SOX12 was upregulated in SKCM
Based on pan-cancer RNA-seq data from TCGA, we evaluated SOX12 expression in 33 types of human cancer. Figure 1A shows the differential expression of SOX12 in pan-carcinoma between the tumor and adjacent normal tissues. SOX12 was significantly overexpressed in the vast majority of tumor types in TCGA database compared with normal tissues (25/33, 75.76%), including adrenocortical carcinoma (ACC), breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), acute myeloid leukemia (LAML), brain lower grade glioma (LGG), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), prostate adenocarcinoma (PRAD), skin cutaneous melanoma and stomach adenocarcinoma (SKCM), stomach adenocarcinoma (STAD), thymoma (THYM), uterine corpus endometrial carcinoma (UCEC) and uterine carcinosarcoma (UCS) (P < 0.05). In addition, SOX12 expression was lower in testicular germ cell tumors (TGCT) and thyroid carcinoma (THCA) tissues than that expression in normal tissues (P < 0.05).
Further analysis showed that SOX12 was upregulated in SKCM tumor tissues compared with that in normal tissues (p < 0.001, Fig. 1B). The diagnostic value of SOX12 in patients was identified using ROC curve analysis. As shown in Fig. 1C, SOX12 expression had high sensitivity and specificity for SKCM diagnosis, with an area under the curve (AUC) of 0.918 (95% confidence interval [CI], 0.902–0.933). These results suggest that SOX12 plays a significant regulatory role in SKCM progression.
Association between SOX12 expression and clinicopathologic features
The association between SOX12 expression and the clinical information of 471 patients is summarized in Table 1. High SOX12 expression significant correlated with melanoma ulceration (p = 0.015), OS (p = 0.026), and DSS (p = 0.023). However, we did not observe any correlations between SOX12 expression and other clinicopathological characteristics.
High Expression Of Sox12 Infers A Poor Prognosis For Skcm
Kaplan-Meier analysis was used to assess the relationship between SOX12 expression and the prognosis of patients with SKCM. As shown in Fig. 2A&B, high expression of SOX12 demonstrated a significantly poor prognosis for OS (hazard ratio [HR]: 1.51, p = 0.003) and DSS (HR: 1.58, p = 0.002) in patients with SKCM.
Univariate and multivariate Cox regression analyses were used to assess the predictive value of SOX12 expression for clinical outcomes. Clinical data, including sex, age, T stage, N stage, M stage, pathologic stage, SOX12, melanoma ulceration, melanoma clark level, and breslow depth, for SKCM patients was obtained from TCGA database. As shown in Table 2, high SOX12 expression was associated with significantly poor OS in the univariate (HR: 1.507, 95% CI: 1.150–1.975, p = 0.003) and multivariate (HR: 2.588, 95% CI: 1.695–3.952, p < 0.001) regression analyses. Moreover, the N stage (HR: 12.289, 95% CI: 3.323–45.453, p < 0.001) and melanoma ulceration (HR: 1.539, 95% CI: 1.002–2.366, p = 0.049) were independent risk factors for OS. Univariate Cox analysis demonstrated that age, T stage, M stage, pathologic stage, Clark level, and Breslow depth also had prognostic advantages for clinical outcomes.
Table 2
Univariate and multivariate Cox proportional hazards analysis of SOX12 expression and OS (overall survival) for patients with SKCM in the validation cohort
Characteristics
|
Total(N)
|
Univariate analysis
|
Multivariate analysis
|
Hazard ratio (95% CI)
|
P value
|
Hazard ratio (95% CI)
|
P value
|
Sex
|
456
|
|
|
|
|
Male
|
284
|
Reference
|
|
|
|
Female
|
172
|
0.853 (0.640–1.138)
|
0.281
|
|
|
Age
|
456
|
|
|
|
|
<=60
|
246
|
Reference
|
|
|
|
> 60
|
210
|
1.656 (1.251–2.192)
|
< 0.001**
|
1.022 (0.685–1.524)
|
0.916
|
T stage
|
361
|
|
|
|
|
T1
|
41
|
Reference
|
|
|
|
T2
|
77
|
1.495 (0.811–2.756)
|
0.197
|
1.363 (0.544–3.416)
|
0.509
|
T3
|
90
|
2.097 (1.158–3.798)
|
0.015*
|
1.347 (0.528–3.438)
|
0.533
|
T4
|
153
|
3.711 (2.070–6.653)
|
< 0.001**
|
2.448 (0.777–7.716)
|
0.126
|
N stage
|
402
|
|
|
|
|
N0
|
224
|
Reference
|
|
|
|
N1
|
73
|
1.497 (1.014–2.210)
|
0.043*
|
2.947 (0.828–10.491)
|
0.095
|
N2
|
49
|
1.534 (0.972–2.419)
|
0.066
|
5.052 (1.379–18.515)
|
0.014*
|
N3
|
56
|
2.731 (1.769–4.215)
|
< 0.001**
|
12.289 (3.323–45.453)
|
< 0.001**
|
M stage
|
430
|
|
|
|
|
M0
|
406
|
Reference
|
|
|
|
M1
|
24
|
1.897 (1.029–3.496)
|
0.040*
|
1.200 (0.445–3.234)
|
0.718
|
Pathologic stage
|
410
|
|
|
|
|
I& II
|
217
|
Reference
|
|
|
|
III& IV
|
193
|
1.617 (1.207–2.165)
|
0.001**
|
0.535 (0.159-1.800)
|
0.312
|
SOX12
|
456
|
|
|
|
|
Low
|
228
|
Reference
|
|
|
|
High
|
228
|
1.507 (1.150–1.975)
|
0.003**
|
2.588 (1.695–3.952)
|
< 0.001**
|
Melanoma ulceration
|
313
|
|
|
|
|
No
|
146
|
Reference
|
|
|
|
Yes
|
167
|
2.085 (1.495–2.907)
|
< 0.001**
|
1.539 (1.002–2.366)
|
0.049*
|
Melanoma Clark level
|
315
|
|
|
|
|
I&II
|
19
|
Reference
|
|
|
|
III&IV&V
|
296
|
2.689 (1.188–6.088)
|
0.018*
|
2.295 (0.594–8.871)
|
0.228
|
Breslow depth
|
355
|
|
|
|
|
<=3
|
180
|
Reference
|
|
|
|
> 3
|
175
|
2.651 (1.938–3.627)
|
< 0.001**
|
1.201 (0.590–2.445)
|
0.614
|
Abbreviations: CI, confidence interval; HR, hazard ratio. *P < 0.05, **P < 0.01. |
Similar results were observed for DSS. As shown in Table 3, high SOX12 levels were significant correlated with poor DSS in both univariate (HR: 1.575, 95% CI: 1.180–2.104, p = 0.002) and multivariate (HR: 3.046, 95% CI: 1.934–4.798, p < 0.001) regression analyses. In addition, age, T stage, N stage, M stage, pathologic stage, melanoma ulceration, melanoma Clark level, and Breslow depth also had prognostic advantages for clinical outcomes.
Table 3
Univariate and multivariate Cox proportional hazards analysis of SOX12 expression and DSS (Disease Specific Survival) for patients with SKCM in the validation cohort
Characteristics
|
Total(N)
|
Univariate analysis
|
Multivariate analysis
|
Hazard ratio (95% CI)
|
P value
|
Hazard ratio (95% CI)
|
P value
|
Sex
|
450
|
|
|
|
|
Male
|
278
|
Reference
|
|
|
|
Female
|
172
|
0.862 (0.635–1.170)
|
0.340
|
|
|
Age
|
450
|
|
|
|
|
<=60
|
244
|
Reference
|
|
|
|
> 60
|
206
|
1.699 (1.258–2.294)
|
< 0.001**
|
1.018 (0.666–1.556)
|
0.935
|
T stage
|
356
|
|
|
|
|
T1
|
41
|
Reference
|
|
|
|
T2
|
76
|
1.435 (0.775–2.657)
|
0.250
|
1.226 (0.483–3.113)
|
0.668
|
T3
|
88
|
1.934 (1.060–3.529)
|
0.032*
|
1.310 (0.508–3.378)
|
0.576
|
T4
|
151
|
3.110 (1.715–5.637)
|
< 0.001**
|
2.378 (0.720–7.855)
|
0.155
|
N stage
|
396
|
|
|
|
|
N0
|
221
|
Reference
|
|
|
|
N1
|
71
|
1.279 (0.827–1.976)
|
0.269
|
3.884 (0.851–17.729)
|
0.080
|
N2
|
49
|
1.473 (0.905–2.398)
|
0.120
|
6.820 (1.465–31.751)
|
0.014*
|
N3
|
55
|
2.920 (1.867–4.565)
|
< 0.001**
|
18.399 (3.925–86.241)
|
< 0.001**
|
M stage
|
424
|
|
|
|
|
M0
|
400
|
Reference
|
|
|
|
M1
|
24
|
2.200 (1.190–4.069)
|
0.012*
|
1.265 (0.465–3.437)
|
0.645
|
Pathologic stage
|
405
|
|
|
|
|
I& II
|
215
|
Reference
|
|
|
|
III& IV
|
190
|
1.536 (1.125–2.096)
|
0.007**
|
0.412 (0.096–1.778)
|
0.235
|
SOX12
|
450
|
|
|
|
|
Low
|
225
|
Reference
|
|
|
|
High
|
225
|
1.575 (1.180–2.104)
|
0.002**
|
3.046 (1.934–4.798)
|
< 0.001**
|
Melanoma ulceration
|
309
|
|
|
|
|
No
|
145
|
Reference
|
|
|
|
Yes
|
164
|
1.948 (1.372–2.767)
|
< 0.001**
|
1.498 (0.952–2.356)
|
0.081
|
Melanoma Clark level
|
310
|
|
|
|
|
I&II
|
19
|
Reference
|
|
|
|
III&IV&V
|
291
|
2.949 (1.207–7.209)
|
0.018*
|
2.337 (0.598–9.143)
|
0.222
|
Breslow depth
|
350
|
|
|
|
|
<=3
|
178
|
Reference
|
|
|
|
> 3
|
172
|
2.274 (1.628–3.177)
|
< 0.001**
|
1.039 (0.492–2.194)
|
0.921
|
Abbreviations: CI, confidence interval; HR, hazard ratio. *P < 0.05, **P < 0.01. |
Furthermore, we constructed a nomogram for predicting the OS and DSS of patients with SKCM at 1-, 3-, and 5-years based on SOX12 expression and clinical data (Fig. 2C &D).
We performed a probabilistic calibration analysis and plotted a calibration curve using the R package. As shown in Fig. 2E&F, the calibration curve of the Cox model for OS (Fig. 2E) and DSS (Fig. 2F) showed that the predicted values at 1-, 3-, and 5-years deviated little from the actual values, indicating satisfactory prediction. In summary, these nomograms predicted the survival of patients better than the individual prognostic factors.
Functional enrichment analysis of high and low- SOX12 expression
We screened the coexpressed genes of SOX12 with a Pearson correlation of coefficient |r| > 0.4 and p < 0.001. We analyzed a total of 564 DEGs in the high- and low-SOX12 expression samples, of which 562 and 2 genes positively and negatively correlated with SOX12 expression, respectively. The top 20 associated coexpressed genes of SOX12 are shown as a heat map (Fig. 3).
To further evaluate the potential function of SOX12 in SKCM, 564 DEGs were used for GO and KEGG analyses, of which were mainly involved in transcriptional and epigenetic regulation. The top GO enrichment items in BP, MF, and CC groups were RNA splicing, RNA splicing via transesterification reactions, ribosome biogenesis, rRNA metabolic process, nuclear chromatin, methyltransferase complex, histone methyltransferase complex, preribosome, catalytic activity acting on RNA, transcription coactivator activity, histone binding, and histone methyltransferase activity (Fig. 4A, Table 4).
Table 4
Ontology
|
ID
|
Description
|
GeneRatio
|
BgRatio
|
pvalue
|
p.adjust
|
qvalue
|
BP
|
GO:0042254
|
ribosome biogenesis
|
36/521
|
297/18670
|
1.23e-13
|
4.46e-10
|
4.08e-10
|
BP
|
GO:0016072
|
rRNA metabolic process
|
32/521
|
253/18670
|
9.16e-13
|
1.65e-09
|
1.51e-09
|
BP
|
GO:0008380
|
RNA splicing
|
44/521
|
469/18670
|
2.02e-12
|
2.43e-09
|
2.22e-09
|
BP
|
GO:0000375
|
RNA splicing, via transesterification reactions
|
38/521
|
382/18670
|
1.21e-11
|
1.09e-08
|
9.97e-09
|
CC
|
GO:0034708
|
methyltransferase complex
|
21/534
|
113/19717
|
2.87e-12
|
1.40e-09
|
1.14e-09
|
CC
|
GO:0000790
|
nuclear chromatin
|
37/534
|
377/19717
|
1.47e-11
|
2.50e-09
|
2.04e-09
|
CC
|
GO:0030684
|
preribosome
|
17/534
|
76/19717
|
1.54e-11
|
2.50e-09
|
2.04e-09
|
CC
|
GO:0035097
|
histone methyltransferase complex
|
17/534
|
86/19717
|
1.25e-10
|
1.52e-08
|
1.24e-08
|
MF
|
GO:0003713
|
transcription coactivator activity
|
29/526
|
319/17697
|
1.00e-07
|
5.05e-05
|
4.11e-05
|
MF
|
GO:0140098
|
catalytic activity, acting on RNA
|
32/526
|
386/17697
|
1.84e-07
|
5.05e-05
|
4.11e-05
|
MF
|
GO:0042054
|
histone methyltransferase activity
|
11/526
|
57/17697
|
7.71e-07
|
1.41e-04
|
1.14e-04
|
MF
|
GO:0042393
|
histone binding
|
20/526
|
197/17697
|
1.81e-06
|
2.47e-04
|
2.01e-04
|
In addition, KEGG analysis indicated that SOX12 was implicated in the spliceosome, RNA transport, ribosome biogenesis in eukaryotes, mRNA surveillance pathway, and base excision repair (Fig. 4B, Table 5). These results indicated that SOX12 worked by participating in RNA splicing, histone modification and ribosome biogenesis in SKCM.
Table 5
Ontology
|
ID
|
Description
|
GeneRatio
|
BgRatio
|
pvalue
|
p.adjust
|
qvalue
|
KEGG
|
hsa03040
|
Spliceosome
|
18/218
|
151/8076
|
1.09e-07
|
2.31e-05
|
2.27e-05
|
KEGG
|
hsa03015
|
mRNA surveillance pathway
|
10/218
|
97/8076
|
2.70e-04
|
0.029
|
0.028
|
KEGG
|
hsa03008
|
Ribosome biogenesis in eukaryotes
|
10/218
|
111/8076
|
7.97e-04
|
0.056
|
0.055
|
KEGG
|
hsa03013
|
RNA transport
|
13/218
|
186/8076
|
0.002
|
0.074
|
0.073
|
KEGG
|
hsa03410
|
Base excision repair
|
5/218
|
33/8076
|
0.002
|
0.074
|
0.073
|
GSEA enrichment analysis of high and low- SOX12 expression
To investigate the biological functions of SOX12 in SKCM, we performed GSEA enrichment analysis on the differentially expressed SOX12 genes. GSEA analysis revealed 222 datasets satisfying the criteria of FDR < 0.25 and adjusted p < 0.05 (false discovery rate, FDR) by the enrichment of the MSigDB Collection (c2.cp.v7.2.symbols.gmt). The most significant enrichment pathways were the sumoylation of DNA replication proteins, sumoylation of RNA binding proteins, polo-like kinase-mediated events, formation of the cornified envelope, keratinization, and antigen activates b cell receptor bcr leading to generation of second messengers (Fig. 5, Table 6). These results demonstrated the key pathways related to SOX12.
Table 6
ID
|
NES
|
p value
|
p.adjust
|
FDR
|
REACTOME_SUMOYLATION_OF_DNA_REPLICATION_PROTEINS
|
2.233
|
0.0046
|
0.049
|
0.036
|
REACTOME_SUMOYLATION_OF_RNA_BINDING_PROTEINS
|
2.153
|
0.0045
|
0.049
|
0.036
|
REACTOME_POLO_LIKE_KINASE_MEDIATED_EVENTS
|
2.152
|
0.0030
|
0.045
|
0.033
|
REACTOME_FORMATION_OF_THE_CORNIFIED_ENVELOPE
|
-2.812
|
0.0012
|
0.028
|
0.021
|
REACTOME_KERATINIZATION
|
-2.763
|
0.0011
|
0.028
|
0.021
|
REACTOME_ANTIGEN_ACTIVATES_B_CELL_RECEPTOR_
BCR_LEADING_TO_GENERATION_OF_SECOND_MESSENGERS
|
-2.647
|
0.0012
|
0.028
|
0.021
|
The correlation between SOX12 expression and the infiltration of immune cells
It has been reported that tumor-infiltrating immune cells served as an important indicator of survival and sentinel lymph node status in tumor patients [20, 21]. To evaluate whether SOX12 is involved in tumor immunity, we used the ssGSEA algorithm to evaluate the relationship between SOX12 expression and infiltration levels of 24 immune cells. The correlation between SOX12 expression and infiltrating levels of immune cells was evaluated using the Spearman method. We observed close relationship between SOX12 and dendritic Cells (DC), T cells, and NK cells (Fig. 6A). Furthermore, SOX12 expression was negatively correlated with the infiltration of activated dendritic cells (DCs) (R=-0.316, p < 0.001, Fig. 6B), T cells (R=-0.302, p < 0.001, Fig. 6C), T helper cells (R=-0.217, p < 0.001, Fig. 6D), and CD8 T cells (R=-0.166, p < 0.001, Fig. 6E), whereas SOX12 expression did not correlate with Th2 cells (R=-0.074, p = 0.109, Fig. 6F) and NK cells (R = 0.065, p = 0.161, Fig. 6G). We further quantified the relative abundance of tumor-infiltrating immune cells in the SKCM [22]. As shown in Fig. 6H, the proportion of tumor-infiltrating immune cells was significantly different in the high and low-SOX12 subgroups.
Tumor-infiltrating lymphocytes may serve as independent predictors of sentinel lymph node status and survival in patients with cutaneous melanomas [23]. We further analyzed the potential relationship between SOX12 expression and immune infiltration levels in SKCM using TIMER software. As shown in Fig. 7A, the “SCNA” module analysis suggested that several immune cell infiltrations are involved in the altered SOX12 gene copy numbers, including B cells, CD4 + T cells, macrophages, neutrophils and DCs, in SKCM. The “Gene” module analysis indicated that SOX12 expression was also closely associated with tumor purity. Moreover, SOX12 expression positively correlated with infiltrating levels of B cells and CD4 + T cells and negatively correlated with CD8 + T cells, macrophages, neutrophils and DCs infiltrating levels in SKCM (Fig. 7B). Finally, the effect of immune infiltration cell on the survival of patients with SKCM was evaluated. As shown in Fig. 7C, low levels of B cells, CD8 + T cells, neutrophils, and DCs were significantly associated with poor prognosis of patients with SKCM. Additionally, using TIMER software, the high expression of SOX12 was associated with a poor prognosis in patients with SKCM, and this result was consistent with the results of our previous KM analysis.
The correlation between SOX12 expression and methylation in SKCM
Because the GO and GSEA enrichment analyses confirmed the possible involvement of SOX12 in the regulation of epigenetic modifications, we further investigated the correlation between SOX12 expression and its methylation status. First, UALCAN analysis showed that SOX12 had a slight trend of high methylation levels in SKCM tissues (no statistical significance) (Fig. 8A). We obtained similar results using DiseaseMeth version 2.0, which showed higher methylation of SOX12 in SKCM tissues than in paracancerous normal tissues adjacent to cancer (P = 0.3399, Fig. 8B). Seven methylation sites (cg06006403, cg11871345, cg25575688, cg23922081, cg10625705, cg03342008 and cg03901958) in the DNA sequences of SOX12 were negatively associated with their expression levels, and one methylation site (cg23827867) was positively associated with SOX12 expression (Fig. 8C).
Validation The Sox12 Expression In Skcm Tissues And Cell Lines
To detect SOX12 expression, we randomly selected 10 pairs of SKCM tissues from the 21 pairs collected for qPCR analysis. Compared with the matched normal adjacent tissues, SOX12 expression was significantly increased in tumor tissues (80%, 8/10) (P < 0.001; Fig. 9A). Furthermore, at the cellular level, SOX12 mRNA expression was significantly higher in A375 and A875 cells than in HaCat cells, especially in A875 cells (Fig. 9B). To further confirm SOX12 expression in tissues, we performed HE staining and IHC analysis on the SKCM tissues. As shown in Fig. 9C, the nuclei of SKCM tissues were larger than those of normal tissues, and the former showed megakaryons or multinucleates, a typical pathological feature of cancer. Correspondingly, IHC was used to detect SOX12 protein expression in SKCM tissues. The details of clinicopathologic features are documented in Supplementary Table 1. Among the 21 SKCM cases, 5 cases were negative, and 16 were positive. The overall positivity rate was 76.19%, and five cases (23.81%) were SOX12-negative. Further analysis revealed that SOX12 was predominantly distributed in the nucleus and weakly stained in the cytoplasm (Fig. 9D&E). Based on SOX12 protein expression in tumor tissues, a division of patients with SKCM into low (negative) and high (positive) SOX12 expression groups was performed.