Abstract
Acute myeloid leukaemia (AML) is the most common acute leukaemia in
adults, with an unfavourable outcome and a high rate of recurrence due
to its heterogeneity. Dysregulation of fatty acid metabolism plays a
crucial role in the development of several tumours. However, the value
of fatty acid metabolism (FAM) in the progression of AML remains
unclear. In this study, we obtained RNA sequencing and corresponding
clinicopathological information from the TCGA and GEO databases.
Univariate Cox regression analysis and subsequent LASSO Cox regression
analysis were utilized to identify prognostic FAM-related genes and
develop a potential prognostic risk model. Kaplan-Meier analysis was
used for prognostic significances. We also performed ROC curve to
illustrate that the risk model in prognostic prediction has good
performance. Moreover, significant differences in immune infiltration
landscape were found between high-risk and low-risk groups using
ESTIMATE and CIBERSOT algorithms. In the end, differential expressed
genes (DEGs) were analyzed by gene set enrichment analysis (GSEA) to
preliminarily explore the possible signaling pathways related to the
prognosis of FAM and AML. The results of our study may provide
potential prognostic biomarkers and therapeutic targets for AML
patients, which is conducive to individualized precision therapy.
Keywords: fatty acid metabolism, prognosis signature, acute myeloid
leukaemia
1. Introduction
Acute myeloid leukaemia (AML) is a heterogeneous haematological
malignancy characterized by abnormal proliferation and differentiation
of haemopoietic stem cells (HSCs). AML is highly heterogeneous in its
aetiology, pathogenesis, genetic background, clinical presentation, and
outcomes [[40]1]. Although many significant breakthroughs in AML
immunotherapy and targeted therapy were achieved recently, such as
BCL-2 inhibitors, FLT-3 inhibitors, and IDH1/2 inhibitors,
postremission relapses occur frequently, and we do not yet have a good
understanding of the aetiologies and pathogenesis of AML, which leads
to poor prognosis for AML patients [[41]2,[42]3]. Improving the
efficacy of AML treatment, especially for patients with relapsed and
refractory disease, is still a great challenge. Therefore, further
studies on promising prognostic biomarkers and potential therapeutic
targets are urgently needed to achieve individualized precision
medicine.
Fatty acids, as a crucial part of lipid metabolism, can accumulate to
meet the needs of signalling molecules and membranes for lipid
synthesis. Growing evidence indicates that fatty acid metabolism is
significantly associated with the occurrence and development of
multiple cancer types, including AML [[43]4,[44]5,[45]6]. For instance,
fatty acid oxidation (FAO) activation was shown to play an essential
role in promoting AML cell survival by bone marrow adipocyte
remodelling and lipolysis [[46]7]. In addition, upregulating FAO could
induce resistance to venetoclax with azacitidine (ven/aza) due to RAS
pathway mutations or compensatory adaptations in relapsed disease, so
the inhibition of FAO might provide new insights into strategies for
ven/aza resistance in AML [[47]8]. Moreover, dysregulation of fatty
acids was also demonstrated to affect not only the efficacy of
chemotherapy and radiotherapy for cancer patients, but also
immunotherapy [[48]9,[49]10]. Moreover, very long-chain fatty acid
metabolism also demonstrated a significant connection with the survival
of AML cells. For example, very long-chain acyl-CoA dehydrogenase
(VLCAD) may cause a decrease in AML cell survival and proliferation by
overexpressing and inhibiting fatty FAO in AML cells [[50]11]. In
general, the analysis of fatty acid metabolic pathways in AML will
contribute to understanding the molecular mechanism of AML and further
exploration of novel therapeutic treatments. However, the FAM-related
gene set is not thoroughly investigated in AML.
In this study, we developed a reliable and sensitive prognostic risk
signature based on 11 FAM-related genes, and thus constructed a
nomogram to predict 1-, 3-, and 4-year survival rates. We studied the
potential correlation of the risk score with several essential clinical
parameters, immune infiltration, and drug sensitivity. The results
highlight the essential role of FAM in the AML and provide a novel
perspective to understanding metabolic mechanisms.
2. Materials and Methods
2.1. Data Collection
The gene expression data and corresponding clinical data for 150 AML
samples and 337 normal whole-blood samples were retrieved from The
Cancer Genome Atlas (TCGA) database and Genotype-Tissue Expression
(GTEx) project database, respectively. Study participants with
incomplete clinical information were excluded. The transcriptomic and
clinical data of 140 patients with AML from the [51]GSE37642 data
cohort, which was based on the [52]GPL570 platform, were acquired from
the Gene Expression Omnibus (GEO) database and selected as a test set
to validate our prognostic signature. Furthermore, a total of 254 fatty
acid metabolism-related genes were downloaded from the Molecular
Signatures Database (MSigDB) version 7.4 for the next study.
2.2. Identification and Enrichment Analysis of the Differentially Expressed
Genes (DEGs)
First, Venn analysis was applied to screen overlapping genes between
all RNA-seq data (TCGA and GTEx) and FAM-related genes. Then, we
selected differentially expressed genes (DEGs) related to FAM from
normal and AML samples with the “Limma” R package. A false positive
discovery rate (FDR) < 0.05 and|fold change > 0.3|were set as the
thresholds. The “Pheatmap” package was applied to make a heatmap t of
the DEGs. In addition, gene ontology (GO) analysis and the Kyoto
Encyclopedia of Genes and Genomes (KEGG) analysis were conducted to
systematically explore the major biological features and cell
functional pathways of DEGs using the “org.Hs.e.g.,.db” package and
“clusterProfiler” package. Differences with a p value < 0.05 were
considered statistically significant.
2.3. Construction and Validation of the Prognostic Signature Associated with
FAM
We used data from the TCGA database and the GEO sample [53]GSE37642 as
the training and validation cohort, respectively. The sequencing data
of the FAM-related genes of each sample were integrated with
corresponding survival data. A univariate Cox regression analysis was
conducted to identify key genes associated with the prognosis of AML in
the TCGA cohort. Genes with a p value < 0.05 were selected for least
absolute shrinkage and selection operator (LASSO) analysis to identify
optimal prognostic genes. Based on the expression levels and risk
coefficients of first-rank prognostic genes, a risk signature was
developed, and the risk score for each patient was calculated according
to the equation below:
[MATH:
Risk score=∑i=1<
/mn>nβi<
mi>xi :MATH]
where n,
[MATH:
βi
:MATH]
and
[MATH:
xi
:MATH]
represent the number of signature genes, the value of the correlation
coefficients in the LASSO analysis for the genes, and the expression of
the gene, respectively. Based on the median risk score, AML patients
were separated into high-risk and low-risk categories. Kaplan-Meier
analysis and ROC curve analysis were carried out to evaluate the
predictive value of the risk signature. We also performed principal
component analysis (PCA) to estimate the clustering capacity of the
risk signature. In addition, the prognostic performance of the
signature was validated in the test set.
2.4. Comprehensive Analysis of the Prognostic Risk Score and
Clinicopathological Parameters of AML Patients
After excluding missing data in the training group, we combined the
risk score of each AML sample with the corresponding
clinicopathological characteristics. The connection between the risk
score and several clinical characteristics, including FAB subtype,
gender, age, cytogenetic, and molecular risk, was investigated by the
“limma” R package. Univariate and multivariable Cox regression analyses
were performed to determine the independent prognostic factors among
the above variables. A p < 0.05 was set as the selection criterion.
2.5. Development and Assessment of the Nomogram for AML Patients
A novel prognostic nomogram involving all independent prognostic
indicators was constructed by the “nomogram” package in R software to
further explore individual prognosis. Additionally, calibration, ROC
curve and Cox regression analyses were performed to assess the
predictive power of the nomogram from different angles.
2.6. Gene Set Enrichment Analysis
A total of 216 DEGs were identified between the high-risk and low-risk
groups with the screening criteria of FDR < 0.05 and|fold change > 1|.
Enrichment analysis of the DEGs was performed with the “Metascape”
website. GSEA software was also used to investigate the tumour
hallmarks and KEGG pathways in the different subtypes. The pathways and
hallmarks with p < 0.05 and FDR < 0.25 were regarded as statistically
significant.
2.7. Immune Infiltration Level Analysis
The ESTIMATE algorithm was applied to estimate the composition of
infiltrating stromal and immune cells between different groups.
According to the ESTIMATE algorithm, we calculated the immune score and
stromal score, which are positively related to the ratio of the
corresponding component in the tumour microenvironment (TME), and the
ESTIMATE score, which is the integration of the 2 former scores. Then,
we quantified the difference in immune cell infiltration and immune
function between different groups based on the CIBERSORT algorithm. The
associations between the risk score and immune checkpoints were also
investigated, and the results are shown in box plots.
2.8. Drug Sensitivity Analysis
To explore the potential role of the risk model in the treatment for
AML patients, we calculated the IC50 of the antineoplastic drugs using
the “pRRophetic” R package [[54]12]. IC50 was an important predictor
for evaluating drug response to treatment. All statistical analyses
were presented by the R package “ggplot2.”
2.9. Protein-Protein Interaction (PPI) Network
The DEGs between the two groups were further explored using STRING
online databases, and thus, PPI network data were constructed on the
basis of an interaction score > 0.70 (median confidence). We then
processed and showed the PPI network data through Cytoscape software,
which was used to seek hub genes from all DEGs as a Cytoscape plugin.
All samples were divided into low- and high-expression groups according
to the median expression value of the hub gene. The expression of the
CD163 gene was explored using gene expression profiling interactive
analysis (GEPIA) “[55]http://gepia.cancer-pku.cn/index.html (accessed
on 7 February 2023), which is an analysis tool that contains RNA
sequencing expression data of 9736 tumors and 8587 normal samples from
TCGA and the GTEx projects. Furthermore, the relationship of CD163 with
the prognosis, clinical features, and immune status were further
explored.
2.10. RNA Extraction and Real-Time Quantitative PCR (RT-qPCR)
Bone marrow samples for 10 newly diagnosed AML patients and 10 healthy
volunteers were collected from the Qilu Hospital of Shandong University
in Jinan, China. We used Trizol to extract the total RNA according to
specific protocols. Subsequently, the extracted RNA was reversely
transcribed into cDNA for qPCR using the Evo M-MLV RT Mix Kit. The 2 ×
SYBR Green qPCR master mix was applied in RT-qPCR on the Light Cycler
480 II. Relative expressions of the genes were normalized to GAPDH and
calculated using the 2^−δδct method. The primers used in this study are
shown in [56]Table S1.
3. Results
The detailed flow chart of the study is displayed in [57]Figure 1.
Figure 1.
[58]Figure 1
[59]Open in a new tab
Flow chart of this study.
3.1. Enrichment Analysis of AML Patient Samples
As shown in [60]Figure 2A, a total of 160 FAM-related genes were
identified. Subsequently, 78 genes were further selected for further
study by comparing the expression of FAM-related genes in normal and
AML samples. The top 10 upregulated and downregulated genes were
visualized on a heatmap ([61]Figure 2B). Enrichment analysis was
conducted for these FAM-associated DEGs. Fatty acid metabolism,
fatty−acyl−CoA metabolism, and fatty acid derivative metabolism were
highlighted in the GO analysis ([62]Figure 2C). The fatty acid
metabolism, degradation, biosynthesis, and elongation pathways were
significantly clustered in KEGG analysis ([63]Figure 2D). The results
indicated that fatty acid metabolism might exert a crucial function in
the pathogenesis of AML.
Figure 2.
[64]Figure 2
[65]Open in a new tab
Comparison between AML and normal samples in the TCGA and GTEx
databases. (A) Venn diagram of FAM-related genes in this study. (B)
Heatmap of the top 10 upregulated and downregulated FAM-related DEGs in
AML and normal samples. (C,D) GO and KEGG pathway analyses of DEGs
related to fatty acid metabolism in AML and normal samples.
3.2. Construction and Validation of the Risk Signature
Twenty-four FAM-related genes significantly correlated with the overall
survival (OS) of AML patients were identified using universal Cox
regression analysis in the TCGA set ([66]Figure 3A). Then, LASSO
analysis was applied to further identify the optimal prognostic genes.
Finally, 11 FAM-related genes ([67]Table 1) were identified and
constituted a prognostic risk model ([68]Figure 3B,C). The risk score
of each patient was calculated using the equation: risk score = (0.614
× the expression of the CBR1) + (0.092 × the expression of MAOA) +
(0.219 × the expression of ENO3) + (−0.243 × the expression of OSTC) +
(0.134 × the expression of UROD) + (−0.123 × the expression of PCTP) +
(0.104 × the expression of MAPKAPK2) + (0.257 × the expression of
PLA2G4A) + (0.133 × the expression of EPHX2) + (−0.364 × the expression
of ACSL6) + (0.348 × the expression of IDI1). The median risk score was
applied as a cut-off to categorize AML samples into a high-risk group
(n = 69) and a low-risk group (n = 70). PCA was then utilized to
compare the gene expression levels between high- and low-risk patients
based on FAM-associated genes and the 11 genes of the prognostic
signature ([69]Figure 3D,E). The outcomes revealed that the risk
signature had the best discriminatory ability between the two different
groups.
Figure 3.
[70]Figure 3
[71]Open in a new tab
Construction of the prognostic risk model. (A) Forest plot of 24 fatty
acid metabolism-related genes associated with prognosis identified
using univariate Cox regression analysis. (B,C) LASSO Cox regression
was performed to identify fatty acid metabolism-related genes closely
associated with the prognosis of AML. (D) Principal component analysis
based on all fatty acid metabolism-related genes in AML. (E) Principal
component analysis based on the fatty acid metabolism risk score to
distinguish tumours from normal samples. The group marked in blue
represents low-risk patients, and the group marked in red represents
high-risk patients.
Table 1.
Eleven fatty acid metabolism-related genes detected using LASSO
regression analysis.
Gene Coefficient
CBR1 0.614533407779894
MAOA 0.0923271597975086
ENO3 0.21850701264524
OSTC −0.243479390920753
UROD 0.134439432274683
PCTP −0.123309988656909
MAPKAPK2 0.104413688378819
PLA2G4A 0.25685461709233
EPHX2 0.133099892317426
ACSL6 −0.364168757158391
IDI1 0.347710847872942
[72]Open in a new tab
The differences in prognosis among AML patients were explored between
the different risk groups. The scatter plot displayed that mortality
increased with the risk score ([73]Figure 4A). The Kaplan–Meier
survival analysis also demonstrated that high risk correlated closely
with worse outcome in patients with AML (p < 0.001, [74]Figure 4B).
Then, we plotted ROC curves of the risk signature to measure predictive
sensitivity and specificity, and the areas under the curve (AUCs) for
predicting survival at 1, 3, and 5 years were 0.848, 0.836, and 0.856,
respectively ([75]Figure 4C). The expression of the 11 prognostic
FAM-related genes in the low- and the high-risk group was clearly
exhibited in a heatmap diagram ([76]Figure 4D).
Figure 4.
[77]Figure 4
[78]Open in a new tab
The correlation between risk score and survival status in the training
cohort. (A) Distribution of the risk scores (top) and survival status
of each patient (bottom). (B) Kaplan–Meier curve of the signature. (C)
ROC curves for predicting survival at 1 year, 3 years, and 5 years. (D)
Heatmap depicting the differential expression of the 11 genes in AML
patients.
To validate the predictive capacity of the risk signature, we used the
GEO cohort, which included 140 AML patients, as the test set.
[79]GSE37642 was clustered into low- and high-risk groups based on the
cut-off value determined in the training set. Consistent with the
outcome in the training set, the survival analysis indicated that the
risk score had an inverse relationship with the clinical outcomes of
AML patients ([80]Figure 5A,B). The AUCs were 0.717, 0.668, and 0.69
for predicting survival at 1, 3, and 5 years, respectively ([81]Figure
5C). We further screened four FAM-related genes from signatures and
examined the expressions of these genes by qPCR in 10 AML and 10
healthy control samples. As shown in [82]Figure 5D, the expression
levels of UROD, PCTP, and EPHX2 were relatively higher in controls than
that in AML, while PLA2G4A were higher in AML. In addition, compared to
other known prognostic signatures in AML, such as AJH 2021 [[83]13],
Leu 2020 [[84]14], and JHO 2016 [[85]15], the FAM-related risk
signature had a significantly better capability to predict prognosis
([86]Figure 5E–G).
Figure 5.
[87]Figure 5
[88]Open in a new tab
Validation of the risk signature. (A,B) Risk score distribution and
survival status, Kaplan–Meier curve in the GEO cohort. (C) ROC curves
for predicting survival at 1 year, 3 years, and 5 years in the GEO
cohort. (D) Validation of the expression of UROD, PCTP, PLA2G4A, and
EPHX2 in AML samples by qPCR. (E–G) AUC comparison between the risk
signature based on 11 FAM-related genes and other previously published
prognostic signatures. * p < 0.05 and *** p < 0.001.
Briefly, these results demonstrated that the risk model constructed
based on the FAMs prognostic signature could accurately evaluate the
prognosis of patients with AML.
3.3. Correlation Analysis between Risk Score and Clinicopathological Features
Based on the accuracy of the survival risk predictions, we further
investigated the role of the risk signature in predicting AML
progression, and the associations between the risk score and
clinicopathological characteristics were explored. Significant
differences were shown between different groups in age, cytogenetic
risk, FAB subtype, and molecular risk (all p < 0.001) ([89]Figure
S1A–D). However, no correlation was found between the risk score and
gender (p > 0.05; [90]Figure S1E).
3.4. Construction of a Nomogram for AML Patients
To identify the independent predictors of survival in AML patients,
univariate and multivariate Cox regression analyses were performed. The
results indicate that risk score and age were the only two independent
prognostic factors for predicting the OS of AML patients (p < 0.001;
[91]Figure 6A,B). Subsequently, a prognostic nomogram including
independent risk factors (age and risk score) was constructed to
predict the survival probabilities of AML patients at 1, 3, and 4 years
([92]Figure 6C). The AUCs showed that the nomogram (AUC = 0.864) had
higher sensitivity and specificity than other single prognostic
factors, such as age (AUC = 0.790), cytogenetic risk (AUC = 0.645),
molecular risk (AUC = 0.680), and risk score (AUC = 0.852) ([93]Figure
6D). In addition, the calibration curves demonstrated good prediction
performance, and our model was similar to the ideal model in estimating
1-, 3-, and 4-year OS ([94]Figure 6E). Moreover, Cox regression
analyses were conducted to confirm that the nomogram score was an
independent predictive factor for AML prognosis in all participants
([95]Figure 6F,G). Collectively, we validated the prominently
predictive ability of the tumour prognostic signature and revealed a
high potential for clinical utility from multiple perspectives.
Figure 6.
[96]Figure 6
[97]Open in a new tab
Development and assessment of the nomogram for patients with AML. (A,B)
Univariate and multivariate Cox regression analyses of clinical
parameters in patients with AML. (C) Nomogram to predict the 1-, 3-,
and 4-year OS of AML patients. (D) ROC curves of the nomogram, risk
score, and clinical characteristics in predicting prognosis. (E)
Calibration plot analysis to evaluate the predictive ability of the
nomogram. The x-axis is nomogram-predicted survival, and the y-axis is
actual survival. (F,G) Univariate and multivariate Cox regression
analyses were conducted to identify if the nomogram score was an
independent predictor in AML patients. Green squares: Hazard ratio (HR)
< 1; red squares: HR > 1.
3.5. Functional and Annotation Analyses
To distinguish the biological functions and networks related to the
risk signature, we screened 216 DEGs in the low- and high-risk group
for enrichment analyses. Metascape analysis results establish that some
immune-associated pathways, including cytokine signalling in the immune
system, regulation of interleukin-12 production, regulation of
macrophage-derived foam cell differentiation, and several pathways
associated with the development of malignant tumours, such as positive
regulation of phosphatidylinositol 3-kinase signalling and
proteoglycans [[98]16,[99]17] in cancer, were significantly clustered
([100]Figure 7A). Moreover, GSEA was also conducted for functional
enrichment analysis. The results using the Hallmark database are shown
in [101]Figure 7B. Most metabolism-related pathways, such as fatty acid
metabolism and adipogenesis, and several immune pathways, such as TNF-α
signalling via NF-κB pathways and IL6-JAK-STAT3 signalling, were
significantly enriched in the high-risk group. Additionally, the
results using the KEGG database demonstrate that fatty acid substance
metabolism pathways, such as arachidonic acid and butanoate acid
metabolism, glycerolipid metabolism, propanoate metabolism, pyruvate
metabolism, and associated fatty acid metabolism pathways, such as
citrate cycle tac cycle signalling pathways (TCA cycle) and PPAR
signalling pathways, were greatly clustered in the high-risk group
([102]Figure 7C). All these results validated that FAM-related
signatures are closely connected to the immune response, which is
crucial for AML.
Figure 7.
[103]Figure 7
[104]Open in a new tab
Function and pathway enrichment analysis. (A) Function and pathway
enrichment analysis by Metascape. The image shows the histogram of the
top 20 enriched pathways. (B,C) GSEA results.
3.6. The Landscape of the Tumour Microenvironment (TME) and Immune Cell
Infiltration in AML Patients
According to the enrichment analysis, pathways related to fatty acid
metabolism and immunity were significantly highlighted in the high-risk
group, suggesting that exploration of the correlation between immune
status and risk score was essential. The result from CIBERSORT
algorithm revealed that monocytes and macrophage M2 cells infiltrated
at a greater rate in the high-risk group, while resting mast cells were
markedly activated in the low-risk group (p < 0.05; [105]Figure 8A).
Higher immune, stromal, and ESTIMATE scores were found in the patients
with the high risk using the ESTIMATE algorithm (p < 0.05, [106]Figure
8B,D,F), representing a relatively high infiltration of immune and
stromal cells in the high-risk TME, which was associated with worse
prognosis [[107]18,[108]19]. Regarding immune-associated functions, the
type I IFN and II response, APC co-inhibition and stimulation,
checkpoint, chemokine receptors (CCRs), inflammation promotion, HLA, T
cell co-stimulation, and parainflammation were better activated in the
high-risk group (p < 0.05, [109]Figure 8C). Moreover, there were
drastic differences in immune checkpoint expression. The expression
levels of immune checkpoints, such as CTLA4 and PDCD1, were positively
correlated with the risk score (p < 0.05, [110]Figure 8E).
Figure 8.
[111]Figure 8
[112]Open in a new tab
Comparison of the tumour microenvironment (TME) and immune cell
infiltration between the two risk groups. (A) Box plot of the fraction
of 22 immune cells in the high- and low-risk group. (B,D,F) Immune
score, stromal score, and ESTIMATE score between different risk groups.
(C) Box plot of the scores of 13 immune-related functions in the high-
and low-risk group. (E) Differences in the expression of 15 checkpoints
in the two risk groups. * p < 0.05; ** p < 0.01; and *** p < 0.001.
3.7. Analysis of Drug Sensitivity in the Two Risk Groups
Chemotherapy and targeted drug therapy are considered as vital
strategies in the clinical management of AML patients. Therefore, it is
necessary to explore differences in drug sensitivity between the
different risk groups. In our study, the traditional cytotoxic drugs
were more sensitive in the low-risk group and included histone
deacetylase (HDAC) inhibitors, such as vorinostat and parthenolide
([113]Figure 9A,B), the BCL-2 inhibitor navitoclax ([114]Figure 9C),
and midostaurin ([115]Figure 9D), which were demonstrated to
consolidate chemotherapy and encourage efficacy in AML patients with a
FLT3 mutation [[116]20]. GNF-2, as an allosteric inhibitor of BCR-ABL,
also showed higher sensitivity in the low-risk group ([117]Figure 9E).
Moreover, a more sensitive response to cytarabine (Ara-C) was seen in
the low-risk group than in the high-risk group ([118]Figure 9F). Ara-C
is a first-line agent with excellent activity in AML. However, the
PI3K/mTOR inhibitors BEZ235 and AZD8055 were less sensitive in the
low-risk group ([119]Figure 9G,H). According to above results, the two
risk groups responded dramatically different to chemotherapy.
Figure 9.
[120]Figure 9
[121]Open in a new tab
Drug sensitivity predictions in different risk groups. (A–H)
Vorinostat, parthenolide, navitoclax, midostaurin, GNF-2, cytarabine
(Ara-C), BEZ235, and AZD8055.
3.8. PPI Analysis
To further investigate the discrepancies between the low- and high-risk
group, the DEGs of the two risk groups were introduced into the online
database STRING to analyze their expression profiles. Relevant PPIs
were obtained and are visualized in [122]Figure S2A. As shown in
[123]Figure 10A, the interacting genes were processed using Cytoscape
software (version 3.9.0), which is an open source software platform for
visualizing complex networks and integrating these with any type of
attribute data. Cytoscape and the plug-in app “cytoHubba” were applied
to parse the network. The ten highest-scored genes (including CD163,
FN1, FCGR3A, ITGB3, SERPINE1, THBS1, ITGA2B, CD14, MMP2, and CCR1) in
the network were identified as the hub genes on the basis of the “MCC”
algorithm, which are shown in [124]Figure 10B. Subsequently, we
analyzed CD163 expression level in AML patients by GEPIA dataset.
Result show CD163 mRNA expression was significantly higher in AML
tissues compared to that in the corresponding healthy bone marrow
samples ([125]Figure 10C). Kaplan-Meier analysis was also performed to
determine whether hub gene expression correlated with AML survival.
Except for FN1 and FCGR3A, the other eight hub genes were significantly
related to the prognoses of AML patients (p < 0.05, [126]Figure S2B–J;
[127]Figure 10D). Among all hub genes, CD163 was observed to be the
most relevant prognostic factor. In addition, the selection of the core
genes in the PPI network also indicated that CD163 was centrally
located. These results suggest that CD163 may serve as a novel
therapeutic target in AML. Therefore, we divided all AML samples into
two groups according to the CD163 mRNA median expression value. The
correlation study between CD163 expression and AML clinical status
revealed that age, cytogenetic risk, molecular risk, and FAB subtype
were closely related to CD163 expression ([128]Figure 10E–H),
suggesting an inverse correlation between CD163 expression and the
clinical status and prognosis of AML. In addition, we also conducted an
investigation to explore the specific differences in immune cell
infiltration between groups with high and low CD163 expression. The
results show that NK, B cell, and CD8+ T cell infiltration was
significantly increased in tumours with low CD163 expression
([129]Figure 10I).
Figure 10.
[130]Figure 10
[131]Open in a new tab
Protein-protein interaction (PPI) network. (A) PPI network processed by
Cytoscape. Upregulated genes in the high-risk score group are marked in
red, and those in the low-risk score group are presented in green. (B)
Top 10 hub genes selected by cytoHubba. (C) The expression of CD163
genes in AML patients and paired normal bone marrow samples was
analyzed by GEPIA. (D) Survival analysis for the subgroup classified by
CD163 mRNA expression. (E–H) The CD163 mRNA expression levels differed
among different groups based on clinical features, including age,
molecular risk, FAB subtype, and cytogenetic risk. (I) Immune cell
infiltration in patients with different expression levels of CD163. The
text continues here. * p < 0.05; ** p < 0.01; and *** p < 0.001.
4. Discussion
AML is an extremely heterogeneous disease due to its complicated
genetic and molecular mechanisms. Despite excellent breakthroughs in
targeted therapy, the complete remission (CR) rates and long-term
survival of AML patients are still unsatisfactory [[132]21]. Therefore,
the identification of effective prognostic biomarkers is urgently
needed. FAM is involved in membrane synthesis, energy generation, and
signal transduction in AML, including tumorigenesis and the progression
of cancer [[133]22]. Although recent literature demonstrated that
multiple FAM-related genes could greatly improve prognostic prediction
compared with a single biomarker in AML [[134]23], we further
investigated the correlation between FAM-related genes and the OS of
AML patients, and thus selected 11 new genes to construct a risk model
for prognostic prediction and therapeutic decision guidance. Moreover,
we provided a hybrid nomogram model combining clinical factors with an
effective quantitative scoring method that could predict the OS of AML
patients.
Among these 11 new genes in the prognostic signature, carbonyl
reductase 1 (CBR1) was illustrated to have an essential role in the
metabolism of multiple drugs, such as daunorubicin, doxorubicin,
anthracycline, and haloperidol [[135]24,[136]25]. Daunorubicin, as a
great choice for patients with haematologic malignancies, including
AML, was demonstrated to be affected by CBR1 [[137]26]. Phospholipase
A2-IVA (PLA2G4A) not only plays a vital role in the development of
various solid tumours [[138]27,[139]28], but was also illustrated to
independently predict OS in patients with non-M3/NPM1 WT AML [[140]29].
ACSL6, an isoform of the acyl-CoA synthetase long-chain family (ACSL),
can promote leukaemogenesis by ASL6-ETV6 fusion gene formation
[[141]30]. MAPKAPK2 [[142]31], MAOA [[143]32], and EPHX2 [[144]33] were
demonstrated to affect tumour development in various contexts. ENO3,
OSTC, and UROD serve as diagnostic or prognostic biomarkers in other
cancers [[145]34,[146]35,[147]36]. To date, the function of PCTP and
IDI1 in the progression of cancers is still unclear, and further
experimental work is needed. Generally, the above evidence suggests
that these 11 newly identified genes have the potential to become novel
markers for the prediction of prognosis and to become targets for
molecular targeted therapy in patients with AML.
Functional and annotation analyses indicated that FAM-associated
pathways and immune-related hallmarks were markedly highlighted in the
high-risk group, suggesting that the level of fatty acids may affect
the function and phenotype of infiltrating immune cells in the
microenvironment [[148]37]. Immune regulation plays an essential role
in the initiation and development of AML. The number and proportion of
infiltrating immune cells can impact cancer progression, prognosis
prediction, and immunotherapy response [[149]38]. According to the
immune cell infiltration results, monocyte and macrophage M2 cells were
highly expressed in the high-risk group, which are signs of immune
evasion. The polarization of macrophages towards the M2 phenotype is
favoured for the development of tumour cells, angiogenesis, and
immunosuppression. In addition, a higher number of M2-like macrophages
in the TME is associated with poorer outcomes in several malignant
tumours, including AML [[150]39,[151]40]. This evidence indicates that
the poor prognosis of high-risk patients is positively associated with
immunosuppression in the TME, and these differences would favour tumour
progression and immunotherapy response.
In view of the significant discrepancies observed between the low-risk
and high-risk groups, further investigations of the different genes in
each group were needed. It was discovered that CD163 is crucial. As a
haemoglobin scavenger receptor, the overexpression of CD163 was
observed to be significantly correlated with tumour progression and
poor prognosis in multiple human cancers [[152]41], which is consistent
with our results. In addition, CD163 was identified as a vital
biomarker of M2 macrophage activation and may predict the invasion and
prognosis of malignancies [[153]42,[154]43]. Therefore, CD163 might
provide a potential therapeutic target for the treatment of
haematologic malignancies.
Overall, this study meticulously constructed and verified a novel
prognostic signature based on FAM-associated genes for patients with
AML. A comprehensive investigation of the associations of signalling
pathways, drug sensitivity, and immune infiltration with the prognostic
risk model for leukaemia was conducted, and these findings may provide
a novel basis for more precise targeted therapies in AML. However, a
few shortcomings and drawbacks of our study should be taken into
consideration. The results based on bioinformatics in this paper were
not verified by in vitro experiments. In the future, we will
experimentally verify the significance of fatty acid metabolism in AML.
5. Conclusions
A novel fatty acid metabolism-associated risk model was successfully
constructed and demonstrated a reliable prognostic ability for
chemotherapy sensitivity and response to future antitumour
immunotherapy. In addition, we generated a nomogram to accurately
predict the 1-, 3-, and 4-year survival rates using only several
variables from AML patients.
Supplementary Materials
The following supporting information can be downloaded at:
[155]https://www.mdpi.com/article/10.3390/curroncol30020193/s1, Figure
S1: Correlation between the risk score and clinicopathological
characteristics in the TCGA cohort. Figure S2: Construction of the PPI
network and prognostic value of the top 10 hub genes. Table S1: The
primer sequences of four FAM-related genes.
[156]Click here for additional data file.^ (708.6KB, zip)
Author Contributions
Writing—original draft, N.W.; methodology, X.B.; data curation, X.W.;
conceptualization, G.M., F.Z. and J.Y.; supervision, D.W. F.L. and C.J.
All authors have read and agreed to the published version of the
manuscript.
Institutional Review Board Statement
The studies involving human participants were reviewed and approved by
the Ethics Committee of Shandong University School of Medicine
(SDULCLL2020-1-14). The patients/participants provided their written
informed consent to participate in this study. The patients involved in
the database TCGA, GTEx and GEO database, which belong to public
databases, have obtained ethical approval. Users can download relevant
data for free for research and publish relevant articles.
Informed Consent Statement
Not applicable.
Data Availability Statement
The data used to support the findings of this study are available from
the corresponding authors upon request.
Conflicts of Interest
The authors declare that they have no conflict of interest regarding
the publication of this study.
Funding Statement
This work was supported by grants from the Distinguished Taishan
Scholars in Climbing Plan (tspd20210321), the National Natural Science
Foundation of China (82070160, 82000165, 82170182), the Major Research
Plan of the National Natural Science Foundation of China (91942306),
the 68th China postdoctoral Science Foundation (2020M682171), the Key
Program of Natural Science Foundation of Shandong Province
(ZR2020KH016, ZR2021MH302), the Fundamental Research Funds for the
Central Universities (2022JC012), the Independently Cultivate
Innovative Teams of Jinan, Shandong Province (2021GXRC050) and the
Clinical Practical New Technology and Development Fund of Qilu
Hospital, Shandong University (2019-5).
Footnotes
Disclaimer/Publisher’s Note: The statements, opinions and data
contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI
and/or the editor(s) disclaim responsibility for any injury to people
or property resulting from any ideas, methods, instructions or products
referred to in the content.
References