly longer OS (HR $=$ 1.692, 95% CI $=$ 1.121–2.556, $p=$ 0.012) (Fig. 3B, Table 3). Based on its obvious significance, the CpG site, cg08619119, has been chosen to examine the association between SIRT5 expression and its DNA methylation status. We used Wanderer to collect data, and carried out Pearson correlation analysis to identify that SIRT5 transcriptional level was negatively correlated with the DNA methylation values of the CpG site (cg08619119) (Fig. 3C). These outcomes indicate that the effects of DNA methylation may inhibit the expression of SIRT5 and induce tumorigenesis in GBM.

3.4 Analysis of genes coexpressed with SIRT5 in GBM

To further understand the potential role of SIRT5 in GBM carcinogenesis, datasets (TCGA-GBM) were extracted from TCGA using the GlioVis platform to screen differentially expressed genes that interact with SIRT5 (Table S1). The cutoff was set as $|$ logFC $|⩾$ 1 and $P⩽$ 0.05 to filter significantly differentially coexpressed genes. As shown in Fig. 4A and B, we found a total of 70 differently expressed genes, of which 4 and 66 were identified as upregulated and downregulated in GBM, respectively. This finding indicated that most coexpressed genes were downregulated. We then selected these coexpressed genes to construct a protein-protein interaction (PPI) network using the STRING database and Cytoscape software (Fig. 4C). To further investigate the biological role of the coexpressed genes, we performed functional enrichment analysis using FunRich. Our results suggest that the top enriched biological processes were cell communication (29%), cell growth and/or maintenance (17.4%) and transport (15.9%) (Fig. 4D). We also found that the expressed protein was mostly distributed in the plasma membrane (46.4%), followed by extracellular (36.2%) and exosomes (27.5%). Transporter activity (8.7%), cell adhesion molecule activity (7.2%) and the extracellular matrix structural constituent (7.2%) were the main functions among these coexpressed genes. KEGG analysis suggested that the genes that coexpressed with SIRT5 were functionally enriched in several pathways, mostly those involving epithelial-to-mesenchymal transition (EMT, 23.2%) (Fig. 4E). Collectively, these findings preliminarily demonstrated the network of genes coexpressed with SIRT5 and the potential biological pathways in GBM.

Table S1

Significant differentially coexpressed genes that interacted with SIRT5 in glioblastoma

	logFC		AveExpr		$t$		$P$ .value	adj. $P$ .val	B
EPCAM	1.	313	4.	564	4.	958	1.281E-06	0.00000851	4.	96
KRT18	1.	19	5.	333	4.	418	0.00001462	0.00006673	2.	66
KRT19	1.	109	4.	238	4.	733	3.632E-06	0.0000205	3.	97
ANXA3	1.	051	4.	657	5.	907	1.081E-08	1.723E-07	9.	51
DDR2	$-$ 1		6.	17	$-$ 8		4.168E-14	1.557E-11	21.	5
SOX9	$-$ 1.	01	10.	02	$-$ 6.	45	5.54E-10	1.733E-08	12.	4
C3	$-$ 1.	01	11.	01	$-$ 5.	97	7.858E-09	1.324E-07	9.	82
IGFBP5	$-$ 1.	01	6.	692	$-$ 6.	79	7.373E-11	3.658E-09	14.	3
FAM107A	$-$ 1.	01	9.	036	$-$ 4.	51	9.992E-06	0.00004831	3.	02
VSIG4	$-$ 1.	01	9.	226	$-$ 4.	83	0.00000236	0.00001427	4.	38
C21orf62	$-$ 1.	02	6.	425	$-$ 4.	81	0.00000253	0.00001512	4.	31
NACC2	$-$ 1.	02	9.	039	$-$ 7.	04	1.724E-11	1.248E-09	15.	7
MAP1B	$-$ 1.	03	9.	983	$-$ 6.	29	1.359E-09	3.551E-08	11.	5
AGT	$-$ 1.	03	10.	38	$-$ 4.	16	0.00004272	0.0001675	1.	65
DPYSL3	$-$ 1.	04	8.	631	$-$ 7.	07	1.441E-11	1.095E-09	15.	9
CX3CR1	$-$ 1.	04	9.	178	$-$ 4.	6	6.466E-06	0.00003309	3.	43
DPYD	$-$ 1.	05	8.	374	$-$ 6		6.623E-09	1.156E-07	9.	98
BAALC	$-$ 1.	05	10.	77	$-$ 4.	46	0.000012	0.00005663	2.	84
PTN	$-$ 1.	05	10.	88	$-$ 4.	83	2.356E-06	0.00001426	4.	38
SCG2	$-$ 1.	05	8.	712	$-$ 4.	59	6.912E-06	0.00003505	3.	36
NID1	$-$ 1.	06	6.	892	$-$ 7.	8	1.534E-13	3.977E-11	20.	3
AKAP12	$-$ 1.	06	9.	167	$-$ 6.	28	1.434E-09	3.672E-08	11.	5
EDNRB	$-$ 1.	06	8.	906	$-$ 5.	31	2.349E-07	2.046E-06	6.	57
FAM70A	$-$ 1.	06	8.	842	$-$ 5.	38	1.665E-07	1.554E-06	6.	9
COL4A2	$-$ 1.	06	8.	667	$-$ 7.	17	7.86E-12	7.03E-10	16.	5
CORO2B	$-$ 1.	06	8.	234	$-$ 5.	89	1.191E-08	1.852E-07	9.	42
ANGPT2	$-$ 1.	07	6.	545	$-$ 6.	87	4.574E-11	2.63E-09	14.	8
C1R	$-$ 1.	07	8.	936	$-$ 6.	49	4.214E-10	1.42E-08	12.	6
CRYAB	$-$ 1.	08	11.	32	$-$ 4.	85	2.149E-06	0.00001316	4.	47
CHL1	$-$ 1.	08	8.	35	$-$ 5.	01	9.922E-07	6.886E-06	5.	2
APLNR	$-$ 1.	08	8.	347	$-$ 5.	45	1.179E-07	1.181E-06	7.	23
LPL	$-$ 1.	09	9.	54	$-$ 5.	24	3.337E-07	2.763E-06	6.	24
MLC1	$-$ 1.	1	8.	179	$-$ 6.	02	5.857E-09	1.058E-07	10.	1
GPM6A	$-$ 1.	1	11.	05	$-$ 3.	99	0.00008446	0.0003015	1.	01
PLTP	$-$ 1.	1	8.	844	$-$ 7.	72	2.421E-13	5.302E-11	19.	8
ITPKB	$-$ 1.	11	8.	49	$-$ 7.	71	2.69E-13	5.695E-11	19.	7
SERPINE1	$-$ 1.	11	7.	303	$-$ 5.	91	1.048E-08	1.683E-07	9.	54
SLC1A3	$-$ 1.	11	11.	25	$-$ 5.	45	1.191E-07	1.191E-06	7.	22
AQP4	$-$ 1.	11	8.	623	$-$ 5.	07	7.483E-07	5.456E-06	5.	47
PDPN	$-$ 1.	12	7.	827	$-$ 5.	34	1.987E-07	1.784E-06	6.	73
TUBB2B	$-$ 1.	12	10.	93	$-$ 4.	8	2.634E-06	0.00001562	4.	28
NTRK2	$-$ 1.	13	8.	433	$-$ 5.	64	4.416E-08	5.327E-07	8.	17
ZIC1	$-$ 1.	13	9.	242	$-$ 5.	35	1.913E-07	1.735E-06	6.	77
COL4A1	$-$ 1.	13	9.	303	$-$ 6.	48	4.607E-10	1.504E-08	12.	5
DIRAS3	$-$ 1.	13	6.	666	$-$ 6.	2	2.198E-09	5.086E-08	11
ATP1B2	$-$ 1.	14	9.	421	$-$ 5.	24	3.278E-07	2.721E-06	6.	26
CD163	$-$ 1.	19	8.	849	$-$ 5.	09	6.954E-07	5.111E-06	5.	54
ID4	$-$ 1.	19	9.	384	$-$ 6.	05	4.991E-09	9.24E-08	10.	3
VCAM1	$-$ 1.	2	8.	35	$-$ 5.	73	2.753E-08	3.616E-07	8.	62
TNC	$-$ 1.	2	7.	711	$-$ 7.	57	6.436E-13	1.12E-10	18.	9
TRIM9	$-$ 1.	21	8.	588	$-$ 5.	6	5.534E-08	6.412E-07	7.	95
MAN1C1	$-$ 1.	25	7.	424	$-$ 7.	38	2.059E-12	2.54E-10	17.	8
ABCA1	$-$ 1.	28	8.	15	$-$ 8.	78	2.197E-16	5.582E-13	26.	6
GPM6B	$-$ 1.	35	11.	48	$-$ 5.	76	2.35E-08	3.181E-07	8.	77
SRPX	$-$ 1.	36	9.	701	$-$ 6.	36	9.09E-10	2.583E-08	11.	9
RCAN1	$-$ 1.	36	7.	217	$-$ 7.	34	2.794E-12	3.223E-10	17.	5
CHI3L2	$-$ 1.	4	7.	915	$-$ 5.	23	3.424E-07	2.824E-06	6.	21
AQP1	$-$ 1.	41	9.	146	$-$ 5.	21	3.87E-07	3.125E-06	6.	1
GAP43	$-$ 1.	45	9.	207	$-$ 6.	07	4.59E-09	8.712E-08	10.	3
EFEMP1	$-$ 1.	48	9.	545	$-$ 7.	6	5.345E-13	9.561E-11	19.	1

Table S1, continued
	logFC		AveExpr		$t$		$P$ .value	adj. $P$ .val	B
PTPRZ1	$-$ 1.	49	11.	64	$-$ 5.	5	8.865E-08	9.406E-07	7.	5
PTX3	$-$ 1.	49	7.	324	$-$ 6.	01	6.152E-09	1.093E-07	10.	1
CHI3L1	$-$ 1.	51	10.	73	$-$ 4.	97	1.205E-06	8.098E-06	5.	02
SERPINA3	$-$ 1.	51	11.	28	$-$ 6.	18	2.394E-09	5.427E-08	11
LTF	$-$ 1.	58	8.	108	$-$ 4.	29	0.0000257	0.0001077	2.	13
NES	$-$ 1.	65	9.	937	$-$ 7.	57	6.587E-13	1.12E-10	18.	9
FABP7	$-$ 1.	67	10.	32	$-$ 5.	77	2.195E-08	3.011E-07	8.	84
PMP2	$-$ 1.	72	10.	78	$-$ 5.	6	5.53E-08	6.412E-07	7.	95
FCGBP	$-$ 1.	9	9.	474	$-$ 7.	77	1.781E-13	4.35E-11	20.	1

4. Discussion

In this study, we aimed to investigate important and novel biomarkers among 7 sirtuins and understand their biological functions. Intriguingly, we found that only the expression levels of SIRT5 decreased in GBM, with remarkable consistency among three independent datasets from these open bioinformatics platforms. We also revealed that the expression levels SIRT5 was positively associated with the length of survival.

SIRT5, a member of the sirtuin family, is an enigmatic protein with weak deacetylase activity [25] and major localization in the mitochondrial matrix [26]. Currently, mRNA expression of SIRT5 is detected across a wide variety of cancers, and its expression tends to be distinct in various cancers. SIRT5 mRNA is elevated in human lung cancers [22] and invasive breast tumors [27] and serves as a candidate biomarker for poor survival. By contrast, SIRT5 mRNA expression is significantly decreased in head and neck squamous cell carcinoma as a cancer suppressor [28]. To date, there are no reports on the function of SIRT5 in association with GBM. In this study, we first demonstrate that SIRT5 is significantly downregulated in GBM. To further confirm the prognostic value of SIRT5, we used UALCAN and GlioVis datasets to confirm that highly expressed SIRT5 was significantly associated with a better prognosis.

Abnormal DNA methylation is observed in a variety of pathologies, such as aging and tumorigenesis [29]. Several studies have confirmed that the functions of sirtuins could be influenced by DNA methylation or that they may participate in the regulation of DNA methylation. Aberrant CpG DNA methylation could induce aberrant expression of SIRT1-targeting miRNA, which regulates malignant SIRT1 elevation and the deregulation of p53-dependent events in tumorigenesis [30]. Through deacetylation of chromatin regulation marker H4K16, SIRT2 can modulate the mitotic deposition of H4K20 methylation as tumor repressor, which regulates cell cycle progression and genome stability [31]. However, the relationship between SIRT5 and DNA methylation is still unclear. Our study revealed that the expression of SIRT5 is negatively correlated with DNA methylation status of cg12078157, and meanwhile, its global DNA methylation values are obviously increased in GBM tissues. Also future work is needed to identify the detailed mechanism.

Construction of the SIRT5-related genes expression network and the results of functional enrichment analysis facilitated the identification of possible and important biological processes. From KEGG pathway analysis, EMT signaling pathway was found to be the most likely pathway associated with SIRT5 in GBM. EMT is a biological process in which epithelial cells undergo the loss of cell polarity and epithelial organization, culminating in a mesenchymal phenotype [32]. Previous studies have confirmed that EMT is one of the mechanisms that confer the invasive and migration properties of GBM cells [33]. Although to date there have been no studies investigating the role of SIRT5 or other sirtuins in EMT in GBM development, some research has shown that sirtuins can regulate EMT in other tumors. Overexpression of SIRT1 promotes EMT and metastasis in colorectal cancer [34] and breast cancer [35]. Similarly, upregulation of SIRT2 can mediate EMT via protein kinase B/glycogen synthase kinase-3 $\beta$ / $\beta$ -catenin signaling in hepatocellular carcinoma [36]. On the contrary, SIRT4 acted as tumor suppressor and repressed cell migration and invasion in gastric cancer via suppression of EMT through promoting E-cadherin expression [37]. Our study indicates that the suppression role of SIRT5 via EMT is warranted for further exploration regarding GBM.

In conclusion, this is the first report that SIRT5 is downregulated in GBM, which is correlated with poor outcome. Moreover, we demonstrated the negative relationship between SIRT5 expression and its DNA methylation values, indicating that the transcription level of SIRT5 could be modulated by the DNA methylation. Our findings suggest that SIRT5 serves as a promising prognostic factor for GBM and represses the tumorigenesis.