Abstract
Prostate cancer (PCa) is a common and serious health issue among older
men globally. Metabolic reprogramming, particularly involving lactate
and mitochondria, plays a key role in PCa progression, but studies
linking these factors to prognosis are limited. To identify novel
prognostic markers of PCa based on lactate-mitochondria-related genes
(LMRGs), RNA sequencing data and clinical information of PCa from The
Cancer Genome Atlas (TCGA) and the cBioPortal database were used to
construct a lactate-mitochondria-related risk signature. Here, we
established a novel nine-LMRG risk signature for PCa, and Kaplan-Meier
curves confirmed a worse prognosis for high-risk subgroups in the TCGA
dataset. Meanwhile, a nomogram that effectively predicts the prognosis
of PCa patients was also constructed. Next, close associations between
the lactate-mitochondria-related signature and the immune
microenvironment were examined to clarify the role of LMRGs in shaping
the immune landscape. Furthermore, as the only lactate-related gene
among the nine key prognostic risk genes, myeloperoxidase (MPO) was
identified as a key factor that mediates lactate production in vitro
and in vivo through attenuation of the glycolytic pathway. More
importantly, MPO significantly inhibited PCa cell migration, invasion,
and epithelial–mesenchymal transition (EMT), indicating its potential
as an anticancer gene. Additionally, PCa with high MPO expression is
highly sensitive to chemotherapeutic agents and mitochondrial
inhibitors, highlighting its potential as an improved therapeutic
strategy for PCa management.
Keywords: prostate cancer (PCa), lactate-mitochondria-related genes
(LMRGs), prognosis, myeloperoxidase (MPO), metastasis, drug sensitivity
Introduction
Prostate cancer (PCa) is a malignant tumor that originates from the
epithelial cells of the prostate gland, and cause significant mortality
among elderly men. Currently, approximately 1.4 million new cases and
0.4 million deaths are reported annually ([45]Bray et al., 2024).
Although advancements in technology have significantly enhanced the
early diagnosis of PCa, the survival rate has not significantly
improved, because of the high rate of bone metastasis and insensitive
to treatment. Radical prostatectomy (RP) or radiotherapy is the
first-antineoplastic treatment for patients in early stage with
localized PCa ([46]Siegel et al., 2017). However, 30%–50% of patients
will still progress to biochemical relapse after taking treatment
([47]Lalonde et al., 2014). Around 20% of intermediate-risk patients
face biochemical failure within 18 months of initial local treatment
([48]Shao et al., 2009; [49]Nichol et al., 2005). The oncogenic
mechanisms that drive PCa are not yet well understood, making it
challenging to implement targeted therapy for high-risk PCa and
castration-resistant prostate cancer (CRPC) ([50]Xie et al., 2018;
[51]Li et al., 2020). Therefore, gaining a deeper understanding of the
various characteristics of PCa and identifying of effective prognostic
indicators are essential for developing more effective treatment
strategies for PCa.
Metabolic reprogramming is a acknowledged hallmark of cancer and
involves pathways such as glycolysis, oxidative phosphorylation, and
mitochondrial metabolism ([52]Li et al., 2020; [53]Mohsen et al., 2019;
[54]Delaunay et al., 2022). This shift in metabolic processes of cancer
cells enables them to satisfy their heightened energy requirements,
support rapid growth, and survive under various stress conditions.
Additionally, cancer cells often modulate the levels of metabolic
byproducts, such as lactate, to influence the tumor microenvironment,
thereby affecting the behavior of both cancer cells and immune cells
([55]Sarah et al., 2020). This intricate interplay between metabolic
pathways and cellular interactions is critical for tumor development
and progression, and thus this interplay is a focal point for potential
therapeutic interventions.
Traditionally, lactate has been considered a metabolic waste product
excreted by glycolytic prostate cancer cells into the microenvironment
([56]Pereira‐Nunes et al., 2020). However, with deeper research, it has
been discovered that lactate serves not only as the preferred energy
substrate for PCa cells but also plays a role in reprogramming their
metabolism through interactions with cancer-associated fibroblasts
(CAFs) ([57]Pierre et al., 2008; [58]Fiaschi et al., 2012). Metabolic
reprogramming in PCa is frequently associated with altered
mitochondrial function, as mitochondria are not only the site of
oxidative phosphorylation but are also the central hub of multiple
metabolic pathways, including the tricarboxylic acid cycle and fatty
acid oxidation ([59]Mamouni et al., 2021). Many studies suggest that,
metabolic reprogramming is associated with changes in mitochondrial
bioenergetics, biogenesis and dynamics during PCa developmentt
([60]Vikramdeo et al., 2023; [61]Haokun et al., 2024; [62]Xiao et al.,
2018). The increase in ROS and sulfide oxidation flux, along with the
reduction of ATP generation, may exacerbate PCa malfunction and lead to
higher grade malignancy ([63]Bee et al., 2021). Elucidation the
intricate molecular mechanisms of mitochondrial-participated metabolic
reprogramming and exploration the role of these mechanisms in
carcinogenesis progression are essential for identifying innovative
therapeutic targets and strategies for PCa.
Lactate metabolism, which promotes tumor cell proliferation and
metastasis via modulating the tumor microenvironment, enhancing
angiogenesis, and suppressing the immune response, is also particularly
critical for cancer progression. Lactate is a byproduct of glycolysis
and is especially enriched in rapidly growing tumors. Accumulated
studies have confirmed that lactate serves as a high-energy substrate
that shuttles between the cytoplasm (glycolysis) and mitochondria
(oxidative phosphorylation) ([64]Daniel and Kane, 2014; [65]Safer et
al., 1971; [66]George et al., 1999). Lactate participates in
mitochondrial oxidative reactions through the lactate-malate-aspartate
shuttle and functions as an energy substrate that enhances energy
support and regulates androgen metabolism; therefore, lactate
potentially offers new therapeutic avenues for PCa ([67]Daniel and
Kane, 2014; [68]Glancy et al., 2021).
Although numerous studies have demonstrated the importance of lactate
and mitochondrial function on tumor progression, the specific genes
related to the association between these two pathways that could serve
as prognostic markers for PCa remain unclear. This knowledge gap
hinders the development of precise prognostic tools and effective
therapeutic strategies. This study aimed to explore and identify LMRGs
with prognostic significance in PCa. We conducted a comprehensive
bioinformatics analysis using data from large-scale PCa studies.
Differentially expressed genes were identified via statistical methods,
and survival ratio analysis was conducted to assess their prognostic
significance. We established a novel nine-LMRG signature and
accompanying nomogram that can accurately forecast the outcome of PCa
patients. Furthermore, we demonstrated that myeloperoxidase (MPO), a
lactate metabolism-related gene, plays a critical role in mediating
lactate production by modulating the glycolytic signaling pathway,
which results in significant inhibition of EMT, migration, and
invasiveness of PCa cells. Additionally, PCa cells with higher levels
of MPO expression show more sensitivity to chemotherapeutic agents and
mitochondrial inhibitors, which highlights its potential to improve the
current therapeutic strategies for PCa management.
Materials and methods
Data collection
Fragments per kilobase transcript (FPKM) data and corresponding
clinical information of PCa patients were obtained from the TCGA-PRAD
dataset of TCGA database ([69]https://portal.gdc.cancer.gov), which
includes 52 normal samples and 502 tumor samples. Clinical data
([70]Supplementary Table S1) were obtained from cBioPortal’s collection
of clinical data from PCa patients ([71]https://www.cbioportal.org).
External clinical data of TCGA-PRAD ([72]Supplementary Table S2) were
obtained from the UCSC Xena platform
([73]https://xenabrowser.net/datapages/). The data were processed and
analyzed via Perl software (version Strawberry-perl-5.30.0.1;
[74]https://www.perl.org) and the R Bioconductor package in R software
(version R-4.4.1).
Screening of lactate and mitochondria-related genes
The list of lactate-related genes ([75]Supplementary Table S3) was
compiled from relevant literature on lactate-associated gene sets
([76]Jiang et al., 2023). The list of mitochondria-associated genes
presented in [77]Supplementary Table S4 was compiled from well-curated
datasets ([78]Chang et al., 2023), including the MitoCarta 3.0 database
([79]Rath et al., 2020) and the molecular signatures database (MSigDB)
([80]Mootha et al., 2003; [81]Subramanian et al., 2005). Differentially
expressed genes (DEGs) between PCa and normal tissues were identified
using the “limma” package in R. Genes were considered significantly
differentially expressed in PCa samples relative to normal tissues if
they had an absolute log2-fold change (logFC) greater than 0.585
(equivalent to a fold change exceeding 1.5) and a false discovery rate
(FDR) below 0.05.
Development of prognostic risk features based on LMRGs
To further refine the DEGs, we applied a univariate Cox regression
analysis with a p-value <0.05 as the selection criterion to identify
the genes as lactate-mitochondria-related markers. To avoid
overfitting, we employed the “glmnet” and “survival” packages for LASSO
Cox regression analysis ([82]Friedman et al., 2010). Following LASSO
regression, multivariate Cox regression analysis was performed to
establish LMRGs based on the selected markers. The entire cohort was
randomly divided into training and testing groups at a 1:1 ratio for
internal validation. Patients were classified into high-risk and
low-risk categories on the basis of the risk scores from the training,
testing, and overall groups via median split values. The risk score was
calculated according to the following formula:
[MATH: Risk Score=∑i=1<
/mn>nGeneexp×Coefi
:MATH]
Here, “n” represents the number of mRNAs associated with PCa prognosis,
and “i” denotes the ith LMRG. The expression levels of LMRGs and the
regression coefficients are represented by Geneexp and Coefi,
respectively. The risk score for each patient was predicted using the
“predict” function included in the “survival” R package. Patients were
divided into two subgroups, the LMRG-high-risk subgroup and the
LMRG-low-risk subgroup, according to the median LMRG score.
Validation of prognostic risk features
The prognostic value of the LMRGs was assessed via Kaplan‒Meier (KM)
survival analysis, which compared the progression-free survival (PFS)
rates between the two LMRG groups from the TCGA database. We further
explored the model’s predictive capabilities concerning clinical
variables such as age, T stage, N stage, and the risk score. To ensure
the robustness of the model, multivariate independent prognostic
analysis was conducted, and ROC curves were generated for these
clinical features using the “timeROC,” “survival,” and “survminer” R
packages. In addition, we used prognostic data from TCGA-PRAD,
including disease-free interval (DFI), disease-specific survival (DSS),
and overall survival (OS), for external validation to evaluate the
model’s generalizability.
Immune microenvironment evaluation
Following the global immune classification of solid tumors developed by
([83]Thórsson et al., 2018), we identified the following four distinct
immune subtypes: C1 (wound healing), C2 (IFN-γ-dominant), C3
(inflammatory), C4 (lymphocyte-depleted). To determine the relationship
between risk scores and immune phenotypes, the “ggpubr” R package was
utilized. To evaluate immune cell infiltration in PCa patients, both
the “CIBERSORT” and “xCell” R packages were utilized. “CIBERSORT”
estimated the abundances of 22 specific immune cell types, while
“xCell” provided a broader analysis, assessing 64 distinct cell types
based on gene expression data. The immune-related characteristics were
defined according to the previous studies, and the scores were
calculated via gene set variation analysis (GSVA).
Drug sensitivity calculation
We used the “OncoPredict” ([84]Maeser et al., 2021) R package to
evaluate the potential clinical applications of LMRGs in PCa treatment.
OncoPredict predicts drug response according to the RNA-Seq gene
expression data. Specifically, we utilized the calcPhenotype function
to estimate the drug response curves (IC50 values) for commonly used
chemotherapeutic drugs via baseline tumor gene expression data from the
TCGA database. The Wilcoxon signed-rank test was then employed to
compare the IC50 values between different LMRG risk groups to identify
statistically significant differences.
Functional enrichment
To examine the activity of key pathways and identify differences
between the high- and low-risk groups, gene set variation analysis
(GSVA) was conducted. The R package “clusterProfiler” ([85]Wu T. et
al., 2021) was employed to conduct Gene Ontology (GO) and Kyoto
Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses on
the nine-LMRG, showcasing the top 10 results. These results were
visualized using bubble charts. The GO analysis covered terms related
to biological processes (BP), cellular components (CC), and molecular
functions (MF).
Antibodies and reagents
Anti-MPO (Cat# 11117-1-AP) antibody was purchased from Proteintech.
Anti- FAK (Cat# 3285), anti-pFAK Tyr925 (Cat# 3284), anti-Snail (Cat#
3879), anti-E-Cadherin (Cat# 3195) and anti-actin (Cat# 4970)
antibodies were purchased from Cell Signaling Technology. Goat
anti-rabbit (Cat# 31210) and anti-mouse (Cat# 31160) secondary
antibodies were purchased from Thermo Fisher Scientific. Docetaxel
(Cat# HY-B0011) and paclitaxel (Cat# HY-B0015) were purchased from
MedChemExpress. antimycin A (Cat# A8674) was purchased from
Sigma-Aldrich.
Cells and cell culture
Normal prostate cells WPMY-1, RWPE1, RWPE2, and prostate adenocarcinoma
cells C4-2, LNCaP, PC3, and DU145 were obtained from Xiamen Immocell
Biotechnology Co., Ltd. (Xiamen, China). These cells were cultured in
Dulbecco’s Modified Eagle’s Medium (DMEM) (Sigma Aldrich, St. Louis,
United States) or in RPMI-1640 medium (Sigma Aldrich, St. Louis, United
States), according to the supplier’s specifications. Both media were
supplemented with 10% fetal bovine serum (FBS) (Genial Biologicals,
Inc., Brighton, United States). Regular mycoplasma tests confirmed no
contamination.
Western blot
Cells were lysed using ELB lysis buffer containing both protease and
phosphatase inhibitor cocktails. After lysis, the samples underwent
centrifugation at 14,000 × g for 15 min at 4°C. The supernatants
obtained were mixed with 2 × SDS sample buffer and heated at 95°C–100°C
for 5–10 min to ensure denaturation. Subsequently, the samples were
applied to SDS-PAGE gels for electrophoresis, transferred onto PVDF
membranes, and then analyzed via immunoblotting using specific
antibodies.
Generation of the lentiviral system
The oligonucleotides for short hairpin RNAs (shRNAs) were subcloned
into the lentiviral vector pLL3.7, and expressed in LNCaP cells. First,
lentiviruses were generated in HEK293T cells through co-transfection
with the pLL3.7 vector carrying shRNA sequences, packaging plasmids,
and polyethylenimine (PEI) for 48 h. After harvesting the viral
supernatants, viruses were concentrated by centrifuging at 75,000 × g
for 1.5 h, and subsequently filtered through 0.45 μm pore-size
membranes (Millipore). The LNCaP cells were infected with the freshly
isolated lentiviruses, and knockdown efficiency was tested using
reverse transcription PCR (RT-PCR) after 48 h of incubation. The
oligonucleotide sequences for the construction of the shRNA-targeted
mRNAs were listed as below: Non-Targeting Control (NTC)-shRNA, 5′-
CCTAAGGTTAAGTCGCCCTCG-3′; MPO-shRNA-1, 5′-
GCCATGGTCCAGATCATCACT-3′; MPO-shRNA-2, 5′-
GCAGTACACTTCCTGCATTGA-3′; MPO-shRNA-3, 5′-
GGTTATGTGTATGTGCCATTT-3′.
Extracellular acidification rate (ECAR) assay
To assess the extracellular acidification rate of the cells, a
Glycolysis Stress Test Kit from Agilent Technologies (Santa Clara, CA,
United States) and a Seahorse XFe96 Extracellular Flux Analyzer were
employed. In brief, 10 ([86]Shao et al., 2009) cells were plated per
well in Seahorse XF96 cell microplate and incubated for 24 h. Glucose
(10 mM), followed by the oxidative phosphorylation inhibitor oligomycin
(1 μM) and the glycolysis inhibitor 2-DG (50 mM), were sequentially
injected into each well. The Seahorse XFe96 instrument was used to
capture the dynamic fluorescence signals. Finally, all the data were
normalized to the cell count.
Cell cycle detection
To detect the cell cycle, cells were initially collected and rinsed
with PBS. They were then fixed in 70% ethanol, followed by another PBS
wash. Afterward, the cells were stained with propidium iodide (PI) for
30 min in darkness. Flow cytometry was used to analyze the stained
cells.
In vitro cell migration and invasion assays
In vitro cell migration experiments were conducted using 8-mm polyester
Transwell chambers (Corning, New York, United States). For cell
invasion assays, a method akin to the migration assay was followed, as
previously detailed by Wang et al. ([87]Wang et al., 2019). The
protocol was similar to that used for the migration experiments, except
that the chambers were coated with growth factor-reduced Matrigel
beforehand. Briefly, LNCaP cells were placed, in triplicate, into the
Transwell chambers at a density of 2 × 10 ([88]Nichol et al., 2005)
cells per well in 0.1% BSA RPMI 1640 medium. As a chemoattractant,
conditioned medium from NIH3T3 cells was collected and added to the
lower chamber. Following a 16-h incubation period, any non-invading or
non-migrating cells were removed from the upper membrane surface. The
cells that had migrated or invaded to the underside of the Transwell
insert were then fixed with methanol and stained with crystal violet.
Images of five random fields at ×10 magnification were captured for
each membrane and analyzed using ImageJ software.
Metabolite analysis by LC‒MS
A total of 5 × 10^6 cells were thoroughly rinsed 3 times with cold PBS
(4°C). Metabolites from each sample group were extracted using 1.6 mL
of 80% methanol chilled to −80°C. The extracts, along with the cells,
were placed into 2 mL tubes, subjected to vortex mixing for 1 min, and
then centrifuged at 140,00× g and 4°C for 10 min. The supernatants were
dried using a vacuum centrifuge (Labconco Corporation). The dried
samples were reconstituted with 200 μL of 50% acetonitrile. A volume of
2 μL from each sample was injected into a QTRAP 5500 mass spectrometer
(SCIEX) connected to a UPLC system (AB Sciex, ExionLC AD system), and
separated on a ZIC-pHILIC column (SeQuant, 5 μm, 100 × 2.1 mm, Merck).
For the mobile phases, buffer A was made up of 15 mM ammonium acetate
with the pH adjusted to 9.7 using ammonium hydroxide, while buffer B
contained 90% acetonitrile. The column was kept at 40°C and the samples
at 10°C. A flow rate of 0.2 mL/min was maintained, with the gradient
set as follows: 95% B from 0 to 2 min, 45% B from 15 to 18 min, and
then back to 95% B from 18 to 22 min. The QTRAP instrument was operated
in negative ion mode using multiple reaction monitoring (MRM).
Lactate detection
Lactate release in the cell culture supernatants was analyzed using
Lactate-Glo assay kits (Promega Corp., Madison, WI, United States)
according to the manufacturer’s protocols. To quantify lactate
production, the culture medium was diluted at a 1:20 ratio with PBS and
then plated in 96-well plates. An equivalent amount of lactate
detection reagent, which includes reductase, lactate dehydrogenase,
reductase substrate, luciferin detection solution, and NAD, was added
to the wells. The plates were incubated at room temperature for 1 h,
after which luminescence was measured and normalized to the count of
viable cells.
MTT assay
Briefly, a total of 6 × 10³ LNCaP cells were plated in triplicate in a
96-well plate and treated with the indicated reagents in a final volume
of 200 μL per well for a duration of 72 h at 37°C. Cells receiving DMSO
acted as the control group. After then, 10 μL of MTT solution (5 mg/mL)
was added to each well, and the plates were incubated for another 4 h
at 37°C. Afterward, the MTT-containing medium was discarded, and 150 μL
of DMSO was re-added to each well to dissolve the formazan crystals.
The plates were incubated for an additional 10 min, and the absorbance
at 490 nm was then recorded using a microplate reader.
Statistical analysis
Differential expression box plots for MPO were obtained from TNMplot
([89]Bartha and Győrffy, 2020) (accessed on 15 August 2024). The
differences in the proportions of clinical characteristics were
analyzed via the chi-square test. Differences between KM curves were
assessed via the log-rank test. A p-value less than 0.05 indicated
statistical significance. Statistical analyses were performed via R or
GraphPad software. The current study investigated the publicly
available data, and no ethical approval was required. The logical flow
of the study is illustrated in [90]Figure 1.
FIGURE 1.
[91]FIGURE 1
[92]Open in a new tab
Research process flowchart.
Results
Identification of lactate-mitochondria-related differentially expressed genes
We initially integrated 206 lactate-related genes and 2,030
mitochondria-related genes to form a LMRG set. [93]Figure 1 shows the
workflow of the LMRG signature analysis. Briefly, RNA-seq data of 52
normal samples and 502 prostate adenocarcinoma (PRAD) samples were
downloaded from the TCGA database. In addition, the clinical
characterization and prognostic data of 494 TCGA-PRAD patients from
cBioPortal were integrated, and we censored the data showing NA. As
shown in the volcano plot ([94]Figure 2A), 443
lactate-mitochondria-associated DEGs were identified in the TCGA-PRAD
dataset ([95]Supplementary Table S5). From these, the top 50
upregulated genes and the top 50 downregulated genes were selected
according to the logFC and visualized in a heatmap ([96]Figure 2B).
FIGURE 2.
[97]FIGURE 2
[98]Open in a new tab
Identification of differentially expressed LMRGs and construction of
prognostic risk model. (A) Volcano plot and (B) heatmaps of
differentially expressed genes between PRAD and normal prostate tissue
of LMRGs. (C) The LASSO regression coefficient spectrum. (D)
Cross-validation of parameter selection in the LASSO model.
Construction of prognostic genes and a risk model for LMRGs
We integrated patient survival data and performed univariate regression
analysis, using a p-value below 0.05 as the statistical threshold to
determine genes significantly linked to survival rates. Among 443 DEGs,
145 LMRGs showed significant associations with prognosis
([99]Supplementary Table S6). Further analysis was performed on the 145
LMRGs using LASSO and multivariate Cox regression ([100]Figures 2C, D),
ultimately resulting in the construction of a prognostic risk signature
based on nine-LMRGs. The risk score was calculated using the following
formula:
[MATH: Risk Score=MPO expression×−1.9763<
mo>+KCNJ11 expression×−0.5236<
mo>+SPATA18 expression<
/mtext>×−0.3166<
mo>+ACSM1 expression×−0.2261<
mo>+HJURP expression×0.3196+<
mrow>REXO2 expression×0.4176+<
mrow>SLC25A29 expression
×0.5170+<
mrow>SCO2 expression×0.9966+<
mrow>SIAH3 expression×2.5355
:MATH]
Validation of the risk model
First, the aforementioned Risk Score model was used to score the PRAD
samples. Based on the median risk score, the samples were classified
into high-risk and low-risk groups. The entire cohort with accessible
clinical data (n = 416) was then randomly split into a training set (n
= 210) and a testing set (n = 206) in a 1:1 ratio, showing no
significant differences in clinical characteristics between the groups
(P > 0.05, [101]Supplementary Table S7). Patient prognosis was
effectively differentiated between the high-risk and low-risk groups,
with the low-risk group consistently showing better outcomes
([102]Figure 3A). For survival analysis, patients received risk scores
and were categorized into high-risk and low-risk groups. Overall,
higher risk scores correlated with increased recurrence rates in PCa
patients ([103]Figures 3B, C). In the low-risk group, genes ACSM1,
KCNJ11, SPATA18, and MPO were highly expressed, while the other five
key genes showed high expression in the high-risk group ([104]Figure
3D). The area under the ROC curve (AUC) values for 1, 3, and 5-year
predictions were all above 0.740 ([105]Figure 3E). We validated similar
outcomes in both the training and testing sets ([106]Supplementary
Figures S1A–J), confirming the model’s accuracy and effectiveness. To
assess the model’s generalizability, we conducted an extensive analysis
using prognostic data for disease-free interval (DFI), disease-specific
survival (DSS), and overall survival (OS) from TCGA-PRAD via the UCSC
Xena platform ([107]Supplementary Figures S2A–C). The findings show
that the risk model accurately predicts PCa recurrence and is
applicable to various other prognostic scenarios, underscoring its
potential as a reliable tool for evaluating survival outcomes in PCa
patients.
FIGURE 3.
[108]FIGURE 3
[109]Open in a new tab
Validation of the prognostic risk model in the overall group. (A)
Overall group recurrence survival curves. (B) Risk scores and group
distribution in the overall group. (C) Recurrence status distribution
map. (D) Frequency distribution of key genes in high- and low-risk
groups. (E) ROC curves for 1, 3, and 5 years.
Independent prognostic value of risk features
To assess the prognostic predictive power of the nine-LMRGs risk
signature and other clinical factors for PCa, univariate ([110]Figure
4A) and multivariate Cox regression analyses ([111]Figure 4B) were
performed. The results showed that the risk signature met the
significance threshold (p < 0.05) in both analyses, which indicates
that the nine-LMRGs risk signature has superior predictive ability
compared with known clinical factors such as age, T stage, and N stage.
According to the above analyses, we constructed a nomogram using the
prognostic risk signature to predict the 1-, 3-, and 5-year survival
probabilities of patients ([112]Figure 4C). To assess the reliability
of the model, 1-year, 3-year, and 5-year calibration curves were
plotted. The result showed that the points closely aligned with the
standard line, indicating a high degree of accuracy and consistency
between the predicted and observed outcomes ([113]Figure 4D).
Additionally, AUC curves of the risk model and nomogram were compared
with those of other clinical features, and the results showed that the
nine-LMRGs risk signature was the strongest predictor than those of
other clinical characteristics ([114]Figure 4E). These results
highlight the superior effectiveness of the risk model in predicting
patient prognosis.
FIGURE 4.
[115]FIGURE 4
[116]Open in a new tab
Determination of the effectiveness of the prognostic risk model. (A)
Univariate and (B) multivariate forest plots of risk scores. (C)
Construction of the nomogram. (D) Calibration curves for recurrence at
1, 3, and 5 years. (E) Comparison of the predictive effectiveness of
the risk score and nomogram with other clinical characteristics; ***P <
0.001, **P < 0.01, *P < 0.05.
Differences in immunity and drugs sensitivity between risk groups
Then we explored the correlation between risk scores and previously
reported immune subtypes. Correlation analysis revealed that the C3
(inflammatory) subtype was significantly different from the C1 (wound
healing), C2 (IFN-gamma-dominant) and C4 (lymphocyte-depleted) subtypes
([117]Figure 5A) and had the highest immune infiltration rate in both
high- and low-risk groups ([118]Figure 5B). Moreover, the frequencies
of the C1, C2, and C4 subtypes were higher in the high-risk group
compared to the low-risk group ([119]Figure 5B). To delve deeper into
the associations between PCa risk groups and immune cell infiltration,
we utilized the CIBERSORT algorithm. This analysis demonstrated that
the infiltration of regulatory T cells (Tregs) was significantly higher
in the high-risk group, whereas the infiltration levels of memory B
cells, resting memory CD4 T cells, and resting mast cells were markedly
lower compared to the low-risk group ([120]Figure 5C). To pinpoint
specific immune function subtypes that are activated in PCa, we
conducted further analysis of the correlation between immune functions
and risk scores using the CIBERSORT algorithm. The results showed that
in the low-risk group, immune functions were significantly linked to
mast cells, while T-cell co-stimulation was more pronounced in the
high-risk group ([121]Figure 5D).
FIGURE 5.
[122]FIGURE 5
[123]Open in a new tab
Comparison of immune profiles and drug sensitivity between risk groups.
(A) Intergroup comparison of immune subtypes. (B) Proportions of immune
subtypes in high- and low-risk groups. (C) Immune cell correlation of
risk groups. (D) Immune function correlation of risk groups. (E–H) Drug
sensitivity analysis of Docetaxel (E), Paclitaxel (F), Irinotecan (G)
and AZD5363 (H); ***P < 0.001, **P < 0.01, *P < 0.05.
We then assessed the sensitivity of high- and low-risk groups to common
anticancer drugs to find potential treatment strategies for PCa. The
findings indicated that the low-risk group responded better to clinical
chemotherapy and targeted therapies, such as docetaxel, paclitaxel, and
irinotecan ([124]Figures 5E–G). Meanwhile, the high-risk group showed
greater sensitivity to the AKT inhibitor AZD5363 ([125]Figure 5H).
These insights offer a theoretical foundation for selecting clinical
drugs in PCa treatment.
Enrichment analysis of key LMRGs and pathway comparison between groups
To further investigate the genes comprising the risk model, we
conducted GO functional enrichment analysis and KEGG pathway enrichment
analysis on the nine-LMRGs. As shown in the results, the BP terms were
involved in non-membrane-bounded organelle assembly, muscle system
process, and muscle tissue development. The CCs were mainly enriched in
sarcomere, myofibril, and contractile fiber, whereas the MFs were
enriched predominantly in actin binding, hormone activity, and
structural constituent of muscle ([126]Supplementary Figure S3A). The
KEGG enrichment results highlighted pathways related to the
cytoskeleton in muscle cells and motor proteins ([127]Supplementary
Figure S3B).
We subsequently performed GSVA to explore potential pathways
differentiating the risk groups. The results indicated that the
high-risk group was enriched primarily in the BASE EXCISION REPAIR,
HOMOLOGOUS RECOMBINATION, and DNA REPLICATION pathways, whereas the
low-risk group was enriched in the GLYCOLYSIS GLUCONEOGENESIS,
PROPANOATE METABOLISM, and PYRUVATE METABOLISM pathways
([128]Supplementary Figure S3C). The enrichment of these pathways in
the high-risk group suggests a focus on responding to DNA damage and
maintaining genomic stability, possibly due to oxidative stress and
mitochondrial dysfunction caused by lactate accumulation, which
exacerbates DNA damage. In contrast, the low-risk group was enriched in
pathways related to glycolysis/gluconeogenesis, propanoate metabolism,
and pyruvate metabolism, emphasizing the maintenance of mitochondrial
energy metabolism and metabolic homeostasis, thus preventing excessive
lactate accumulation. Based on pathway analysis, the results revealed
that lactate serves as a key factor in associating
mitochondrial-related functions and influences genomic stability as
well as metabolic pathways between the high- and low-risk groups.
Therefore, we selected the lactate-related genes for further
investigation.
MPO inhibits lactate production in PCa
As the only lactate-related gene among the 9 key prognostic risk genes,
MPO was identified as a protective factor according to the univariate
KM survival curve ([129]Figure 6A). Additional analysis via the TNMplot
revealed that MPO expression in PCa tissues was significantly lower
than that in normal tissues ([130]Figure 6B). We further discovered
that MPO expression levels are significantly associated with the
infiltration of various immune cells, including monocytes and dendritic
cells ([131]Supplementary Figures S4A, B). We further examined MPO
expression levels in PCa cell lines with varying degrees of malignancy.
Compared with normal prostate cells (WPMY-1, PNT1A, RWPE1, RWPE2) and
poorly metastatic PCa cells (C4-2, LNCaP), moderately metastatic PCa
cells (DU145, PC3) presented lower levels of MPO expression
([132]Figure 6C).
FIGURE 6.
[133]FIGURE 6
[134]Open in a new tab
MPO inhibits lactate production in PCa. (A) Univariate survival
analysis of MPO in TCGA-PRAD. (B) Differential gene expression analysis
of MPO in the TNM-Ploter. (C) MPO protein levels in PCa cell lines. (D)
The endogenous MPO expression in LNCaP cells was knocked down using a
lentiviral system. Actin was used to determine the amount of loading
proteins. (E) The ECAR of LNCaP cells was detected. (F) Intracellular
metabolites were extracted from LNCaP cells, and glycolytic metabolites
were measured by LC-MS. (G, H) The level of extracellular lactate in
the medium of LNCaP cells (G) and PC3 (H) was detected. All data are
presented as the mean ± SD of three independent experiments, ***P <
0.001, **P < 0.01, *P < 0.05.
MPO selectively oxidizes thiol-containing proteins, particularly those
involved in the glycolysis pathway, which leads to the disruption of
basal glycolysis ([135]Love et al., 2016). To further confirm this, we
established stable MPO knockdown LNCaP PCa cells, after which the
knockdown efficiency was measured by Western blot analysis ([136]Figure
6D). As demonstrated in [137]Figure 6E, MPO silencing resulted in
increases in both the basal and maximum ECARs in LNCaP cells. Moreover,
as shown in [138]Figure 6F, the levels of glycolytic metabolites,
particularly pyruvate and lactate, were elevated in the absence of MPO,
which supports its role in the inhibition of glycolysis. In addition to
increasing the ECAR, the absence of MPO also led to increased
extracellular lactate release ([139]Figure 6G). We further
overexpressed MPO in PC3 cells, where endogenous MPO is barely
detectable. Our study demonstrated that MPO overexpression
significantly reduced extracellular lactate release, as expected
([140]Figure 6H). In conclusion, our data confirm that MPO is not only
linked to the lactate pathway but that it also significantly suppresses
glycolysis and lactate production.
MPO inhibits prostate cancer metastasis
We subsequently explored the role of MPO in PCa. Our results indicated
that silencing MPO did not significantly impact cell proliferation
([141]Figure 7A) or cell cycle progression ([142]Figures 7B, C).
However, MPO knockdown via shRNA markedly enhanced the migration
([143]Figure 7D) and invasion ([144]Figure 7E) abilities of LNCaP
cells. These results imply that MPO plays more pivotal role in
regulating metastasis than in influencing cell proliferation in PCa.
FIGURE 7.
[145]FIGURE 7
[146]Open in a new tab
MPO inhibits prostate cancer metastasis. (A) The cell number of LNCaP
cells was counted at different time points. (B, C) Cell cycle analysis
through PI staining was detected by FACS (B), and the cell cycle
distribution was presented (C). (D, E) LNCaP cells were analyzed for
cell migration and invasion. Representative images of crystal
violet-stained migrated (D) or invaded (E) cells are presented (scale
bar, 100 μL). (F) The Tyr925-phosphorylated level of FAK and total FAK,
E-Cadherin, and Snail in LNCaP cells were detected by Western blot
analysis. Actin was used as a loading control. All data are presented
as the mean ± SD of three independent experiments; ns, not significant.
Focal adhesion kinase (FAK) is a tyrosine kinase situated at
extracellular matrix adhesion sites and is crucial for cell motility.
Furthermore, E-cadherin and Snail are well-known biomarkers of EMT.
LNCaP cells with MPO-shRNA exhibited increased total protein levels of
Snail and higher phosphorylation of FAK at the Tyr925 site, along with
a reduced protein level of E-cadherin ([147]Figure 7F). Collectively,
these findings demonstrate that MPO is crucial for metastasis
modulation in PCa LNCaP cells.
PCa with high MPO expression is highly sensitive to drugs
To investigate whether MPO affects PCa cell sensitivity to clinical
chemotherapy, we knocked down the expression of MPO in LNCaP cells.
Comparing with the cells in control group, the cells in MPO silencing
group lose the sensitivity to docetaxel ([148]Figure 8A) and paclitaxel
([149]Figure 8B). Moreover, overexpression of MPO significantly
increased sensitivity to docetaxel ([150]Supplementary Figure S5A) and
paclitaxel ([151]Supplementary Figure S5B). According to drug
sensitivity analysis in TCGA-PRAD, it was found that high MPO
expression is associated with increased sensitivity to paclitaxel
([152]Figure 8C). This is consistent with our experimental conclusions.
FIGURE 8.
[153]FIGURE 8
[154]Open in a new tab
PCa with high MPO expression exhibits higher drug sensitivity. (A, B)
LNCaP cells were treated with docetaxel (A) or paclitaxel (B) for 72 h,
and then cell viability was determined by MTT assays. (C) Drug
sensitivity analysis of TCGA-PRAD to paclitaxel. (D) LNCaP cells were
treated with Antimycin A for 72 h, and then cell viability was
determined by MTT assays. (E–H) LNCaP cells were treated with the
Antimycin A (10 μM) and analyzed for cell migration and invasion.
Representative images of crystal violet-stained migrated (E) or invaded
(G) cells are presented (scale bar, 100 μL). Quantification for
migrated cells (F) or invaded cells (H) are presented. All data are
presented as the mean ± SD of three independent experiments; ***P <
0.001, *P < 0.05; ns, not significant.
We then investigated, whether different expression levels of MPO
influence the reliance of PCa cells on mitochondrial
metabolism/oxidative phosphorylation. Antimycin A is wildly used for
mitochondrial dysfunction, which blocks mitochondrial complex III
(cytochrome bc1 complex) of the respiratory chain via disrupting
electron transport and ATP synthesis ([155]Tzung et al., 2001). LNCaP
control cells are vulnerable to mitochondrial inhibitors. Compared with
them, LNCaP cells with silenced MPO exhibited decreased sensitivity to
antimycin A ([156]Figure 8D). Similarly, antimycin A significantly
inhibited both migration and invasion, whereas MPO knockdown diminished
this inhibitory effect on migration and invasion ([157]Figures 8E–H).
These findings suggest a close relationship between mitochondrial
function and the glycolytic pathway, which is characterized by lactate
production. Therefore, targeting both pathways could offer an effective
approach for the treatment of PCa.
Discussion
PCa is the most common malignant cancer among men in Western countries,
with an incidence rate that continues to increase ([158]Bray et al.,
2024; [159]Sung et al., 2021). Despite the approval of abiraterone and
enzalutamide, mortality has only slightly decreased ([160]Bono et al.,
2011; [161]Howard et al., 2012). The deaths of PCa patients can be
broadly categorized into cancer causes and noncancer causes. The risk
of death due to noncancer causes is influenced by factors such as
stage, ethnicity, and treatment variations ([162]Guo et al., 2022;
[163]Guo et al., 2020). The incidence and mortality rates of PCa among
African American men in the United States are significantly higher than
those among White men, by at least 1.7 times ([164]Siegel et al.,
2022). While genetic predisposition contributes to the increased
incidence of PCa in African Americans, social factors play a more
critical role in prognosis. Studies have shown that African American
and white men with similar PCa stages can achieve comparable outcomes
when provided with equal access to healthcare ([165]Rana et al., 2020;
[166]Bergengren et al., 2023; [167]Dess et al., 2019) Uncontrollable
factors such as age, family history, and germline variations also pose
significant threats to survival ([168]Bergengren et al., 2023).
Aberrant metabolism in PCa, which involves the regulation of various
metabolic pathways, has been identified as a key factor in disease
progression and metastasis ([169]Chen et al., 2023). Lactate is
recognized as a crucial signaling molecule in cellular metabolism.
Lactate produced by cancer cells is secreted into the extracellular
environment, where it induces tumor progression by affecting the tumor
microenvironment ([170]Certo et al., 2020; [171]Ippolito et al., 2019).
In addition to being a byproduct of energy metabolism, lactate can act
as a signaling molecule that regulates mitochondrial dynamics
([172]Zhou et al., 2024). As organelles involved in biosynthesis and
energy generation, mitochondria enable cells to rapidly adapt to their
environment and are considered important mediators of tumorigenesis.
Recent experimental studies have confirmed the importance of balancing
mitochondrial function and glycolysis in cancer metastasis
([173]Delaunay et al., 2022). Therefore, a comprehensive investigation
into the roles of lactate and mitochondria-related genes in the
pathogenesis and prognosis of PCa is urgently needed.
In this study, we identified and validated LMRGs that play a
significant role in the prognosis of PCa patients. By integrating
large-scale RNA-seq data from the TCGA-PRAD dataset, we screened and
analyzed DEGs associated with lactate metabolism and mitochondrial
function. Among the 443 identified DEGs, we focused on 145 genes
significantly related to prognosis. Ultimately, we developed a
prognostic risk model based on LMRGs via LASSO and multivariate Cox
regression analysis. Our prognostic risk model effectively stratified
patients into high-risk and low-risk groups, with significant
differences in PFS observed between the two groups. The model
demonstrated its validity not only in PFS but also in OS, DSS and DFI
outcomes. To explore the underlying mechanisms, we analyzed the
differences in immune and pathway activities between risk groups. The
GSVA results revealed strong associations between different risk groups
and various metabolic pathways involving lactate and mitochondria,
which further confirms their critical role in PCa prognosis.
Additionally, GO and KEGG enrichment analyses of risk-related genes
highlighted key cellular functions and pathways involved. The analysis
of immune cell infiltration and related immune functions, combined with
drug sensitivity analysis of common PCa treatments, suggested that the
risk model could effectively guide clinical therapy for PCa patients.
Notably, the drug sensitivity analysis included AZD5363 and docetaxel,
both of which are included in the ProCAID clinical trial, which
indicates potential benefits for guiding related clinical studies
([174]Crabb et al., 2017). Although our drug sensitivity analysis did
not identify the key drugs used in androgen deprivation therapy (ADT)
for prostate cancer, AZD5363 combined with docetaxel has shown an
increase in median overall survival (mOS) in metastatic
castration-resistant prostate cancer (mCRPC). This potentially
indicates that our model has a unique predictive value for the mCRPC
population ([175]Simon et al., 2022). Furthermore, we conducted an
in-depth analysis of the protective gene MPO within the risk model,
which highlights its association with immune cell subpopulations and
confirms its crucial role as a lactate-related gene that can influence
PCa cell migration and invasion.
Our GO analysis of the nine mitochondria-related genes revealed that
the genes were associated with muscle development, adaptation, and
function, particularly at the levels of skeletal and striated muscle.
These functions are crucial for organismal movement and environmental
adaptation ([176]Stefano et al., 2011), which may be linked to the
metabolic plasticity of tumors. The CC terms were enriched in specific
muscle cell structures, and the MF terms were related primarily to
muscle contraction and the regulation of muscle structure. This
indicates that the key genes in the risk model are involved in motor
function and metabolic regulation ([177]Olivença, 2023; [178]Brendan et
al., 2013). Mitochondria are the primary source of ATP during muscle
contraction and maintenance. During intense exercise and in cases of
mitochondrial dysfunction, lactate accumulates through anaerobic
glycolysis, which provides additional energy to the muscle. However,
lactate accumulation can lead to an acidic environment, which affects
calcium channels and muscle protein function; this in turn alters the
microenvironment and potentially induces tumor development ([179]Ann et
al., 2006; [180]George and Brooks, 2018; [181]Wu Danchen et al., 2021).
Our findings align with the current understanding of metabolic
reprogramming in cancer and extend this understanding, particularly in
the context of PCa. For example, Wang et al. reported that lactate acts
as a signaling molecule to promote tumor immune evasion by modulating
immune cell function ([182]Anushka et al., 2019). Our study further
enriches this field by identifying the relevance of specific lactate
metabolism- and mitochondria-related genes in the prognosis of PCa.
The identification and validation of LMRGs as prognostic biomarkers for
PCa is the most significant contribution of this study. By constructing
robust LMRGs, we not only provide a novel tool for stratifying patients
according to risk, but we also identify potential targets for
therapeutic intervention. This study bridges the gap between metabolic
reprogramming and clinical outcomes, offering a comprehensive
understanding of how alterations in lactate metabolism and
mitochondrial function drive PCa progression and impact patient
survival. The validation of MPO as a key gene and its potential role as
a protective factor in PCa further highlights the importance of these
key targets and their metabolic pathways in cancer biology. In summary,
our research contributes to the development of more personalized and
effective treatment strategies for PCa.
Research has demonstrated that MPO is significantly linked to the risk
of prostate cancer (PCa) through single nucleotide polymorphisms
([183]Ding et al., 2013). MPO is a peroxidase that contains heme and
catalyzes the formation of oxidants, such as hypochlorous acid (HOCl)
and hypothiocyanous acid (HOSCN), from H[2]O^2 and halide or
pseudo-halide ions. These oxidants selectively oxidize proteins
containing thiol groups, particularly those involved in the glycolytic
pathway ([184]Love et al., 2016; [185]Lin et al., 2024). In this study,
we revealed that MPO is connected to lactate metabolism and is vital in
modulating lactate production by influencing the glycolytic pathway.
This regulation leads to a marked reduction in migration, invasion, and
EMT in PCa cells. The transforming growth factor-β (TGF-β) pathway
plays a key role in fostering tumor metastasis and EMT across various
tissue types ([186]Anushka et al., 2019). Previous studies revealed
that lactate can influence TGF-β-related pathways, thereby enhancing a
tumor’s invasive characteristics ([187]Baumann et al., 2009). For
example, PCa cells can exploit lactate to promote PKM2/HIF-1-mediated
transcriptional regulation and facilitate EMT ([188]Giannoni et al.,
2015). Therefore, substantial evidence indicates that lactate could be
an upstream regulator of TGF-β, a crucial factor in EMT. The study by
([189]Çakıcı et al., 2024) confirmed that reducing HIF-1α expression
through the combination of inhibitors and chemotherapy drugs, thereby
inhibiting the metabolic reprogramming of the EMT mechanism, can affect
PC cell apoptosis and metastasis. Lactate enhances HIF-1α lactylation
through the lactate transporter MCT-1, stimulating angiogenesis in PCa
and influencing PCa proliferation and migration ([190]Yongwen et al.,
2022). Further research is needed to determine whether these mechanisms
can form a conceptual and functional feedback loop in the tumorigenesis
of PCa.
Furthermore, PCa cells that exhibit elevated MPO levels are sensitive
to mitochondrial inhibitors. We treated PCa cells with antimycin A to
disrupt electron transport, which resulted in mitochondrial
dysfunction. Compared with control cells, which are vulnerable to the
mitochondrial inhibitor antimycin A, LNCaP cells in which MPO was
knocked down exhibited decreased sensitivity, migration, and invasion
in response to this inhibitor. These data indicate that both
mitochondrial metabolism and glycolysis may be required for PCa
progression and metastasis. Compared with normal cells, cancer cells
exhibit increased glycolysis due to the Warburg effect ([191]Liberti et
al., 2016). However, recent studies have shown that oxidative
phosphorylation (OXPHOS) is as important as glycolysis in certain
cancers ([192]Wang and Patti, 2023). Several drugs, including
metformin, atovaquone, and arsenic trioxide, are used clinically as
OXPHOS inhibitors ([193]Weinberg et al., 2015; [194]Ashton et al.,
2016; [195]Wang et al., 2016). Therefore, in this study, we highlight
novel applications of OXPHOS inhibitors in PCa cells with high MPO
expression, highlighting their potential to improve therapeutic
strategies for PCa management.
Conclusion
In summary, this study is the first to construct a prognostic model for
PCa based on LMRGs and provides effective guidance for the prognosis
and drug treatment of PCa patients. Additionally, we investigated the
role of the lactate-related gene MPO as a key factor that mediates
lactate production by attenuating the glycolytic pathway, which leads
to significant inhibition of migration, invasion, and EMT and increased
drug sensitivity in PCa cells. The prognostic model and the MPO gene
identified in this study not only offer insights into the metabolic
basis of PCa but also present potential strategies for its treatment.
Funding Statement
The author(s) declare that financial support was received for the
research, authorship, and/or publication of this article. This work was
generously supported by funding from the National Natural Science
Foundation of China (No. 82073017); the Fundamental Research Funds for
the Central Universities, Grant numbers ZYGX2021J024 and
Y030222059002005; Fujian Clinical Research Center for Radiation and
Therapy of Digestive, Respiratory and Genitourinary Malignancies
(2021Y2014); the National Natural Science Foundation of China
(82473376); Science and Technology Pilot Program of Fujian Province,
China (2021Y0053).
Data availability statement
The original contributions presented in the study are included in the
article/[196]Supplementary Material, further inquiries can be directed
to the corresponding authors.
Ethics statement
Ethical approval was not required for the studies involving humans
because TCGA and GEO data, used in this study, belong to public
databases. The patients involved in the database have obtained ethical
approval. Users can download relevant data for free for research and
publish relevant articles. The studies were conducted in accordance
with the local legislation and institutional requirements. The
participants provided their written informed consent to participate in
this study.
Author contributions
YW: Funding acquisition, Project administration, Writing–original
draft, Writing–review and editing. RC: Data curation, Project
administration, Software, Writing–original draft. F-LJ:
Conceptualization, Writing–review and editing. XJ: Conceptualization,
Writing–review and editing. YuZ: Project administration, Validation,
Writing–review and editing. YiZ: Formal Analysis, Validation,
Writing–review and editing. XH: Data curation, Formal Analysis,
Software, Writing–review and editing. CL: Formal Analysis, Methodology,
Software, Writing–review and editing. W-JW: Conceptualization, Funding
acquisition, Supervision, Writing–original draft, Writing–review and
editing. SQ: Funding acquisition, Supervision, Writing–original draft,
Writing–review and editing.
Conflict of interest
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Generative AI statement
The author(s) declare that no Generative AI was used in the creation of
this manuscript.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or claim
that may be made by its manufacturer, is not guaranteed or endorsed by
the publisher.
Supplementary material
The Supplementary Material for this article can be found online at:
[197]https://www.frontiersin.org/articles/10.3389/fgene.2024.1515045/fu
ll#supplementary-material
[198]Table2.xlsx^ (68.3KB, xlsx)
[199]Image3.jpeg^ (1.9MB, jpeg)
[200]Table3.xlsx^ (14.6KB, xlsx)
[201]Image1.jpeg^ (1.6MB, jpeg)
[202]Image4.jpeg^ (554.3KB, jpeg)
[203]Image2.jpeg^ (707.8KB, jpeg)
[204]Image5.jpeg^ (245.4KB, jpeg)
[205]Table6.xlsx^ (19.2KB, xlsx)
[206]Table4.xlsx^ (40.6KB, xlsx)
[207]Table1.xlsx^ (170KB, xlsx)
[208]Table5.xlsx^ (47.2KB, xlsx)
[209]Table7.xlsx^ (33.7KB, xlsx)
References