Abstract
Background: Pancreatic cancer (PC) is a malignant gastrointestinal
tumor with a terrible prognosis. Cuproptosis is a recently discovered
form of cell death. This study is intended to explore the relationship
between cuproptosis-related lncRNAs (CRLncs) signature with the
prognosis and the tumor microenvironment (TME) of PC.
Methods: Transcript sequencing data of PC samples with clinical
information were obtained from the Cancer Genome Atlas (TCGA).
Univariate Cox regression analysis and LASSO regression analysis were
employed to construct the prognostic signature based on CRLncs
associated with PC survival. A nomogram was created according to this
signature, and the signaling pathway enrichment was analyzed.
Subsequently, we explored the link between this prognostic signature
with the mutational landscape and TME. Eventually, drug sensitivity was
predicted based on this signature.
Results: Forty-six of 159 CRLncs were most significantly relevant to
the prognosis of PC, and a 6-lncRNA prognostic signature was
established. The expression level of signature lncRNAs were detected in
PC cell lines. The AUC value of the ROC curve for this risk score
predicting 5-year survival in PC was .944, which was an independent
prognostic factor for PC. The risk score was tightly related to the
mutational pattern of PC, especially the driver genes of PC.
Single-sample gene set enrichment analysis (ssGSEA) demonstrated a
significant correlation between signature with the TME of PC.
Ultimately, compounds were measured for therapy in high-risk and
low-risk PC patients, respectively.
Conclusion: A prognostic signature of CRLncs for PC was established in
the current study, which may serve as a promising marker for the
outcomes of PC patients and has important forecasting roles for gene
mutations, immune cell infiltration, and drug sensitivity in PC.
Keywords: pancreatic cancer, cuproptosis, lncRNA signature, mutation
landscape, tumor microenvironment
1 Introduction
The number of newly diagnosed pancreatic cancer (PC) in the United
States broke 60,000 in 2021, with the incidence increasing by over 1%
annually. After numerous years of efforts worldwide, the 5-year
survival rate for PC has only just surpassed 10% ([36]Rahib et al.,
2014; [37]Siegel et al., 2021; [38]Siegel et al., 2022). The risk
elements for PC are not well-defined and the main ones generally
accepted are family history, tobacco, alcohol abuse, obesity, and type
II diabetes ([39]Mizrahi et al., 2020; [40]Huang et al., 2021). Lacking
effective early screening measures, localized PC patients often are
asymptomatic or have minimal symptoms, and approximately half of newly
diagnosed PC cases are advanced diseases at diagnosis, with an average
survival time of less than 1 year ([41]Mizrahi et al., 2020). The high
heterogeneity of PC contributes to the challenge of predicting its
prognosis, and satisfactory results are not achievable by traditional
TNM staging alone. Given the current dilemma in the prognostic
prediction of PC, new strategies for screening high-risk patients to
detect PC at an early stage are urgently necessary to provide practical
clinical benefit ([42]Mizrahi et al., 2020). Hence, a prognostic model
based on molecular profiling to screen for high-risk patients is
appreciated.
Tsvetkov et al. found that copper triggered the aggregation of
mitochondrial lipid acylated proteins to induce regulated cell death
(RCD) and named this unique form of death “cuproptosis” ([43]Tang et
al., 2022; [44]Tsvetkov et al., 2022). Studies have been proposed
before this to suggest a strong relationship between copper with
cancer. Yu et al. suggested that blocking SLC31A1-dependent copper
uptake increased PC cell autophagy to resist cell death ([45]Yu et al.,
2019). Inhibition of copper transport induces apoptosis and inhibits
tumor angiogenesis in triple-negative breast cancer cells
([46]Karginova et al., 2019). Of course, the link between copper with
tumor is not only in the cuproptosis but also as a possible therapeutic
target for treatment ([47]Safi et al., 2014). Disulfiram (DSF) acts as
a copper carrier to induce copper-dependent oxidative stress and
mediates antitumor efficacy in inflammatory breast cancer
([48]Allensworth et al., 2015). Thus, this form of death brings promise
for the therapy of PC.
Long non-coding RNAs (LncRNAs) are described as RNAs consisting of over
200 nucleotides without the ability to encode proteins ([49]Gibb et
al., 2011; [50]Serghiou et al., 2016). LncRNAs exert a central role in
PC pathogenesis as regulators of cancer pathways, regulating cellular
processes including but not limited to cell cycle, apoptosis, and
epithelial-mesenchymal transition to influence biological behaviors
such as tumor growth, migration, and invasion ([51]Kornienko et al.,
2013; [52]Sharma et al., 2021). The critical role of LncRNAs in cancer
is what makes them proposed as important biomarkers of cancer outcome
and are emerging as promising candidates for biomarker development in
various cancers, including PC ([53]Esteller, 2011; [54]Sharma et al.,
2021; [55]Toden et al., 2021). The potential value of prognostic
signature regarding cuproptosis-associated lncRNAs as a prognostic
predictor for PC is currently uncertain.
The current study utilized CRLncs to create a signature for PC as well
as a nomogram to predict the prognosis of PC patients. The intimate
association of signature with mutational pattern in PC samples and
immune infiltration landscape in TME was revealed. In this study, the
high expression of signature lncRNAs in PC cell lines was investigated
by PCR experiment. This study identified a prognostic signature of
CRLncs as a potentially promising marker that could be applied for
forecasting the long-term survival time and guiding the therapeutic
regimen for PC patients.
2 Materials and methods
2.1 Data source
Transcript sequencing data and clinical details of PC patients were
available in the latest version of TCGA (July 2022,
[56]https://portal.gdc.cancer.gov/). The raw copy number variation
(CNV) data of PC was extracted from UCSC Xena
([57]http://xena.ucsc.edu/). Nineteen cuproptosis-related genes (CRGs)
([58]Supplementary Table S1) were searched from works of literature
([59]Yu et al., 2019; [60]Gao et al., 2021; [61]Tang et al., 2022).
2.2 To identify CRLncs
The “igraph” package was utilized to draw the correlations within CRGs.
The cuproptosis-related lncRNAs were identified via Pearson correlation
analysis with a correlation coefficient = .35 and p = .05 considered
relevant.
2.3 CRLncs prognostic signature
First, the prognostic value of CRLncs was assessed by univariate Cox
analysis, whereby a threshold of cox = .05 and p < .05 was accepted. In
the current study, The least absolute shrinkage and selection operator
(LASSO) Cox regression algorithm was employed to remove overfitting
between CRLncs and construct CRLncs signature, which was run through
the “glmnet” package ([62]Engebretsen and Bohlin, 2019). Risk score
formula:
[MATH:
Risk score=∑in(CoefCRLncs×ExpCRLncs) :MATH]
The TCGA samples were randomly divided into train cohort and test
cohort initially. The risk of PC samples was scored and patients were
classified as high-risk and low-risk groups with a cut-off of median
risk. Kaplan-Meier (K-M) survival curves were plotted to assess the
progression-free survival (PFS) and overall survival (OS) of PC
patients. Time-dependent receiver operating characteristic (ROC) curves
were drawn via the “timeROC” package to determine the performance of
this signature. Principal Components Analysis (PCA) was performed for
each risk subtype by the “scatterplot3d” package.
2.4 To construct a nomogram
Univariate combined with multivariate Cox regression analyses to
determine whether this risk score is a prognostic factor independently
of other clinical features. A nomogram was constructed with the “rms”
package to predict the survival rate of individual patients, and
calibration curves were employed to measure the performance of the
nomogram model. The concordance index (C-index) curve was plotted to
detect the stability of the risk score and nomogram prediction power.
The decision curve analysis (DCA) was carried out to determine the
contribution of risk score and nomogram in clinical decision-making.
2.5 Functional enrichment analysis
Initially, differences in active signaling pathways among different
risk PC samples were explored by Gene Set Enrichment Analysis (GSEA).
To understand the biological functions of risk-related differential
genes and the potential signaling enrichment pathways, the gene
ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and
Genomes (KEGG) pathway enrichment analysis were conducted with the
package of “clusterProfiler”, and presented as bar graphs, bubble
plots, and functional clustering circles, respectively.
2.6 Mutation landscape
Firstly, the “maftools” R package was employed to describe the
mutational patterns of CRGs. The “maftools” and “Rcircos” packages were
employed for CNV analysis. All mutation frequencies and oncoplot
waterfall maps of PC patients were drawn by the “maftools” package. The
association between risk score and the frequency of mutations in the
classical driver genes of PC (KRAS, TP53, SMAD4, and CDKN2A) was
further explored ([63]Kanda et al., 2012). The association between TMB
and patient OS was analyzed in a tumor mutation load analysis, and the
“maftools” package was finally utilized to compare the mutation
frequency between different risk groups.
2.7 Immune infiltration analysis
First, the CIBERSORT algorithm was used to reveal the immune
infiltration pattern of PC samples in TCGA, and the “corrplot” package
was utilized to explore the correlation among various immune cells.
Alternatively, the ssGSEA was employed to compare differences in the
level of immune cell infiltration and immune function in different risk
PC samples. Eventually, the expression levels of HLA family genes and
immune checkpoint genes were compared between patients with different
risks.
2.8 Compound sensitivity analysis
To investigate the potential of this risk score for PC therapy, we
explored the drug therapy data in the Cancer Therapeutics Response
Portal (CTRP) database with the “oncoPredict” package ([64]Maeser et
al., 2021). This study focused on the differences in sensitivity of PC
patients with different risk profiles to compounds currently in
clinical use. Through literature review, we collected 24 compounds that
are at least in the clinical trial stage for PC clinical treatment
([65]Sally et al., 2022).
2.9 Cell culture
The pancreatic cancer cell lines of AsPC-1, Capan-2, CFPAC-1, MIA
PaCa-2, and SW1990 were used to detect the expression levels of
lncRNAs, and HPDE cell lines were employed as normal controls. AsPC-1,
Capan-2, and CFPAC-1 were cultured with 1,640 complete medium, while
MIA PaCa-2 and SW1990 were cultured with DMEM complete medium.
2.10 Quantitative real-time PCR
Details of the full procedure of the PCR operation were carried out in
accordance with previous study (30333874). PCR primers for FAM83A-AS1,
PAN3-AS1, and SENCR were purchased from SangonBiotech (Sangon,
Shanghai, China). The primer sequences were as follows: FAM83A-AS1: 5’
-AAG AGC ATG AAA GAC TGA GGA AGC G-3’ (forward), 5’ -TCC AGG AGG TCG
GTG CCA TTG-3’ (reverse); PAN3-AS1: 5’ -CTT CCT CTC CCC GTT TCC TTT CTT
C-3’ (forward), 5’ -CAA GAG GTT AGC GTA ATC GGT CCA G-3’ (reverse);
SENCR: 5’ -GCT TTC AGG AGA ATG CGG AGA GAC-3’ (forward), 5’ -TTC TGG
CTG AAT GAG GAG CAA TGT G-3’ (reverse); β-actin: 5’ -CCT GGC ACC CAG
CAC AAT-3’ (forward), 5’ -GGG CCG GAC TCG TCA TAC-3’ (reverse). The
expression of FAM83A-AS1, PAN3-AS1, and SENCR was standardized by the
internal control β-actin. In addition, fold changes in FAM83A-AS1,
PAN3-AS1, and SENCR were calculated by 2^−ΔΔCT.
3 Statistical analysis
All analysis processes were carried out with the R software
([66]https://www.r-project.org/, version 4.2.1) plus its application
packages. The Student’s t-test was employed to compare the statistical
difference between two groups. The log-rank test was performed for the
Kaplan-Meier survival analysis. Drug sensitivity differences were
compared with the Wilcoxon test. p-value <.05 is considered
statistically significant if not otherwise stated. *p < .05, **p < .01;
***p < .001; ns, not significant.
4 Results
4.1 Genetic variation of CRGs in PC
This study progressed according to the flow chart ([67]Figure 1). The
TCGA dataset of 179 PC was downloaded, accompanied by four normal
pancreatic tissue samples. The mutation waterfall plot of the nineteen
CRGs in PC samples indicated that the frequency of CDKN2A mutation was
the highest (17%), NLRP3, ATP7A, NFE2L2, ATP7B, FDX1, LIAS, DLAT,
PDHA1, GLS, and DBT were mutated at 1%, and the remaining were
unmutated ([68]Figure 2A). The genetic instability of CRGs in PC
samples is commonly visible ([69]Figures 2B, C). The interconnections
among these CRGs are very obvious ([70]Figure 2D).
FIGURE 1.
[71]FIGURE 1
[72]Open in a new tab
The flowchart for analyzing cuproptosis-related lncRNAs prognostic
signature.
FIGURE 2.
[73]FIGURE 2
[74]Open in a new tab
Genetic variations of cuproptosis-related genes (CRGs). (A) The
mutation waterfall plot of the CRGs in PC. (B) Copy number variation
(CNV) frequencies of CRGs. (C) The sites of CRGs with CNV on
chromosomes. (D) The interconnections among CRGs. (E) Identification of
cuproptosis-related lncRNAs (CRLncs).
4.2 Establishing and validating a CRLncs signature
In the TCGA dataset, altogether 159 CRLncs were identified ([75]Figure
2E). Initially, forty-six of the above CRLncs were identified as
prognostically associated with PC using univariate Cox regression
analysis, of which four were risk factors and the rest were protective
factors ([76]Figure 3A). The heatmap showed that all 46 CRLncs
mentioned above were differentially expressed in PC tissues ([77]Figure
3B). A strong positive regulatory relationship was observed between
these CRLncs and CRGs, with most CRLncs being closely related to GLS,
LIAS, and NLRP3 ([78]Figure 3C). Subsequently, LASSO Cox regression
analysis was performed to obtain a CRLncs signature with an optimal λ
value ([79]Figures 3D, E). The signature is the sum of the product of
the coefficient of the six CRLncs with their expression, as follows:
[MATH: Risk score=0.2738*expression of FAM83A−AS1+−0.2123* expression of AL137186.1+−0.0024*expression of SENCR+−0.0193*expression of PAN3−AS1+−0.5518*expression of Z97832.2+
−0.2967*expression of AL139243.1 :MATH]
FIGURE 3.
[80]FIGURE 3
[81]Open in a new tab
Establishment of the CRLncs signature for PC. (A) Recognition of the
prognostic CRLncs in PC via univariate Cox regression analysis. (B)
Heatmap of the expression of prognostic CRLncs. (C) The ggalluvial
diagram of relationship between CRGs and prognostic CRLncs. (D,E) The
LASSO Cox regression analysis was performed depending on the optimal λ
value. (F) The relationship between CRGs with signature CRLncs. (G–K)
The PCA of PC patients in the TCGA-train, test, and entire cohort. *p <
.05, **p < .01; ***p < .001; ns, not significant.
According to this model, the risk score was calculated for individual
patients in the training cohort, the validation cohort, and the entire
TCGA cohort. PC patients were divided into high- and low-risk groups
based on the median risk score. The correlation between the six
signature CRLncs with CRGs was next assessed, with distinct positive
relationships between [82]AL137186.1, [83]AL139243.1, and SENCR with
NLRP3. There was a significant positive connection between FAM83A-AS1
with GCSH and PAN3-AS1 with [84]Z97832.2 ([85]Figure 3F). PCA analysis
demonstrated that the high-risk samples in the three cohorts were
independent of the low-risk samples ([86]Figures 3G–K). The risk score
and survival status map were plotted ([87]Figures 4A, B). As the K-M
curves suggested, low-risk patients shared significantly superior OS
and PFS to high-risk patients ([88]Figures 4C, D). The area under the
ROC curve (AUC) values for 1/3/5 years were 0.794/0.796/0.944 in the
training cohort, 0.573/0.653/0.615 in the validation cohort, and
0.703/0.721/0.757 in the whole cohort, respectively ([89]Figure 4E).
The expression of six signature CRLncs was substantially different
between the high- and low-risk groups, with FAM83A-AS1 being more
highly expressed in the high-risk group and the remaining being
expressed at higher levels in the opposite risk group ([90]Figure 4F).
High expression of FAM83A-AS1, PAN3-AS1, and SENCR in PC cell lines was
demonstrated by PCR ([91]Figure 4G).
FIGURE 4.
[92]FIGURE 4
[93]Open in a new tab
Construction of the CRLncs prognostic signature. (A–F) The risk score,
survival status, Kaplan-Meier plot survival cure of OS and PFS, ROC
curve predicted the 1/3/5-year OS, and heatmap of 6 signature CRLncs in
different risk PC patients in the TCGA-train, test, and entire cohort.
(G) The PCR of signature lncRNAs expression levels. *p < .05, **p <
.01; ***p < .001; ns, not significant.
4.3 Building a nomogram
The results of univariate and multivariate Cox regression analyses
suggested that this risk score was an independent prognostic factor for
PC ([94]Figures 5A, B). Alternatively, a nomogram was generated to
score patients and predict their OS in 1, 3, and 5 years ([95]Figure
5C), and the calibration curve showed that the nomogram performed well
([96]Figure 5D). The concordance index curve demonstrated that risk
score and nomogram outperformed age, gender, and stage ([97]Figure 5E).
The DCA curve demonstrated the contribution of risk and nomogram in
clinical decision-making ([98]Figure 5F). Consistently, the AUC values
for risk score predicting 0.5/1/2/3/4/5-year overall survival were
.713, .707, .698, .740, .801, and .781, respectively, and the AUC
values for nomogram predicting .5/1/2/3/4/5-year OS were .637, .591,
.694, .721, .737, and .698, which were better than other clinical
characteristics ([99]Figures 5G–L). Further, the interrelationship
between the signature CRLncs was assessed and a remarkable positive
correlation was found between PAN3-AS1 with [100]Z97832.2 ([101]Figure
6A). The relationship between these signature CRLncs and risk score
with clinical features is intimate ([102]Table 1). [103]AL137186.1 was
highly expressed in patients with M0 or T3-4 ([104]Figures 6B, C).
PAN3-AS1 was more abundant in younger or N0 patients ([105]Figures 6D,
E). PAN3-AS1, [106]Z97832.2, and risk score were expressed at higher
levels in patients with high stage and T-staging ([107]Figures 6F–J).
The heat map illustrates the distribution of risk score with clinical
characteristics ([108]Figure 6K). The outcomes of PC patients in the
low-risk group were better than their counterparts, across all clinical
characteristic subgroups ([109]Figures 6L–W).
FIGURE 5.
[110]FIGURE 5
[111]Open in a new tab
A nomogram for PC patients in TCGA dataset. (A,B) Independent
prognostic value of risk score by univariate and multivariate analysis.
(C) A nomogram of the risk score for predicting the 1-, 3-, and 5-year
OS of PC patients. (D) Calibration curves of this nomogram for the
prediction of 1/3/5-year OS of PC patients. (E) The concordance index
curve of risk score and nomogram. (F) The DCA curve of risk score and
nomogram. (G–L) The AUC values for risk score and nomogram predict
0.5/1/2/3/4/5-year OS. *p < .05, **p < .01; ***p < .001; ns, not
significant.
FIGURE 6.
[112]FIGURE 6
[113]Open in a new tab
Relationship between risk score and clinical characteristics. (A) Chord
diagram of the relationship between the six signature CRLncs. (B–J) The
relationship between these signature CRLncs and risk score with
clinical features. (K) The clinical characteristics heatmap. (L–W) The
survival of PC patients in different risk group across clinical
characteristic subgroups.
TABLE 1.
The distribution of clinical features in three cohorts.
Covariates Type Train cohort Test cohort Entire cohort
Age ≤65 43 (50%) 46 (53.49%) 89 (51.74%)
Age >65 43 (50%) 40 (46.51%) 83 (48.26%)
Gender FEMALE 40 (46.51%) 38 (44.19%) 78 (45.35%)
Gender MALE 46 (53.49%) 48 (55.81%) 94 (54.65%)
Stage Stage I 7 (8.14%) 13 (15.12%) 20 (11.63%)
Stage Stage II 74 (86.05%) 68 (79.07%) 142 (82.56%)
Stage Stage III 1 (1.16%) 2 (2.33%) 3 (1.74%)
Stage Stage IV 2 (2.33%) 2 (2.33%) 4 (2.33%)
Stage unknow 2 (2.33%) 1 (1.16%) 3 (1.74%)
T T1 2 (2.33%) 4 (4.65%) 6 (3.49%)
T T2 8 (9.3%) 15 (17.44%) 23 (13.37%)
T T3 74 (86.05%) 64 (74.42%) 138 (80.23%)
T T4 1 (1.16%) 2 (2.33%) 3 (1.74%)
T unknow 1 (1.16%) 1 (1.16%) 2 (1.16%)
N N0 26 (30.23%) 21 (24.42%) 47 (27.33%)
N N1 57 (66.28%) 64 (74.42%) 121 (70.35%)
N unknow 3 (3.49%) 1 (1.16%) 4 (2.33%)
M M0 38 (44.19%) 38 (44.19%) 76 (44.19%)
M M1 2 (2.33%) 2 (2.33%) 4 (2.33%)
M unknow 46 (53.49%) 46 (53.49%) 92 (53.49%)
[114]Open in a new tab
4.4 Functional enrichment
Initially, GSEA shows that high-risk patients are predominantly
enriched in the signaling pathways of calcium signaling pathway,
primary immunodeficiency, and type II diabetes mellitus ([115]Figures
7A, B). Malignant activities such as cell proliferation are more active
in high-risk patients. To further explore the functional enrichment of
risk differentially expressed genes, a risk difference analysis was
performed with the threshold set at logFC = 1, p < .05. Among them, the
GO terms with the highest enrichment were, biological process:
regulation of cytosolic calcium ion concentration; cellular component:
T cell receptor complex; and molecular function: metal ion
transmembrane transporter activity ([116]Figures 7C, D). KEGG terms
were mainly enriched in the T cell receptor signaling pathway and the
cAMP signaling pathway ([117]Figures 7E, F). The GO and KEGG enrichment
circle plots show the number of differential genes enriched in the top
eighteen items ([118]Figures 7G, H).
FIGURE 7.
[119]FIGURE 7
[120]Open in a new tab
Enrichment Analysis of GSEA. (A,B) GSEA of the top 5 up-regulated and
down-regulated KEGG signaling pathways enriched in high and low-risk
groups. Enrichment Analysis of risk differentially expressed genes.
(C,D) The bar and bubble plots of GO terms. (E,F) The bar and bubble
charts of KEGG terms. (G,H) The GO and KEGG enrichment circle plots.
4.5 Tumor mutation load analysis
The most common variant classification among all mutations was missense
mutation, and the most frequent variant type was SNP ([121]Figure 8A).
Mutations occurred in 145 of 173 PC samples (83.82%), with 62% of the
samples harboring mutations in KRAS and 58% of the samples harboring
mutations in TP53 ([122]Figure 8B). PC patients carrying mutated KRAS,
TP53, and CDKN2A had higher risk score than wild-type patients, while
SMAD4 mutation was not associated with risk ([123]Figures 8C–F). The
TMB of the PC samples in the TCGA database was scored and they were
categorized into two mutation groups. Mutations have a significant
impact on survival time, with patients in the low-mutation group
enjoying a much longer survival time than their counterparts
([124]Figure 8G). Then, combining signature risk and TMB, we found that
patients with high TMB from the high-risk group suffered the worst
prognostic outcome ([125]Figure 8H). The mutation landscape maps were
dramatically diverse among patients with different risks, with a
substantially higher mutation frequency in high-risk patients than in
low-risk patients. The frequency of KRAS mutation was (80% vs. 47%) and
TP53 mutation was (73% vs. 44%) in patients with high and low risks
([126]Figures 8I, J).
FIGURE 8.
[127]FIGURE 8
[128]Open in a new tab
Mutation analysis of samples. (A,B) Waterfall plot for the PC patients
in TCGA database. (C–F) The relationship between CDKN2A, KRAS, TP53,
and SMAD4 gene mutation with risk score. (G,H) The relationship between
TMB and PC survival. (I,J) Waterfall plots in different risk groups.
4.6 Relationship between signature with immune infiltration
The proportion of immune cells varied significantly among PC samples in
the TCGA ([129]Figures 9A, B). Pearson’s correlation heatmap suggested
that the correlation differed among immune cells. There was a
significantly positive correlation between B cells memory and CD4^+ T
cells naive, while CD8^+ T cells were markedly negative related to
Macrophages M0 ([130]Figure 9C). The stromal score, immune score, and
ESTIMATE score were all at a much lower level in the high-risk sample
([131]Figure 9D). The ssGSEA revealed that the infiltration levels of B
cells, CD8^+ T cells, iDCs, Mast cells, Neutrophils, pDCs, T helper
cells, Th1 cells, TIL, and Treg were remarkably lower in high-risk
patients ([132]Figure 9E). In high-risk samples, immune functions of
CCR, checkpoint, cytolytic activity, HLA, inflammation promoting, T
cell coinhibition, T cell costimulation, and type II IFN reponse were
less active in high-risk patients ([133]Figure 9F). The ssGSEA results
are consistent with the trend of lower a immune score. Given that
checkpoint and HLA levels were lower in high-risk patients, checkpoint
and HLA gene expression levels were then assessed. Overall, most of the
immune checkpoints were substantially lower in the high-risk group,
including CTLA4, PD-1 (also known as PDCD1, or CD279), and others
([134]Figure 9G). Similarly, the vast majority of HLA genes were at low
levels in high-risk patients ([135]Figure 9H).
FIGURE 9.
[136]FIGURE 9
[137]Open in a new tab
Immune infiltration analysis. (A,B) The bar chart and heatmap revealed
the percentage of 22 infiltrating immune cells in the TCGA. (C) The
correlation between infiltrating immune cells. (D) Comparison of the
Stromal, Immune, and ESTIMATE scores between different risk groups.
(E,F) The ssGSEA of the proportion of immune cells infiltration and
immune function in different risk groups. (G) The differential
expression of immune checkpoints between low- and high-risk groups. (H)
The differential expression of HLA genes between different risk groups.
*p < .05, **p < .01; ***p < .001; ns, not significant.
4.7 Drug susceptibility analysis
Initially, we identified six clinically used drugs (Gemcitabine,
Paclitaxel, Oxaliplatin, Olaparib, Fluorouracil, and Erlotinib) from
the candidate drugs for PC in the CTRP database ([138]Figure 10A).
High-risk patients were more sensitive to Erlotinib and Fluorouracil,
while the sensitivity to Olaparib and Oxaliplatin was higher in
low-risk patients ([139]Figures 10B–G).
FIGURE 10.
[140]FIGURE 10
[141]Open in a new tab
Drug sensitivity analysis. (A) Venn diagram of the clinical use of the
drugs for PC. (B–G) The sensitivity to six drugs (Erlotinib,
Fluorouracil, Gemcitabine, Paclitaxel, Olaparib, and Oxaliplatin) in
different risk patients.
5 Discussion
PC is expected to be the second major cause of cancer mortality by the
third decade of the 20th century, imposing a huge burden on social
development and the physical and mental health of the population
([142]Rahib et al., 2014; [143]Siegel et al., 2021) (33433946,
24840647). The strong relationship between lncRNAs with cancer speaks
for itself, and the potential of lncRNAs as accurate diagnostic,
therapeutic, and prognostic biomarkers for malignancy is enormous
([144]Ponting et al., 2009; [145]Hauptman and Glavac, 2013) (23443164,
19239885). Consequently, we worked to establish a CRLncs prognostic
signature for prognostic prediction of PC to guide personalized
clinical management of patients.
In the current study, univariate Cox regression analysis identified 46
CRLncs associated with PC prognosis and developed a 6-CRLnc signature
for PC. The model classified patients into two levels of high- and
low-risk, with low-risk patients enjoying better clinical outcomes and
this risk score being an independent prognostic factor for PC. In this
study, six promising lncRNAs were identified, namely FAM83A-AS1,
[146]AL137186.1, SENCR, PAN3-AS1, [147]Z97832.2, and [148]AL139243.1.
Recent studies reported that lncRNA FAM83A-AS1 exerted an important
role in the proliferation, migration, invasion, autophagy, and
progression of epithelial-mesenchymal transition (EMT) in lung
adenocarcinoma ([149]Chen et al., 2022; [150]Huang et al., 2022;
[151]Zhao et al., 2022) (35002507, 35164653, 35635086). The biological
roles of lncRNA SENCR are mainly in stabilizing vascular endothelial
cell adhesion junctions and attenuating the proliferation and migration
of vascular smooth muscle cells ([152]Zou et al., 2015; [153]Lyu et
al., 2019; [154]Song et al., 2022) (30584103, 35351345, 26349960). Ping
et al. found that PAN3-AS1 was negatively associated with the prognosis
of PC ([155]Ping et al., 2022) (35330729). Nevertheless, no relevant
reports on [156]AL137186.1, [157]Z97832.2, and [158]AL139243.1 have
been reported in cancer. The current study did not explore the
potential functions and molecular mechanisms of these lncRNAs, but to a
certain extent, it provided a theoretical basis for the link between
lncRNAs with PC.
Replication immortality and cell cycle dysregulation are vital
hallmarks of cancer cells, and DNA replication is a central process in
cell proliferation ([159]Hanahan and Weinberg, 2011; [160]Qu et al.,
2017) (21376230, 28182015). In the present study, GSEA results
suggested that cell cycle and DNA replication related signaling
pathways were more active in high-risk patients. Hence, high-risk
patients exhibited more active tumor proliferative activity and faster
tumor growth and progression, which is consistent with their higher
malignancy. GSVA indicates that the Notch signaling pathway is more
active in PC patients of the C2 subtype. The Notch signaling pathway
has a vital role in cell development and contributes to cancer
development and progression by fundamentally affecting cellular
processes such as differentiation, proliferation, or migration
([161]Reicher et al., 2018) (29409809). In Mullendore’s work, the Notch
signaling pathway was identified to be involved in tumor initiation and
maintenance of PC ([162]Mullendore et al., 2009) (19258443).
Downregulation of the Notch signaling pathway in PC facilitates
inhibition of PC cell growth and invasion ([163]Wang et al., 2006)
(16510599). This implies that PC patients with the C2 subtype show a
greater propensity for tumor progression and is consistent with a worse
prognostic outcome.
Naturally, this study assessed whether the risk score was associated
with PC gene mutation. The current study revealed that KRAS, CDKN2A,
and TP53 mutations occurred at a considerably higher frequency in
high-risk patients than in the corresponding groups. Among them, KRAS
mutations and CDKN2A alterations were early events in the occurrence of
PC ([164]Kamisawa et al., 2016) (26830752). Mutations in the KRAS gene
will permanently activate KRAS proteins that induce cell proliferation,
migration, transformation, and survival by triggering various
intracellular signaling pathways and transcription factors
([165]Buscail et al., 2020) (32005945). The presence of KRAS mutation
was related to an inferior prognosis for PC patients regardless of
whether they underwent radical surgery ([166]Buscail et al., 2020)
(32005945). Whereas, mutation of CDKN2A increases the risk of PC
([167]Goldstein et al., 1995) (7666916). In human cancers, TP53 is the
most prevalently mutated gene, and wild-type TP53 is a pivotal tumor
suppressor gene in cancer. Nevertheless, about half of human cancers
harbor TP53 mutations ([168]Harris and Hollstein, 1993; [169]Muller and
Vousden, 2014; [170]Yue et al., 2017) (24651012, 8413413, 28390900).
Numerous experiments have confirmed that mutated TP53 will acquire
malignant biological functions in tumorigenesis, including promoting
tumor cell survival, proliferation, migration, and invasion, enhancing
chemoresistance, and promoting cancer metabolism ([171]Dittmer et al.,
1993; [172]Brosh and Rotter, 2009; [173]Muller and Vousden, 2013;
[174]Zhang et al., 2013) (23263379, 19693097, 8099841, 24343302).
Thereby, high-risk PC patients are expected to develop disease
progression.
A considerable amount of studies have covered that the infiltration of
immune cells in the TME is intimately related to the prognosis of
various cancers (e.g., ovarian, pancreatic, liver, breast, and
colorectal cancers) ([175]Zhang et al., 2003; [176]Ino et al., 2013)
(12529460, 23385730). B cells play an influential and complex role in
the host’s immune response to malignancy but are often
underappreciated. Given the reports, B cells can act both as
antigen-presenting cells to activate T cell responses to tumor cells
and also produce tumor-specific antibodies to bolster anti-tumor
immunity ([177]Fremd et al., 2013; [178]Ko et al., 2020) (32730744,
24073382). Dendritic cells (DCs) are the most potent antigen-presenting
cells in the body and play a pivotal role in the immune response, but
are often suppressed in TME ([179]Kranz et al., 2016; [180]Chen et al.,
2018) (27281205, 29290766). Interstitial dendritic cells (iDC) could
bind antigens and stimulate T lymphocyte responses, posing a
significant threat to organ graft survival after organ transplantation
([181]Hart and McKenzie, 1990) (2152498). The impaired activity of
plasmacytoid dendritic cells (pDC) is associated with immunodeficiency
status or inefficient immune response to tumors ([182]Reizis, 2019)
(30650380). Regulatory T cells (Treg) exhibit a strong suppressive
capacity and are critical mediators of tumor-associated
immunosuppression ([183]Liu et al., 2016; [184]Plitas and Rudensky,
2016) (27590281, 26787424). Yet other studies pointed to the opposite
function of Treg, that of tumor restriction. Studies suggested that
Treg may promote tumor development by limiting anti-tumor immunity and
also limit tumor development by restricting the mesenchymal environment
required for its growth and metastasis ([185]Tzankov et al., 2008;
[186]Liu et al., 2016) (26787424, 18223287). The exact role of Treg in
the PC microenvironment is not yet fully understood, and the current
study found higher levels of Treg infiltration in PC patients with low
risk, which may be explained by the potential for Treg to limit PC cell
growth and metastasis hence limiting tumor progression. The cytotoxic
potential of CD8^+ T cells is the backbone of current immunotherapy,
and their high infiltration levels are associated with improved OS in
cancer ([187]Manzo et al., 2020; [188]Orhan et al., 2020) (32491160,
32334338). High infiltration of tumor-infiltrating lymphocytes (TIL)
could provide a rationale for differentiating immunogenic hot tumors
from degenerative or cold tumors ([189]Orhan et al., 2020) (32334338).
The presence of TIL and high infiltration of CD8^+ T cells often
predicts response to immunotherapy and prognosis ([190]Galon and Bruni,
2019; [191]Orhan et al., 2020) (30610226, 32334338). The present study
observed markedly higher levels of both CD8^+ T cells and TIL in
low-risk PC patients, with additional higher levels of expression of
immune checkpoints like CTLA4 and PD-1, which seems to imply that
patients with low-risk could benefit more from receiving immunotherapy.
Of course, the currently accepted view is that high levels of CD8^+ T
cells and TIL infiltration do not represent an absolute enhancement of
the efficacy of immunotherapy for PC and that the activity and
depletion of immune cells, as well as the nature of TME, profoundly
influence the response to immunotherapy ([192]Galon and Bruni, 2020)
(31940273). This could also explain, to some extent, the failure of
patients with different risk PC to show significantly different
responses to immunotherapy in the current study. Overall, the road to
tangible benefits from immunotherapy for PC patients is difficult and
urgently awaits further exploration.
Indeed, the current study has some limitations. The reliability and
stability of this prognostic signature should be further validated in a
prospective study at a large PC research center. All these concerns
should be refined in future studies.
6 Conclusion
Novel lncRNA signature and nomogram for cuproptosis were constructed to
predict the prognosis of PC patients. Molecular characterization based
on this model provides a new perspective on the progression and subtype
of PC. Differences in immune cell infiltration and immune checkpoint
expression may be an important indication for the prognosis and
treatment of PC patients. The differences in drug sensitivity might be
a promising opportunity for the treatment of different risk PC
patients.
Acknowledgments