www.nature.

com/cgt Cancer Gene Therapy

ARTICLE OPEN

NCBP2 and TFRC are novel prognostic biomarkers in oral

squamous cell carcinoma
Rahul Arora 1,7, Logan Haynes1,7, Mehul Kumar1, Reid McNeil1, Jahanshah Ashkani 2, Steven C. Nakoneshny 3,
T. Wayne Matthews3,4, Shamir Chandarana3,4, Robert D. Hart3,4, Steven J. M. Jones 2, Joseph C. Dort3,4,5, Doha Itani6,
✉
Ayan Chanda 1,3 and Pinaki Bose 1,3,5

© The Author(s) 2023

There are few prognostic biomarkers and targeted therapeutics currently in use for the clinical management of oral squamous cell
carcinoma (OSCC) and patient outcomes remain poor in this disease. A majority of mutations in OSCC are loss-of-function events in
tumour suppressor genes that are refractory to conventional modes of targeting. Interestingly, the chromosomal segment 3q22-
3q29 is amplified in many epithelial cancers, including OSCC. We hypothesized that some of the 468 genes located on 3q22-3q29
might be drivers of oral carcinogenesis and could be exploited as potential prognostic biomarkers and therapeutic targets. Our
integrative analysis of copy number variation (CNV), gene expression and clinical data from The Cancer Genome Atlas (TCGA),
identified two candidate genes: NCBP2, TFRC, whose expression positively correlates with worse overall survival (OS) in HPV-
1234567890();,:

negative OSCC patients. Expression of NCBP2 and TFRC is significantly higher in tumour cells compared to most normal human
tissues. High NCBP2 and TFRC protein abundance is associated with worse overall, disease-specific survival, and progression-free
interval in an in-house cohort of HPV-negative OSCC patients. Finally, due to a lack of evidence for the role of NCBP2 in
carcinogenesis, we tested if modulating NCBP2 levels in human OSCC cell lines affected their carcinogenic behaviour. We found
that NCBP2 depletion reduced OSCC cell proliferation, migration, and invasion. Differential expression analysis revealed the
upregulation of several tumour-promoting genes in patients with high NCBP2 expression. We thus propose both NCBP2 and TFRC
as novel prognostic and potentially therapeutic biomarkers for HPV-negative OSCC.

Cancer Gene Therapy (2023) 30:752–765; https://doi.org/10.1038/s41417-022-00578-8

INTRODUCTION [5, 6]. Furthermore, OSCC prognostication and treatment selection

Head and neck cancers are the sixth most common cancer still relies heavily on the tumour-node-metastasis (TNM) staging
worldwide, resulting in more than 700,000 newly diagnosed cases system, where tumour biology has limited impact on treatment
in 2020 and over 400,000 deaths [1]. The vast majority of head and decisions. Hence, identifying improved targeted treatments and
neck cancers are squamous cell carcinomas and oral squamous prognostic markers are major priorities in OSCC.
cell carcinoma (OSCC) is the most common head and neck cancer Identification of novel oncogenic drivers may permit more
[1]. Smoking and/or alcohol consumption are canonical risk factors precise treatment selection and reduce treatment-related morbid-
for OSCC, increasing the chance of incidence by up to 30 times. ity in low-risk patients while improving the management of high-
However, the human papillomavirus (HPV) has recently emerged risk patients. Prior studies aiming to characterize the genomic
as a major causal agent [2, 3]. These two aetiologic subsets landscape in HNSCC have reported that loss of function mutations
represent distinct pathologies. However, HPV-positivity is mostly in several tumour suppressor genes—such as TP53, NOTCH1, and
limited to oropharyngeal cancers in the head and neck region and CDKN2A—drive carcinogenesis in a majority of cases [7]. However,
is uncommon in other anatomical subtypes, including OSCC [4]. tumour suppressor alterations are challenging drug targets
OSCC patients exhibit significant symptom burden; primary because it is difficult to restore gene function. A few mutated
treatment involves surgical resection, which is associated with oncogenes such as the EGFR and PIK3CA have been reported but
high morbidity [5, 6]. Most OSCC patients are also treated with targeting these alterations has yielded limited success in OSCC
post-operative radiotherapy (PORT) and/or chemotherapy. These [8–10].
treatment modalities have an adverse impact on quality-of-life. The overexpression of non-mutated genes has been associated
There have been limited advances in OSCC management, with with tumour progression. Genes can be overexpressed as a result
overall survival (OS) improving by only 5% in the last 20 years of the amplification of genomic loci, loss of expression of negative

1
Department of Biochemistry and Molecular Biology, Cumming School of Medicine, University of Calgary, Calgary, Canada. 2Canada’s Michael Smith Genome Sciences Centre,
Vancouver, BC, Canada. 3Ohlson Research Initiative, Arnie Charbonneau Cancer Institute, Cumming School of Medicine, University of Calgary, Calgary, Canada. 4Department of
Surgery, Section of Otolaryngology-Head & Neck Surgery, Cumming School of Medicine, University of Calgary, Calgary, Canada. 5Department of Oncology, Cumming School of
Medicine, University of Calgary, Calgary T2N 4N1 AB, Canada. 6Department of Anatomic and Molecular Pathology, Dalhousie University, Saint John, NB, Canada. 7These authors
contributed equally: Rahul Arora, Logan Haynes. ✉email: [email protected]

Received: 13 July 2022 Revised: 1 December 2022 Accepted: 9 December 2022

Published online: 12 January 2023
 R. Arora et al.
753

Fig. 1 Identification of NCBP2 and TFRC as potential prognostic biomarkers in OSCC. A Filtering scheme to identify 3q22-3q29 genes of
prognostic significance in HPV-negative OSCC. DEA differential expression analysis. Scatter plots showing the expression of B TFRC (Spearman
ρ = 0.46, p < 1.5e−15) and C NCBP2 (Spearman ρ = 0.67, p < 2.2e−16) to be positively correlated to 3q22-3q29 CNV status in TCGA HPV-
negative OSCC patients.

regulators, or increased transcription due to aberrant enhancer genes, TFRC and NCBP2 that are located on the highly vulnerable
activity. Gene amplification is a frequent genomic alteration in telomeric region 3q29, were considered for downstream analyses
cancers whereby there is an increase in copy number of a sub- and validation. High TFRC and NCBP2 protein levels were associated
chromosomal region containing the gene of interest. Gene with worse prognosis in OSCC patient-derived tissue microarrays.
amplification can occur due to genomic instability and/or loss of Since TFRC expression has been previously linked to OSCC cell
cell cycle control, both of which are hallmarks of carcinogenesis growth and is considered a potential therapeutic target [18], we
[11–13]. Usually, segments of the genome that are amplified contain investigated the role of NCBP2 in regulating the proliferation,
many genes and only a select few of these might contribute to migration and invasion of OSCC cells in culture. Finally, we performed
carcinogenesis and progression [11–13]. Developing inhibitors for differential expression analysis between top and bottom-quartile
these targets, and companion diagnostics to identify patients NCBP2-expressing TCGA OSCC patients to identify potential tumour-
suitable for targeted therapy could improve prognosis and alleviate promoting factors that are upregulated downstream of NCBP2. Our
treatment-related morbidity of cancer patients. Interestingly, the studies have established NCBP2 as a bona fide prognostic marker
chromosomal cytoband 3q22-3q29 is frequently amplified in a wide and potential therapeutic target in OSCC.
range of squamous cell carcinomas, including OSCC [14–17].
Here, we analysed OSCC genomes from The Cancer Genome Atlas
(TCGA) to investigate the biological and clinical significance of METHODS
frequently amplified genes located on the cytobands 3q22-3q29. We Patient cohorts
devised a simple, yet effective filtering technique (Fig. 1A) to identify Demographic, survival, gene expression and copy number data for 282
genes of clinical relevance (i.e., prognosis) among the 468 genes HPV-negative OSCC patients and 26 normal oral squamous cell samples
located on the cytobands 3q22–3q29. Out of four potential hits, two were obtained from the TCGA data portal. Gene expression data were also

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
754
obtained for another 263 HPV-negative OSCC patients and 139 normal oral blocks with sufficient tumour content. Three 0.6 mm cores were randomly
squamous cell samples from five published publicly available datasets sampled for each patient from tumour-bearing areas of the FFPE tissue
[19–23] (Table S1). Gene expression data for normal tissue samples were blocks using a beecher manual tissue microarrayer (Beecher Instruments
downloaded from the Genotype-Tissue Expression (GTEx) project. Sample Inc., WI, USA) and arrayed on TMA blocks. Tissue represented ~70 patients
sizes were determined by data availability. (in triplicate) on each TMA block. Normal oral cavity squamous epithelium
The University of Calgary OSCC cohort consisted of 175 histologically tissue cores were also included. TMA slides were prepared using 4 μm-
confirmed, surgically resected, treatment-naïve patients diagnosed between thick sections from the TMA block.
2009 and 2013 (Table 1). The median age was 62.5 years and the median
follow-up for the cohort was 5.8 years. Treatment and outcome information is
prospectively updated and is current as of January 31, 2022. Patient material Immuno-histochemistry (IHC)
TMA slides were deparaffinized and rehydrated before performing heat-
and clinical data use to abide by guidelines discussed in the Tri-council Policy
induced epitope retrieval. Endogenous peroxidase activity was quenched
Statement for Research with Human Subjects (Canada). This study was
performed in accordance with reporting recommendations for tumour with a peroxidase block, and slides were blocked and permeabilized using
marker prognostic studies (REMARK) guidelines [24] and was approved by rodent block with 0.2% Triton X. Slides were incubated for one hour with
the Health Research Ethics Board of Alberta (HREBA). anti-NCBP2 (abcam ab91560) or anti-TFRC antibodies (abcam ab84036) for
one hour followed by an HRP-conjugated mouse anti-rabbit secondary
antibody (Dako EnVison+ kit, K4065), and finally for 2 min with 3,3′-
Determination of 3q22-3q29 amplification status diaminobenzidine tetrahydrochloride (DAB) to visualize bound antibodies.
Genomic Identification of Significant Targets in Cancer (GISTIC) level 2 data Slides were counterstained with hematoxylin, dehydrated, and mounted.
containing amplification statuses for notable amplifications of each TCGA- Multiple antibody concentrations for NCBP2 (1:500, 1:1000, 1:2000 and
OSCC sample was downloaded from the Broad Institute’s Genome Data 1:2500) and TFRC (1:100, 1:500, 1:1000, 1:1500, 1:2000, 1:2500 and 1:5000)
Analysis Center (GDAC) Firehose. Specifically, the file analysed was were used to optimize staining for both proteins in control mouse liver and
“all_lesions.conf_99.txt”. The 3q22-3q29 amplification status came from normal oral cavity squamous epithelium core samples (Fig. S1). The study
peak 3 of this file (3q26.33, region limits: chr3:131468786-198022430; pathologist (DI) assessed stained slides to select the optimal antibody
3q22-3q29). 282 samples were considered for differential expression concentration that would permit for scoring of differential protein expression
analyses between 3q22-3q29 amplified (n = 41 samples) and non- levels within TMA cores. After staining optimization, final primary antibody
amplified (n = 236 samples) TCGA HPV-negative OSCC samples. Amplifica- concentrations of 1:2000 for NCBP2 and 1:1500 TFRC were selected. TMAs
tion data was unavailable for 7 samples. Markers for this cytoband region were scored by the pathologist (DI) with a score of 0 indicating negative
were determined by filtering the “all_data_by_genes.txt” GISTIC2 output staining, and scores of 1, 2 or 3 indicating positive staining of increasing
file (downloaded from GDAC firehose) for 3q22-29 (n = 468 genes). intensity for each protein. Disagreements between cores from the same case
were resolved by taking the maximum score for that patient.
Differential expression analysis (DEA)
Level III mRNA-sequencing data (raw counts) was used to perform DEA using Statistical and survival analysis
the DESeq2 Bioconductor package [25]. DEA was performed between Univariate Cox proportional hazards regression was used to determine the
primary TCGA HPV-negative OSCC tumour versus normal tissue samples, and association between biomarkers and clinical covariates and OS, DSS, or PFI.
between 3q22-3q29 amplified versus non-amplified primary tumour samples. Covariates showing significant associations in univariate analysis were
Thresholds for differential expression were set at an absolute fold change adjusted with p-value correction or multivariate models depending on the
cut-off of 1.5 and false discovery rate (FDR) of 0.1% (adjusted p < 0.001). analysis. Kaplan–Meier survival curves were used to visualize differences in
Differentially expressed genes were filtered for genes located on the 3q22- survival between groups. For Kaplan–Meier curves with continuous
3q29 cytoband. Filtering steps are described in the flow chart shown in the variables, a cut-point determined by the method outlined by Contal and
appended Fig. 1A. Briefly, the genes overexpressed in 3q22-3q29 amplified O’Quigley [26] was utilized. The protein expression status of University of
tumour samples overlapped with the genes overexpressed on OSCC Calgary OSCC patients was dichotomized based on low (0/1) and high (2/3)
compared to matched normal samples. We also overlapped these commonly NCBP2 and TFRC protein levels for Kaplan–Meier curves.
overexpressed genes with those overexpressed in OSCC samples from five Demographic differences by gene and protein expression levels were
independent HPV-negative OSCC cohorts containing both OSCC and assessed using the Mann-Whitney U test for continuous variables and
matched normal oral cavity squamous epithelium. The final set of genes Pearson’s χ2 test for categorical variables. Spearman’s ρ was calculated for
was subjected to univariate cox proportional hazard analysis with overall the correlation between NCBP2 and TFRC mRNA and protein expression.
survival as an outcome and significant hits were further analysed. NCBP2 and TFRC mRNA expression was compared between tumour and
DEA was performed between top quartile NCBP2 and bottom quartile matched normal tissue using the Wilcoxon signed-rank test. Unpaired t-
NCBP2 expressing OSCC patients in the TCGA. Thresholds for differential test was performed for the in vitro assays. All statistical analyses were
expression were set at an absolute fold change cut-off of 1.5 and FDR of performed using GraphPad Prism (GraphPad, California) or R version 3.6.1.
5% (adjusted p < 0.05). An extensive literature review was performed to
evaluate the roles (if any) of the DE genes in regulating OSCC
tumorigenesis and progression. Immunoblotting analysis
Protein lysates were isolated from the UMSCC29 and CAL33 cell lines
treated with a control siRNA or a pool of siRNA targeting NCBP2 (NCBP2i),
TCGA OSCC promoter DNA methylation and mRNA gene using TNTE (50 mM Tris, 150 mM NaCl, and 1 mM EDTA) buffer containing
expression analysis 1% Triton X-100 along with protease and phosphatase inhibitors, and 0.1%
R studio (version 4.2.0) and R package TCGAbiolinks (version 2.24.3) were SDS. Cell extracts were collected in Eppendorf tubes and centrifuged at
used to analyse the TCGA HPV-negative OSCC samples. TCGAbiolinks was 14,000×g for 10 min at 4 °C. Equivalent protein amounts of lysates were
used to download DNA methylation beta values (Illumina Human Methyla- resolved by SDS–PAGE followed by transfer to nitrocellulose membranes.
tion 450 array) and transcriptome profiling mRNAseq counts for all samples. Specific proteins on membranes were incubated overnight with primary
DNA methylation beta values and mRNAseq counts were filtered for four antibodies targeting NCBP2 (abcam ab91560 rabbit anti-human NCBP2,
genes (NCBP2, TFRC, RFC4, and GMPS). mRNAseq raw counts were normalized 1:4000), and actin (Santa Cruz sc-47778, 1:2000) in 3% BSA. Membranes
to TPM values. CpG promoter probes for all four genes were identified and were then incubated with HRP-conjugated goat anti-mouse or donkey
filtered for by genomic base pair position. R function cor.test was used to anti-rabbit IgG (Bio-Rad Laboratories, 1:10,000 in 5% skim milk) as
calculate spearman correlation values between CpG promoter probes and secondary antibodies for 1 h at room temperature, followed by incubation
log2TPM (mRNA) counts. Firstly, individual CpG promoter probes were in enhanced chemiluminescence (ECL) (Millipore) reagent and light signal
correlated with log2TPM (mRNA) counts, ρ and p-values are reported. detection using a Chemidoc® Touch Imager (Bio-Rad Laboratories).
Secondly, scatter plots of mean promoter beta value versus log2TPM (mRNA)
counts are shown for the four genes. Cell proliferation assay
5 × 105 UMSCC29 and CAL33 cells were seeded in each well of a 12-well
Tissue microarray (TMA) construction tissue culture plate and treated either with control RNAi or NCBP2i for 24 h
Haematoxylin–eosin (H&E)-stained slides were reviewed by the study after which 5000 cells each were sub-cultured in triplicate in a 96-well
pathologist (DI) to select formalin-fixed paraffin-embedded (FFPE) tissue tissue culture plate overnight. BrDU reagent was added, and proliferation

Cancer Gene Therapy (2023) 30:752 – 765

 Table 1. Demographics of OSCC patients in Ohlson TMA cohort.
Variable n (%) unless otherwise indicated All patients NCBP2 score 0 or 1 NCBP2 score 2 or 3 p-value TFRC score 0 or 1 TFRC score 2 or 3 p-value
Number 175 49 126 – 76 99
Sex 0.554 0.389
Female 65 (37.1) 16 (32.7) 49 (38.9) 25 (32.9) 40 (40.4)
Male 110 (62.9) 33 (67.3) 77 (61.1) 51 (67.1) 59 (59.6)
Median age at diagnosis, years (IQR) 62.5(53.9–74.8) 62.5(51.9–76.7) 62.8(54.6–74.6) 0.774 62.8(53.2–74.0) 62.5(54.6–75.3) 0.682
Pathologic T stage 0.130 0.401
T1/T2 105 (60.0) 26 (53.1) 79 (62.7) 49 (64.5) 56 (56.6)
T3/T4 62 (35.4) 23 (46.9) 39 (31.0) 24 (31.6) 38 (38.4)

Cancer Gene Therapy (2023) 30:752 – 765

Missing 8 (4.6) 0 (0.0 8 (6.3) 3 (3.9) 5 (5.1)
Pathologic N stage 0.217 0.544
Node-negative 63 (36.0) 22 (44,9) 41 (32.5) 28 (36.8) 35 (35.4)
Node-positive 79 (45.1) 19 (38.8) 60 (47.6) 30 (39.5) 49 (49.5)
Missing 33 (18.9) 8 (16.3) 25 (19.8) 18 (23.7) 15 (15.2)
Overall stage 0.803 0.196
Stage I/II 61 (34.9) 19 (38.8) 42 (33.3) 31 (40.8) 30 (30.3)
Stage III/IV 107 (61.1) 30 (61.2) 77 (61.1) 42 (55.3) 65 (65.6)
Missing 7 (4.0) 0 (0.0) 7 (5.6) 3 (3.9) 4 (4.0)
Extracapsular spread 0.337 0.605
Present 27 (15.4) 5 (10.2) 22 (17.5) 10 (13.2) 17 (17.2)
Absent 148 (84.6) 44 (89.8) 104 (82.5) 66 (86.8) 82 (82.8)
Differentiation 0.314 0.281
Well 38 (21.7) 11 (22.4) 27 (21.4) 20 (26.3) 18 (18.2)
R. Arora et al.

Moderate 99 (56.6) 30 (61.2) 69 (54.8) 42 (55.3) 57 (57.6)

Poor 35 (20.0) 6 (12.2) 29 (23.0) 12 (15.8) 23 (23.2)
Missing 3 (1.7) 2 (4.1) 1 (0.8) 2 (2.6) 1 (1.0)
Alcohol consumption history 0.609 0.810
Never drinker 30 (17.1) 10 (20.4) 20 (15.9) 14 (18.4) 16 (16.2)
Ever drinker 135 (77.1) 36 (73.5) 99 (78.6) 57 (75.0) 78 (78.8)
Missing 10 (5.7) 3 (6.1) 7 (5.6) 5 (6.6) 5 (5.1)
Smoking history 0.647 0.354
Never smoker 48 (26.3) 11 (22.4) 35 (27.8) 23 (30.3) 23 (23.2)
Ever smoker 128 (73.1) 37 (75.5) 91 (72.2) 52 (68.4) 76 (76.7)
Missing 1 (0.6) 1 (2.0) 0 (0) 1 (1.3) 0 (0)
Median pack-years smoked (IQR) 20 (0–35) 25 (1.7–38.5) 18 (0–30) 0.224 13.5 (0–30) 20 (0.2–40.0) 0.015
P-values were calculated with the Mann–Whitney U test for continuous variables and the χ2 test or Fisher’s exact test for categorical variables, as appropriate.
755
 R. Arora et al.
756
was measured using a BrDU Cell Proliferation ELISA kit (Abcam, ab126556). and grown to near confluency in complete growth medium and then 24 h
The mean ± SEM of relative proliferating cells of the independent serum starved by incubating with 0.2% FBS-containing cell culture medium
experiments is plotted on the y-axis versus the experimental conditions in a 5% CO2 humidified incubator at 37 °C. Using a 200 μL pipette tip, a
on the x-axis of a bar graph. scratch was introduced along the midline of the serum-starved cell
monolayers, followed by a PBS wash to remove floating cells, and
incubation of the cells with 0.2% FBS-containing medium for 30 or 48 h, for
In vitro scratch assay UMSCC29 and CAL33 cells, respectively, in a 5% CO2 humidified incubator
5 × 105 UMSCC29 and CAL33 cells were seeded in each well of a 12-well at 37 °C. Scratch closure in each well was followed by imaging the scratch
tissue culture plate and treated either with control RNAi or NCBP2i for 48 h and surrounding cells in each well at ×10 objective of a DIC microscope

Fig. 2 Kaplan–Meier survival curves for TCGA OSCC patients dichotomized by optimized high vs. low NCBP2 and TFRC expression.
A–C Patients with high NCBP2 expression had significantly poorer OS, DSS, and PFI. D–F Patients with high TFRC expression had significantly
poorer OS and DSS, but not PFI. Hazard ratios and p-values quoted are from univariate coxph models, values in square brackets indicate 95%
confidence intervals.

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
757
Table 2. Demographics of OSCC patients in TCGA cohort.
Variable n (%) unless All patients NCBP2 NCBP2 p-value TFRC TFRC p-value
otherwise indicated expression expression expression expression
below median above median below median above median
Number 275 137 138 – 137 138 –
Sex 0.010 0.020
Male 181 (65.8) 80 (58.4) 101 (73.2) 81 (59.1) 100 (72.5)
Female 94 (34.2) 57 (41.6) 37 (26.8) 56 (40.9) 38 (27.5)
Median age at 62 (53–71) 63 (53–73) 61 (54–69) 0.689 61 (52–71) 63 (55–71) 0.343
diagnosis, years (IQR)
Race 0.003 0.052
White 237 (86.2) 128 (93.4) 109 (79.0) 125 (91.2) 112 (81.2)
African-American 19 (6.9) 4 (2.9) 15 (10.9) 4 (2.9) 15 (10.9)
Asian 9 (3.3) 1 (0.7) 8 (5.8) 4 (2.9) 5 (3.6)
American Indian 1 (0.4) 0 (0) 1 (0.7) 0 (0) 1 (0.7)
Unknown 9 (3.3) 4 (2.9) 5 (3.6) 4 (2.9) 5 (3.6)
Oral cavity 0.524 0.127
Oral tongue 112 (40.7) 55 (40.2) 57 (41.3) 67 (48.9) 45 (32.6)
Oral cavity 65 (23.6) 38 (27.7) 27 (19.6) 30 (21.9) 35 (25.4)
Floor of mouth 57 (20.7) 25 (18.3) 32 (23.2) 24 (17.5) 33 (23.9)
Buccal mucosa 22 (8.0) 9 (6.6) 13 (9.4) 8 (5.8) 14 (10.1)
Alveolar ridge 13 (4.7) 6 (4.4) 7 (5.1) 5 (3.7) 8 (5.8)
Hard palate 6 (2.2) 4 (2.9) 1 (1.5) 3 (2.2) 3 (2.2)
Pathologic stage 0.121 0.001
Stage I/II 65 (25.5) 37 (29.8) 28 (21.4) 43 (35.0) 22 (16.7)
Stage III/IV 190 (74.5) 87 (70.2) 103 (78.6) 80 (65.0) 110 (83.3)
Extracapsular spread 0.236 0.073
Present 54 (19.6) 23 (16.8) 31 (22.5)) 21 (15.3) 33 (23.9)
Absent 221 (80.4) 114 (83.2) 107 (77.5) 116 (84.7) 105 (76.1)
Histological grade 0.015 0.086
Grade 1 49 (17.8) 33 (24.1) 16 (11.6) 32 (23.4) 17 (12.3)
Grade 2 168 (61.1) 74 (54.0) 94 (68.1) 75 (54.7) 93 (67.4)
Grade 3 54 (19.6) 27 (19.7) 27 (19.6) 29 (21.2) 25 (18.1)
Alcohol consumption 0.340 0.314
history
0 drinks daily 37 (32.2) 20 (37.7) 17 (27.4) 20 (35.1) 17 (29.3)
1–3 drinks daily 40 (34.8) 15 (28.3) 25 (40.3) 22 (38.6) 18 (31.0)
>3 drinks daily 38 (33.0) 18 (34.0) 20 (32.3) 15 (26.3) 23 (39.7)
Median pack-years 40 (25–54) 37 (20–51) 40 (25–60) 0.101 37 (25–54) 40 (25–60) 0.259
smoked (IQR)
p-values calculated with the Mann–Whitney U test for continuous variables and the χ2 test or Fisher’s exact test for categorical variables, as appropriate.
Significant p-values are bolded.

(Olympus CKX53) coupled to a digital camera at times 0 and 30 or 48 h within a well of a 24-well tissue culture plate and equilibrated with 0.5 mL
after initiating the scratch. Five images were captured along the vertical serum-free DMEM, added both to the upper and lower chambers at 37 °C
axis of the scratch for each experimental condition. The width of each for 2 h. The equilibration media was then gently removed and the upper
scratch was measured at three different positions per image for a total of chamber surface of the insert was coated with 50 μL of 3% Matrigel and
15 measurements using ImageJ (National Institutes of Health, USA), and allowed to solidify at 37 °C for 1 h. 2 × 105 serum-starved OSCC cells were
then averaged per experimental condition. The width average at the resuspended in 0.5 mL of serum-free DMEM and added to the upper
endpoint was subtracted from the width average at 0 h and expressed Matrigel-coated chamber. 500 μL complete growth medium was added to
relative to that at 0 h width for each experimental condition to obtain the lower chamber. Cells were allowed to invade the matrix overnight at
scratch closure and expressed as percent scratch closure. The mean ± SEM 37 °C after which non-adherent cells were removed by PBS washing of cell
of relative scratch closure of the independent experiments is plotted on layers on the upper chamber three times. During the second wash, a
the y-axis versus the experimental conditions on the x-axis of a bar graph. cotton tip applicator was used to gently scrape away the adherent cells on
the upper surface of the membrane. Invading cells were fixed by
In vitro transwell invasion assays immersing the transwell inserts in 100% methanol for 20 min at −20 °C,
Overnight 0.2% FBS-containing media, i.e. serum-starved, control RNAi or followed by staining with 0.5% crystal violet dye (EMD Millipore, Canada)
NCBP2i treated, UMSCC29 and CAL33 cells were used for the transwell for 1 h at room temperature. Six randomly chosen fields of each stained
invasion using polycarbonate filters (24-well inserts, pore size 8 μm; BD membrane were imaged at ×10 objective of a DIC microscope (Olympus
Biosciences, Canada). Prior to the addition of cells, each insert was placed CKX53) coupled to a digital camera. Crystal violet-stained cells in each field

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
758

Fig. 3 Altered expression of NCBP2 and TFRC in OSCC cells. mRNA expression of NCBP2 (A) and TFRC (B) in OSCC relative to various normal
tissues shows higher expression in OSCC as compared to most normal human tissue types. Transcripts per million (TPM) data was obtained
from TCGA and Genotype-Tissue Expression (GTEx) project. Single-cell mRNA expression of NCBP2 (C) and TFRC (D) in 18 head and neck
squamous cell carcinoma as available from Puram et al. [29] showing high expression in OSCC cells as opposed to other cells in the tumour
microenvironment.

were counted using a handheld counter and an average count of cells for expression is altered in malignant tissue [27]. However, we did
the six fields per condition was obtained. Each experiment was repeated at not find a significant correlation between beta-values derived
least three independent times, and invading cell counts at each from TCGA Illumina Infinium Human Methylation 450K array data
experimental condition were expressed relative to the respective control and the expression levels of GMPS, RFC4, NCBP2 and TFRC genes
RNAi-treated condition. The mean ± SEM of relative invading cells of the (Fig. S3 and Table S2) in OSCC.
independent experiments is plotted on the y-axis versus the experimental
conditions on the x-axis of a bar graph. To better understand the clinical relevance of the four genes
driven by 3q22-3q29 amplification, we performed survival analysis
with OS as endpoint in HPV-negative OSCC samples. 3q22-3q29
amplification itself was not associated with OS in HPV-negative
RESULTS OSCC patients (Fig. S4). However, increased expression of all four
The expression of NCBP2 and TFRC genes is associated with genes was associated with worse OS in HPV-negative OSCC samples
clinical outcomes (Cox proportional hazard ratio > 1; p-value < 0.05). Several studies
To identify putative oncogenic drivers among the 468 genes on have described that telomeric aberrations play a critical role in
3q22–3q29, we performed a series of DEA and survival analyses tumourigenesis [28]. Thus, we selected NCBP2 and TFRC, both genes
(Fig. 1A). First, we performed DEA between 3q22-3q29 amplified on the telomeric cytoband 3q29, for further analyses. In con-
(41) vs. non-amplified (228) TCGA-OSCC samples and DEA cordance with the DE analyses described above, we found the gene
between TCGA-OSCC tumours (275) vs. normal samples (26). expression of TFRC and NCBP2 faithfully tracked 3q22-3q29
Then we performed a DE meta-analysis between OSCC and amplification status in TCGA HPV-negative OSCC samples (Fig. 1B, C).
normal samples from six different datasets (Table S1), including Upon performing univariate Cox proportional hazards (Coxph)
TCGA [19–23]. After filtering the genes to those that were analysis in TCGA HPV-negative OSCC patients, high NCBP2 expression
common in all these DE analyses and to the 468 genes on was associated with significantly worse OS (hazard ratio [HR] =
3q22-3q29, we observed that the overexpression of four genes 1.7009 [95% CI: 1.171–2.47], p = 0.00527), but not DSS (HR = 1.3631
(GMPS, RFC4, NCBP2 and TFRC) were directly associated with 3q22- [95% CI: 0.8386–2.216], p = 0.211), or PFI (HR = 1.1368 [95% CI:
3q29 amplification (Fig. S2). We also investigated if the expression 0.7691–1.68], p = 0.52). High TFRC expression was also associated
of the four genes was correlated with the methylation of their with significantly worse OS (HR = 1.3152 [95% CI: 1.069–1.618],
respective promoters, another mechanism by which gene p = 0.0094)], but not DSS [HR = 1.1944 [95% CI: 0.9097–1.568],

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
759

Fig. 4 Kaplan–Meier survival curves for Ohlson OSCC patients dichotomized by optimized high vs. low NCBP2 and TFRC protein
abundance. A Representative TMA slides for each NCBP2 and TFRC staining score established by a trained pathologist. A score of 0 indicates
no antibody staining, and scores of 1, 2 and 3 indicate positive stains of increasing intensity. All positive stains show specific staining primarily
localized to the nucleus for NCBP2 and to the cytoplasm and membrane for TFRC, corresponding to their cellular localizations. Kaplan–Meier
survival curves by high (2/3) vs. low (0/1) NCBP2 and TFRC expression reveal patients with high NCBP2 expression had significantly poorer OS
(B), DSS (C), but not PFI (D). Patients with high TFRC expression had significantly poorer OS (E), DSS (F), and PFI (G). Hazard ratios and p-values
quoted are from univariate coxph models, values in square brackets indicate 95% confidence intervals. Scale bars indicate 50 μm.

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
760
p = 0.201), or PFI [HR = 1.09919 [95% CI: 0.8786–1.375], p = 0.408). TFRC expression was observed in tumour cells, dendritic cells, and
Kaplan–Meier (KM) visualization with optimized cut-points showed macrophages (Fig. 3C, D). Thus, this provides a therapeutic window
that high NCBP2 expression was associated with worse OS, DSS, and for targeting NCBP2 and to a lesser extent TFRC in OSCC patients
PFI (Fig. 2A–C). Similarly, patients with high TFRC expression were with 3q22-29 amplification and NCBP2/TFRC overexpression.
associated with worse OS and DSS, but not PFI (Fig. 2D–F). Although
the continuous mRNA levels of NCBP2 or TFRC are not significantly TFRC and NCBP2 protein abundance is associated with
associated with DSS and PFI in univariate Coxph analysis, the KM demographic differences and patient survival
visualizations indicate that these biomarkers may still retain value for To mitigate the often-observed poor correlation between mRNA and
the risk stratification in OSCC. protein levels and the limitations inherent in TCGA clinical data, we
Clinico-pathological characteristics of the HPV-negative TCGA characterized NCBP2 and TFRC protein expression using IHC in TMAs
OSCC patients stratified by median TFRC and NCBP2 mRNA constructed from a retrospective cohort of OSCC patients at the
expression are presented in Table 2. Patients with high NCBP2 University of Calgary (Fig. 4A). NCBP2 and TFRC specific IHC
expression were more likely to be male, African-American and conditions were optimized using normal mouse and human tissue
Asian by race and present with higher histologic grade. Patients samples (Fig. S1). Of the 183 total patients in the TMA cohort, 8
with high TFRC expression were more likely to be male and (4.3%) could not be assayed for both genes and were removed from
present with a higher pathologic stage. There were no significant further analysis, leaving 175 analysed patients. NCBP2 and TFRC
group differences observed for the median age of diagnosis, proteins were primarily expressed in the nuclear and cytoplasmic
subsite, alcohol consumption history, smoking history, and nodal compartments, respectively, and expression of both proteins was
extracapsular spread. We also observed a significant correlation primarily restricted to squamous cell carcinoma cells in the tumour
between NCBP2 and TFRC expression in OSCC patients (Spearman microenvironment. Table 2 describes the demographics of the
ρ = 0.68, p < 2.2e−16; Fig. S5), which is expected due to their co- University of Calgary OSCC cohort stratified by NCBP2 and TFRC
location on the same chromosomal cytoband that is amplified. protein levels. Patients with high TFRC protein expression had
smoked more pack-years (p = 0.015). NCBP2 and TFRC protein levels
NCBP2 and TFRC are potential therapeutic targets were not correlated in OSCC patients (Spearman’s correlation
Since both increased NCBP2 and TFRC expression are associated with (ρ) = 0.082, p = 0.283; Table S3) unlike the mRNA levels, suggesting
worse survival in OSCC, these genes could potentially be targeted for that protein expression may be influenced by post-transcriptional
therapeutic benefit. However, systemic administered targeted and post-translational modifications.
therapies would also suppress the expression of these genes in We further sought to determine the clinical impact of NCBP2
other high-expressing normal tissues, potentially resulting in adverse and TFRC protein levels in OSCC patients by performing survival
side effects. Analysis of GTEx and TCGA expression data revealed that analyses with OS, DSS and PFI as end-points. Comparisons of OS,
NCBP2 and TFRC expression was negligible across all normal tissues DSS and PFI were first conducted based on the continuous TFRC
evaluated (except for TFRC expression in the bone marrow) and was and NCBP2 protein scores. In univariate Coxph analysis, patients
significantly higher in OSCC samples (Fig. 3A, B). Analysis of single- with higher NCBP2 or TFRC expression were associated with
cell RNAseq data from a study published by Puram et al. [29] also significantly worse OS, DSS, and PFI (Table 3).
shows NCBP2 expression to be significantly higher in OSCC cells Kaplan–Meier visualizations were then performed to further
compared to other cell types in the tumour microenvironment, while assess the prognostic value of these protein biomarkers. TFRC and

Table 3. Cox regression for survival conditions among OSCC patients in the Ohlson TMA cohort.
Variable Univariate analysis Multivariate analysis

p-value Hazard 95% CI for HR p-value Hazard 95% CI for HR

ratio (HR) ratio (HR)
Lower Upper Lower Upper
Overall survival
Age at diagnosis 3.09*10−6 1.035 1.020 1.050 4.75*10−9 1.048 1.032 1.065
Clinical Stage I/II vs. III/IV 0.002 2.143 1.336 3.438 0.243 1.352 0.815 2.243
Extracapsular spread (present) 6.2*10−6 3.038 1.876 4.919 7.03*10−9 5.365 3.038 9.473
TFRC protein score 0.002 1.464 1.157 1.852 5.39*10−4 1.561 1.213 2.009
NCBP2 protein score 0.005 1.436 1.117 1.846 0.016 1.381 1.062 1.796
Disease-specific survival
Median age 0.001 1.03 1.012 1.048 1.95*10−5 1.050 1.027 1.073
−4
Pathological stage Stage I/II 1.04*10 4.045 1.997 8.192 0.012 2.558 1.229 5.324
vs. III/IV
Extracapsular spread (present) 4.08*10−5 3.185 1.831 5.54 1.47*10−6 5.401 2.719 10.729
−4 −4
TFRC protein score 4.49*10 1.677 1.256 2.239 6.14*10 1.717 1.260 2.339
NCBP2 protein score 0.003 1.589 1.172 2.154 0.012 1.486 1.092 2.023
Progression-free interval
Median age 0.076 1.016 0.998 1.035 0.028 1.023 1.002 1.044
Clinical Stage I/II vs. III/IV 0.015 2.062 1.154 3.684 0.260 1.429 0.768 2.661
Extracapsular spread (present) 0.002 2.570 1.432 4.614 0.002 2.862 1.458 5.622
TFRC protein score 0.009 1.469 1.100 1.963 0.016 1.438 1.069 1.935
NCBP2 protein score 0.040 1.392 1.015 1.910 0.049 1.368 1.001 1.871
Significant p-values are bolded.

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
761

Fig. 5 Association of NCBP2 depletion in OSCC cell lines with oncogenic progression. A Immunoblot showing reduction in NCBP2 levels
after siRNA (NCBP2i)-mediated knockdown in UMSCC29 and CAL33 cell lines. β-actin immunoblot was performed as a control. Reduction in
endogenous NCBP2 levels by NCBP2i led to reduced OSCC cell proliferation (B), migration (C), and invasion (D) of both UMSCC29 and CAL33
cells. Scale bars indicate 500 μm. Statistical significance (unpaired t-test): *p < 0.05, **p < 0.01, and ***p < 0.001.

NCBP2 protein scores were dichotomized into high (+2/+3) or appropriate cell line-based analyses. We depleted NCBP2 expres-
low (0/+1) groups. Patients with high NCBP2 protein expression sion in two OSCC cell lines UMSCC29 and CAL33 using a siRNA
had significantly worse OS and DSS, but not PFI (Fig. 4B–D). pool (NCBP2i). Immunoblotting analysis showed a marked
Patients with high TFRC protein expression also had significantly reduction in NCBP2 protein abundance in NCBP2i-treated cells
worse OS, DSS, and PFI (Fig. 4E–G). as compared to those treated with scrambled siRNA (Fig. 5A).
Using a BrDU incorporation assay we observed that depletion of
NCBP2 and TFRC protein expression is associated with clinical endogenous NCBP2 led to reduced cell proliferation of both the
outcomes on multivariate analysis cell lines (Fig. 5B). In vitro scratch assays were performed to test
Since association with survival outcomes may be confounded by the effect of NCBP2 knockdown on the migratory behaviour of
other clinical variables, we performed multivariate Coxph analysis OSCC cells. We found that NCBP2 depletion significantly reduced
to control for relevant clinical covariates. In univariate Coxph, the speed of scratch closure compared to control in both
pathological stage (I/II vs. III/IV), extracapsular spread, and the UMSCC29 and CAL33 cells (Fig. 5C). In addition to migration,
continuous TFRC and NCBP2 protein scores were each significantly invasion plays an important role in the ability of cancer cells to
associated with worse OS, DSS, and PFI. Age at diagnosis was move to sites outside the primary tumour site and initiate
significantly associated with OS and DSS, but not PFI. A metastasis. The effect of NCBP2 knockdown on the invasive ability
multivariate Cox model was constructed using these covariates, of OSCC cells was tested using transwell MatrigelTM assay. NCBP2i
which found age, extracapsular spread, TFRC protein score and significantly reduced the relative number of invading cells as
NCBP2 protein score to be associated with worse OS. Age, clinical compared to the control in both cell lines (Fig. 5D).
stage, extracapsular spread, TFRC protein score and NCBP2 protein
score were all associated with worse DSS in multivariate Cox NCBP2 promotes the expression of several genes involved in
analysis. Age, extracapsular spread, TFRC protein score and NCBP2 tumour progression in OSCC
protein score were all associated with worse PFI on multivariate Since our studies indicate that NCBP2 is a novel tumour-
Cox analysis (Table 3). promoting gene and very little is known about the downstream
effector genes regulated by NCBP2, we performed DEA between
NCBP2 depletion suppresses OSCC cell migration, invasion, top (n = 68) vs. bottom quartile (n = 69) NCBP2 expressing TCGA-
and proliferation OSCC samples. We filtered the resulting gene list based on an
TFRC protein has been shown to regulate the progression of absolute fold change >1.5 and FDR of 5% (p < 0.05) (Fig. 6A). We
several squamous epithelial tumours [18, 30, 31], however, no then performed an extensive literature review to shortlist 12 genes
functional analysis has been performed on the ability of NCBP2 to with well-studied oncogenic roles in OSCC and potential targeted
regulate tumour progression. Therefore, we evaluated the therapies available or in development targeting these genes
functional relevance of NCBP2 on OSCC progression using (Table 4, Fig. 6B). These results corroborate the tumour-promoting

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
762

Fig. 6 Upregulated expression of specific genes in NCBP2-amplified TCGA OSCC patient samples. A Volcano plot showing the differentially
expressed genes in NCBP2 4th quartile vs. 1st quartile expressing HPV-negative OSCC tumours with well-established oncogenic drivers
highlighted. B Boxplots showing the significantly elevated expression of 12 oncogenic drivers of OSCC in NCBP2 high-expressing tumours
compared to low-expressing. Statistical significance (Wilcoxon signed rank test): *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

effects of NCBP2 observed in our functional studies and provide genomic and transcriptomic studies have associated increased
important clues to how downstream NCBP2 signalling may lead to TFRC expression with the prognosis of various cancers
a more aggressive tumour abetting phenotype and how such [30, 31, 34–36], including OSCC. However, there is very little
signalling may be effectively targeted for therapeutic benefit. known about the involvement of NCBP2 in carcinogenesis and
progression [37]. Also, our study is the first to identify that the
expression of NCBP2 and TFRC genes is driven by amplification.
DISCUSSION Interestingly, the amplification of the entire 3q22-29 locus itself
Due to the high morbidity associated with the current clinical was not associated with prognosis (Fig. S4) while the expression of
management of OSCC [1, 3], it is essential to identify putative individual genes within this locus was associated with patient
oncogenes driving carcinogenesis that may also be used as outcomes. This indicates that genes present on amplified
prognostic factors and targeted for therapeutic benefit. Recently, chromosomal regions might be involved in complex cellular
the advancement of high-throughput multi-omic technologies has processes, the impact of which cannot be adequately captured by
facilitated comprehensive molecular profiling of tumour samples querying the amplification status of the region containing
to identify drivers of oncogenesis and progression, which may these genes.
lead to the development of precision oncotherapeutics [32, 33]. Given the growing number of studies describing the prognostic
Here, we analysed OSCC genomes and transcriptomes to identify value of TFRC in squamous cell carcinomas [30, 31, 34–36], we
candidate driver oncogenes on the chromosomal cytobands focused our attention on NCBP2 for further assessing effects on
3q22-3q29, which is frequently amplified in squamous cell cancer aggressiveness. Our in vitro results also suggest that NCBP2
carcinomas [14–17]. Using an intuitive filtering technique to drives OSCC proliferation, migration, and invasion of OSCC cell
analyse TCGA and other publicly available datasets, we identified lines, highlighting the potential for exploiting NCBP2 as a
two genes located on 3q22-3q29—NCBP2, TFRC—with potential therapeutic target (Fig. 5).
clinical relevance in HPV-negative OSCC. Leveraging data from It is also noteworthy that high TFRC expression was associated
multiple datasets of OSCC patients increases the validity of our with higher pack-years smoked (Table 2). Other studies have noted
findings (Fig. 1A). Both NCBP2 and TFRC were found to be that higher cigarette consumption is associated with poor prognosis
amplified and overexpressed in OSCC compared to normal oral and immunosuppression [38, 39]. Interestingly, our single-cell
cavity squamous epithelium, with increased expression of both RNAseq analysis revealed significant expression of TFRC in dendritic
genes associated with worse prognosis (Figs. 1 and 2). Given that cells and macrophages, whereas NCBP2 expression was almost
TCGA lacks protein expression data for these biomarkers, we entirely restricted to tumour cells. Also, TFRC expression was
sought to assess the clinical relevance of NCBP2 and TFRC proteins observed in bone marrow samples from GTEx (Fig. 3B, D). Therefore,
using IHC on TMAs associated with prospectively collected clinical it may be useful to further investigate if TFRC regulates immune
data from an in-house cohort of OSCC patients. OSCC patient responses in the tumour microenvironment.
outcomes significantly differed by NCBP2 and TFRC levels, with Our reported association between DAB IHC-based TFRC and
higher protein expression scores associated with worse OS, DSS NCBP2 protein expression score and poor survival further
and PFI (Fig. 4). Multivariate Cox regression analysis suggests that demonstrates that both TFRC and NCBP2 protein expression
NCBP2 and TFRC are independent prognostic factors in OSCC and may be used as a prognostic marker in the clinical management of
can provide prognostic value in addition to the currently used OSCC. DAB-based IHC staining is a cost-effective and commonly
TNM staging system (Table 3). Any differences in the survival used tool in pathology. Therefore, our DAB IHC-based assay offers
analyses between TCGA and our prospective TMA cohort likely a clinically feasible way to measure biomarker expression in OSCC
reflect the lack of correlation between mRNA and protein patients that is less complex than assessing multigene prognostic
expression. We have also demonstrated that the NCBP2 and TFRC signatures [40]. These novel biomarkers may provide an additional
expression is correlated at the mRNA level, but not at the protein prognostic tool to clinicians besides currently used tumour staging
level. This difference could be explained in part by the loss of approaches, allowing for more informed treatment decisions.
sensitivity in evaluating protein expression semi-quantitatively via Other recent studies have also identified novel prognostic markers
IHC, which may mask the underlying correlation between NCBP2 in oral cancers using IHC [7, 41, 42]. Thus, we propose TFRC and
and TFRC levels. Furthermore, NCBP2 and TFRC may be regulated NCBP2 as novel additions to a growing body of potential
post-transcriptionally or post-translationally in different ways, prognostic biomarkers in OSCC.
reducing the correlation between these proteins compared to We found that NCBP2 expression was significantly upregulated
the correlation in gene expression. in tumour cells compared to normal human tissue samples from
Collectively, our results provide substantial evidence for the role the GTEx consortium, and other cells in the tumour microenviron-
of NCBP2 and TFRC as driver oncogenes in OSCC. Several cancer ment [29] (Fig. 3A, C). This provides a good therapeutic window to

Cancer Gene Therapy (2023) 30:752 – 765

 Table 4. The fold change, adjusted p-value, reported role in OSCC and therapeutics that are known to target genes upregulated in NCBP2-overexpressing patients.
Gene name Protein coded Fold change Adjusted Role in OSCC Known therapeutic interventions
p-value
ALDH1A1 Aldehyde 2.68 5.63E−03 High ALDH1A1 expression was associated with higher TNM Disulfiram (DSF)–copper (Cu) complex [44].
dehydrogenase 1A1 tumour stage and high nodal stage and high mortality rate
in OSCC [43]
EPCAM Epithelial cell 2.75 2.27E−05 EpCAM regulates cyclin D1 expression and localization in Several monoclonal antibodies [48].
adhesion molecule OSCC cells under anchorage-independent conditions [45].
Overexpression of EpCAM was associated significantly with
OSCC tumour size, histological grade, local recurrence of
tumour and patient survival [46, 47]
EpCAM reduction decreased the invasion potential and
proliferation activity of OSCC cells [47]

Cancer Gene Therapy (2023) 30:752 – 765

FGF1 Fibroblast growth 3.39 5.26E−05 FGF1 promotes EMT and thus invasion and metastasis in FGF signalling axis targeting small-molecule receptor TKIs
factor 1 OSCC [49, 50] (non-selective, selective and covalent), monoclonal
FGF2 Fibroblast growth 2.30 1.30E−02 There is a significant association of FGF-2 expression with antibodies, FGF ligand traps and DNA/RNA aptamers [51]
factor 2 malignant transformation of OSCC and high-grade tumours
and correlated with the presence of metastasis and adverse
outcomes [52, 53]
FGFR4 Fibroblast growth 2.20 1.08E−03 FGFR4 is frequently overexpressed in OSCC [54, 55]
factor receptor 4
FN1 Fibronectin 1 2.29 1.55E−03 FN1 is upregulated and correlates to poor prognosis in Expression of fibronectin 1 or its binding to its partners are
OSCC [56] and promotes the proliferation, invasion and attractive targets for novel drug development [58]
migration of OSCC cells [57]
HDAC8 Histone 1.76 4.91E−02 HDAC8 is overexpressed in OSCC tissues and cell lines. Several HDAC inhibitors have shown promise in suppressing
deacetylase 8 HDAC8 silencing enhanced apoptosis, suppressed cell tumour progression [61]. HDAC8-specific inhibitor PCI-
growth and metastasis in OSCC cells [59, 60] 34051 with a >200-fold selectivity over other HDAC isoforms
has shown promise in preclinical models of HCC [62]
IGF2BP2 Insulin growth 1.58 1.55E−03 IGF2BP2 was highly expressed in OSCC and significantly Very recently, a few targeted therapies are being evaluated
factor 2 binding correlated with poor overall survival of OSCC patients. for IGF2BP2 (35023719, 35915142)
R. Arora et al.

protein 2 Apoptosis-, tumour-, and immune-related pathways were

significantly enriched in samples with high IGF2BP2
expression. IGF2BP2 co-expressed genes indicated that
these genes are mainly associated with immunity/
inflammation and tumorigenesis. In addition, IGF2BP2 and
its co-expressed genes are associated with TICs [63, 64]
IGFBP2 Insulin growth 1.88 7.46E−03 IGFBP-2 is strongly correlated with oral cancer metastasis Antisense oligonucleotide or neutralizing antibodies have
factor binding and progression [65] been developed to target IGFBP2 [66]
protein 2
PRKCA Protein kinase 1.53 4.97E−03 PRKCA overexpression is enriched in young OSCC patients There is a lack of isozyme-specific PKC inhibitors. However, a
C alpha and is associated with poor prognosis [67] few are being developed [68]
SOX2 SRY-box 2 2.29 4.38E−03 SOX2 expression is an independent predictor of oral cancer Very few SOX2 targeting therapeutic agents are available
risk in patients with oral leukoplakia [69]. and they are in very early stages of development [72]
SOX2 promotes tumour aggressiveness and epithelial-
mesenchymal transition in OSCC and associated with LNM
[70, 71]
TWIST1 Twist-related 1.51 3.56E−02 TWIST1 is significantly overexpressed in advanced stages of A naturally occurring alkaloid harmine has recently been
protein 1 OSCC and its expression predicted LNM and poor patient shown to promote TWIST1 degradation [75]
survival [73, 74]
763
 R. Arora et al.
764
develop targeted therapies against NCBP2 that might be less toxic 20. Estilo CL, Oc P, Talbot S, Socci ND, Carlson DL, Ghossein R, et al. Oral tongue
to normal cells. Nagai et al. previously demonstrated that a TFRC cancer gene expression profiling: Identification of novel potential prognosticators
antibody inhibited the growth of OSCC tumours in a murine by oligonucleotide microarray analysis. BMC Cancer. 2009;9:11.
xenograft model [18]. However, since we detected TFRC expres- 21. Reis PP, Waldron L, Perez-Ordonez B, Pintilie M, Galloni NN, Xuan Y, et al. A gene
signature in histologically normal surgical margins is predictive of oral carcinoma
sion in the bone marrow and immune cells in the tumour
recurrence. BMC Cancer. 2011;11:437.
microenvironment (Fig. 3B, D), TFRC-targeting might be less 22. Toruner GA, Ulger C, Alkan M, Galante AT, Rinaggio J, Wilk R, et al. Association
desirable compared to ablating NCBP2. IHC-based assessment of between gene expression profile and tumor invasion in oral squamous cell car-
NCBP2 protein levels could be employed as a companion cinoma. Cancer Genet Cytogenet. 2004;154(1):27–35.
diagnostic for potential NCBP2-targeting therapies, helping tailor 23. Ye H, Yu T, Temam S, Ziober BL, Wang J, Schwartz JL, et al. Transcriptomic
treatment to patients whose tumours are driven by NCBP2. We dissection of tongue squamous cell carcinoma. BMC Genomics. 2008;9:69.
also identified potential downstream effectors of NCBP2 in OSCC 24. McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM, et al.
(Fig. 6), some of which may be targeted by available small Reporting recommendations for tumor marker prognostic studies (REMARK). J
molecule inhibitors (Table 4). Deploying such targeted therapeu- Natl Cancer Inst. 2005;97(16):1180–4.
25. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion
tics in the clinic could accelerate the progress in improving OSCC
for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):550.
survival outcomes and help ameliorate the morbidity associated 26. Contal C, O’Quigley J. An application of changepoint methods in studying the effect
with current OSCC treatment regimens. of age on survival in breast cancer. Comput Stat Data Anal. 1999;30(3):253–70.
27. Nishiyama A, Nakanishi M. Navigating the DNA methylation landscape of cancer.
Trends Genet. 2021;37(11):1012–27.
DATA AVAILABILITY 28. Maciejowski J, de Lange T. Telomeres in cancer: tumour suppression and genome
Materials described in the manuscript, including all relevant raw data, will be freely instability. Nat Rev Mol Cell Biol. 2017;18(3):175–86.
available to any researcher wishing to use them for non-commercial purposes, 29. Puram SV, Tirosh I, Parikh AS, Patel AP, Yizhak K, Gillespie S, et al. Single-cell
without breaching participant confidentiality. transcriptomic analysis of primary and metastatic tumor ecosystems in head and
neck. Cancer Cell 2017;171(7):1611–24e24.
30. Huang Y, Huang J, Huang Y, Gan L, Long L, Pu A, et al. TFRC promotes epithelial
REFERENCES ovarian cancer cell proliferation and metastasis via up-regulation of AXIN2
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global expression. Am J Cancer Res. 2020;10(1):131–47.
Cancer Statistics 2020: GLOBOCAN estimates of incidence and mortality world- 31. Chan KT, Choi MY, Lai KK, Tan W, Tung LN, Lam HY, et al. Overexpression of
wide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49. transferrin receptor CD71 and its tumorigenic properties in esophageal squa-
2. Cohen N, Fedewa S, Chen AY. Epidemiology and demographics of the head and mous cell carcinoma. Oncol Rep. 2014;31(3):1296–304.
neck cancer population. Oral Maxillofac Surg Clin North Am. 2018;30(4):381–95. 32. Malone ER, Oliva M, Sabatini PJB, Stockley TL, Siu LL. Molecular profiling for
3. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. precision cancer therapies. Genome Med. 2020;12(1):8.
2020;70(1):7–30. 33. The L. 20 years of precision medicine in oncology. Lancet. 2021;397:1781.
4. Economopoulou P, Kotsantis I, Psyrri A. Special issue about head and neck can- 34. Wei S, Wang Y, Xu H, Kuang Y. Screening of potential biomarkers for chemore-
cers: HPV positive cancers. Int J Mol Sci. 2020;21(9). sistant ovarian carcinoma with miRNA expression profiling data by bioinformatics
5. Chinn SB, Myers JN. Oral cavity carcinoma: current management, controversies, approach. Oncol Lett. 2015;10(4):2427–31.
and future directions. J Clin Oncol. 2015;33(29):3269–76. 35. Shen Y, Li X, Dong D, Zhang B, Xue Y, Shang P. Transferrin receptor 1 in cancer: a
6. Rivera C. Essentials of oral cancer. Int J Clin Exp Pathol. 2015;8(9):11884–94. new sight for cancer therapy. Am J Cancer Res. 2018;8(6):916–31.
7. Cervino G, Fiorillo L, Herford AS, Romeo U, Bianchi A, Crimi S, et al. Molecular 36. Wada S, Noguchi T, Takeno S, Kawahara K. PIK3CA and TFRC located in 3q are
biomarkers related to oral carcinoma: clinical trial outcome evaluation in a lit- new prognostic factors in esophageal squamous cell carcinoma. Ann Surg Oncol.
erature review. Dis Markers. 2019;2019:8040361. 2006;13(7):961–6.
8. Kozaki K, Imoto I, Pimkhaokham A, Hasegawa S, Tsuda H, Omura K, et al. PIK3CA 37. Hu G, Jiang Q, Liu L, Peng H, Wang Y, Li S, et al. Integrated analysis of RNA-
mutation is an oncogenic aberration at advanced stages of oral squamous cell binding proteins associated with the prognosis and immunosuppression in
carcinoma. Cancer Sci. 2006;97(12):1351–8. squamous cell carcinoma of head and neck. Front Genet. 2020;11:571403.
9. Starzynska A, Adamska P, Sejda A, Sakowicz-Burkiewicz M, Adamski LJ, Marvaso 38. de la Iglesia JV, Slebos RJC, Martin-Gomez L, Wang X, Teer JK, Tan AC, et al. Effects
G, et al. Any role of PIK3CA and PTEN biomarkers in the prognosis in oral of tobacco smoking on the tumor immune microenvironment in head and neck
squamous cell carcinoma? Life (Basel). 2020;10. squamous cell carcinoma. Clin Cancer Res. 2020;26(6):1474–85.
10. Taoudi Benchekroun M, Saintigny P, Thomas SM, El-Naggar AK, Papadimi- 39. Cortellini A, De Giglio A, Cannita K, Cortinovis DL, Cornelissen R, Baldessari C, et al.
trakopoulou V, Ren H, et al. Epidermal growth factor receptor expression and gene Smoking status during first-line immunotherapy and chemotherapy in NSCLC
copy number in the risk of oral cancer. Cancer Prev Res (Philos). 2010;3(7):800–9. patients: a case-control matched analysis from a large multicenter study. Thorac
11. Matsui A, Ihara T, Suda H, Mikami H, Semba K. Gene amplification: mechanisms Cancer. 2021;12(6):880–9.
and involvement in cancer. Biomol Concepts. 2013;4(6):567–82. 40. Li H, Zhang X, Yi C, He Y, Chen X, Zhao W, et al. Ferroptosis-related gene signature
12. Ohshima K, Hatakeyama K, Nagashima T, Watanabe Y, Kanto K, Doi Y, et al. predicts the prognosis in Oral squamous cell carcinoma patients. BMC Cancer.
Integrated analysis of gene expression and copy number identified potential 2021;21(1):835.
cancer driver genes with amplification-dependent overexpression in 1,454 solid 41. Chang CF, Chen SL, Sung WW, Hsieh MJ, Hsu HT, Chen LH, et al. PBK/TOPK
tumors. Sci Rep. 2017;7(1):641. expression predicts prognosis in oral cancer. Int J Mol Sci. 2016;17(7).
13. Santarius T, Shipley J, Brewer D, Stratton MR, Cooper CS. A census of amplified 42. Santarelli A, Mascitti M, Rubini C, Bambini F, Giannatempo G, Lo Russo L, et al.
and overexpressed human cancer genes. Nat Rev Cancer. 2010;10(1):59–64. Nuclear survivin as a prognostic factor in squamous-cell carcinoma of the oral
14. Davidson MA, Shanks EJ. 3q26–29 Amplification in head and neck squamous cell cavity. Appl Immunohistochem Mol Morphol. 2017;25(8):566–70.
carcinoma: a review of established and prospective oncogenes. FEBS J. 43. Gupta V, Maurya MK, Agarwal P, Kumar M, Sagar M, Raghuvanshi S, et al. Expression
2017;284(17):2705–31. of aldehyde dehydrogenase 1A1 in oral squamous cell carcinoma and its correlation
15. Fields AP, Justilien V, Murray NR. The chromosome 3q26 OncCassette: a multi- with clinicopathological parameters. Natl J Maxillofac Surg. 2022;13(2):208–15.
genic driver of human cancer. Adv Biol Regul. 2016;60:47–63. 44. Chen SY, Chang YL, Liu ST, Chen GS, Lee SP, Huang SM. Differential cytotoxicity
16. Voutsadakis IA. 3q26 Amplifications in cervical squamous carcinomas. Curr Oncol. mechanisms of copper complexed with disulfiram in oral cancer cells. Int J Mol
2021;28(4):2868–80. Sci. 2021;22(7).
17. Ambele MA, van Zyl A, Pepper MS, van Heerden MB, van Heerden WFP. Ampli- 45. Inoue H, Ohnishi Y, Nakajima M, Kakudo K, Nozaki M. A novel function of EpCAM
fication of 3q26.2, 5q14.3, 8q24.3, 8q22.3, and 14q32.33 are possible common in oral squamous cell carcinoma cells under anchorage-independent conditions.
genetic alterations in oral cancer patients. Front Oncol. 2020;10:683. Int J Oncol. 2011;39(6):1401–5.
18. Nagai K, Nakahata S, Shimosaki S, Tamura T, Kondo Y, Baba T, et al. Development 46. Sen S, Carnelio S. Expression of epithelial cell adhesion molecule (EpCAM) in oral
of a complete human anti-human transferrin receptor C antibody as a novel squamous cell carcinoma. Histopathology 2016;68(6):897–904.
marker of oral dysplasia and oral cancer. Cancer Med. 2014;3(4):1085–99. 47. Yanamoto S, Kawasaki G, Yoshitomi I, Iwamoto T, Hirata K, Mizuno A. Clin-
19. Chen C, Mendez E, Houck J, Fan W, Lohavanichbutr P, Doody D, et al. Gene icopathologic significance of EpCAM expression in squamous cell carcinoma of
expression profiling identifies genes predictive of oral squamous cell carcinoma. the tongue and its possibility as a potential target for tongue cancer gene
Cancer Epidemiol Biomark Prev. 2008;17(8):2152–62. therapy. Oral Oncol. 2007;43(9):869–77.

Cancer Gene Therapy (2023) 30:752 – 765

 R. Arora et al.
765
48. Eyvazi S, Farajnia S, Dastmalchi S, Kanipour F, Zarredar H, Bandehpour M. Anti- 71. Ren ZH, Zhang CP, Ji T. Expression of SOX2 in oral squamous cell carcinoma and
body based EpCAM targeted therapy of cancer, review and update. Curr Cancer the association with lymph node metastasis. Oncol Lett. 2016;11(3):1973–9.
Drug Targets. 2018;18(9):857–68. 72. Zhang S, Xiong X, Sun Y. Functional characterization of SOX2 as an anticancer
49. Jiao J, Zhao X, Liang Y, Tang D, Pan C. FGF1-FGFR1 axis promotes tongue target. Signal Transduct Target Ther. 2020;5(1):135.
squamous cell carcinoma (TSCC) metastasis through epithelial–mesenchymal 73. da Silva SD, Alaoui-Jamali MA, Soares FA, Carraro DM, Brentani HP, Hier M, et al.
transition (EMT). Biochem Biophys Res Commun. 2015;466(3):327–32. TWIST1 is a molecular marker for a poor prognosis in oral cancer and represents a
50. Weaver AN, Burch MB, Cooper TS, Della Manna DL, Wei S, Ojesina AI, et al. Notch potential therapeutic target. Cancer 2014;120(3):352–62.
signaling activation is associated with patient mortality and increased FGF1- 74. Kong YH, Syed Zanaruddin SN, Lau SH, Ramanathan A, Kallarakkal TG, Vincent-
mediated invasion in squamous cell carcinoma of the oral cavity. Mol Cancer Res. Chong VK, et al. Co-expression of TWIST1 and ZEB2 in oral squamous cell car-
2016;14(9):883–91. cinoma is associated with poor survival. PLoS ONE. 2015;10(7):e0134045.
51. Krook MA, Reeser JW, Ernst G, Barker H, Wilberding M, Li G, et al. Fibroblast 75. Yochum ZA, Cades J, Mazzacurati L, Neumann NM, Khetarpal SK, Chatterjee S,
growth factor receptors in cancer: genetic alterations, diagnostics, therapeutic et al. A first-in-class TWIST1 inhibitor with activity in oncogene-driven lung
targets and mechanisms of resistance. Br J Cancer. 2021;124(5):880–92. cancer. Mol Cancer Res. 2017;15(12):1764–76.
52. Nayak S, Goel MM, Makker A, Bhatia V, Chandra S, Kumar S, et al. Fibroblast
growth factor (FGF-2) and its receptors FGFR-2 and FGFR-3 may be putative
biomarkers of malignant transformation of potentially malignant oral lesions into ACKNOWLEDGEMENTS
oral squamous cell carcinoma. PLoS ONE. 2015;10(10):e0138801. The authors would like to thank Dr. Bo Young Ahn for providing guidance regarding
53. Mariz B, Soares CD, de Carvalho MGF, Jorge-Junior J. FGF-2 and FGFR-1 might be the immunohistochemistry staining of tissue microarrays. We are also grateful to the
independent prognostic factors in oral tongue squamous cell carcinoma. Histo- Ohlson Research Initiative for funding the project and providing all necessary
pathology 2019;74(2):311–20. support.
54. Dutra RL, de Carvalho MB, Dos Santos M, Mercante AM, Gazito D, de Cicco R, et al.
FGFR4 profile as a prognostic marker in squamous cell carcinoma of the mouth
and oropharynx. PLoS ONE. 2012;7(11):e50747.
55. Koole K, van Kempen PM, van Bockel LW, Smets T, van der Klooster Z, Dutman AUTHOR CONTRIBUTIONS
AC, et al. FGFR4 is a potential predictive biomarker in oral and oropharyngeal RA, LH, and AC performed the experiments, RA, LH, MK, RM, JA, SCN, DI, and AC
squamous cell carcinoma. Pathobiology 2015;82(6):280–9. performed data analyses, TWM, SC, RDH, and JCD collected the carcinoma samples
56. Xu XL, Liu H, Zhang Y, Zhang SX, Chen Z, Bao Y, et al. SPP1 and FN1 are significant for building the in-house TMA cohort, PB conceived the project, PB and SJMJ
gene biomarkers of tongue squamous cell carcinoma. Oncol Lett. 2021;22(4):713. supervised the project and RA, LH, MK, and AC wrote the paper with input from PB.
57. Zhai J, Luo G. GATA6-induced FN1 activation promotes the proliferation, invasion All authors have read and approved the manuscript.
and migration of oral squamous cell carcinoma cells. Mol Med Rep. 2022;25(3).
58. Spada S, Tocci A, Di Modugno F, Nistico P. Fibronectin as a multiregulatory
molecule crucial in tumor matrisome: from structural and functional features to COMPETING INTERESTS
clinical practice in oncology. J Exp Clin Cancer Res. 2021;40(1):102. The authors declare no competing interests.
59. Ahn MY, Yoon JH. Histone deacetylase 8 as a novel therapeutic target in oral
squamous cell carcinoma. Oncol Rep. 2017;37(1):540–6.
60. Ahn MY. HDAC inhibitor apicidin suppresses murine oral squamous cell carci- ADDITIONAL INFORMATION
noma cell growth in vitro and in vivo via inhibiting HDAC8 expression. Oncol Lett. Supplementary information The online version contains supplementary material
2018;16(5):6552–60. available at https://doi.org/10.1038/s41417-022-00578-8.
61. Bondarev AD, Attwood MM, Jonsson J, Chubarev VN, Tarasov VV, Schioth HB.
Recent developments of HDAC inhibitors: emerging indications and novel Correspondence and requests for materials should be addressed to Pinaki Bose.
molecules. Br J Clin Pharm. 2021;87(12):4577–97.
62. Yang W, Feng Y, Zhou J, Cheung OK, Cao J, Wang J, et al. A selective HDAC8 Reprints and permission information is available at http://www.nature.com/
inhibitor potentiates antitumor immunity and efficacy of immune checkpoint reprints
blockade in hepatocellular carcinoma. Sci Transl Med. 2021;13(588).
63. Wang X, Xu H, Zhou Z, Guo S, Chen R. IGF2BP2 maybe a novel prognostic Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
biomarker in oral squamous cell carcinoma. Biosci Rep. 2022;42(2). in published maps and institutional affiliations.
64. Zhou L, Li H, Cai H, Liu W, Pan E, Yu D, et al. Upregulation of IGF2BP2 promotes
oral squamous cell carcinoma progression that is related to cell proliferation,
metastasis and tumor-infiltrating immune cells. Front Oncol. 2022;12:809589.
65. Tsai YF, Chou HC, Liou MH, Liao EC, Cheng CT, Chang SJ, et al. Role of IGFBP-2 in
oral cancer metastasis. Biochim Biophys Acta Mol Basis Dis. 2021;1867(7):166143. Open Access This article is licensed under a Creative Commons
66. Li T, Forbes ME, Fuller GN, Li J, Yang X, Zhang W. IGFBP2: integrative hub of Attribution 4.0 International License, which permits use, sharing,
developmental and oncogenic signaling network. Oncogene 2020;39(11):2243–57. adaptation, distribution and reproduction in any medium or format, as long as you give
67. Parzefall T, Schnoell J, Monschein L, Foki E, Liu DT, Frohne A, et al. PRKCA appropriate credit to the original author(s) and the source, provide a link to the Creative
Overexpression Is Frequent in Young Oral Tongue Squamous Cell Carcinoma Commons license, and indicate if changes were made. The images or other third party
Patients and Is Associated with Poor Prognosis. Cancers (Basel). 2021;13(9). material in this article are included in the article’s Creative Commons license, unless
68. Sadeghi MM, Salama MF, Hannun YA. Protein kinase C as a therapeutic target in indicated otherwise in a credit line to the material. If material is not included in the
non-small cell lung cancer. Int J Mol Sci. 2021;22(11). article’s Creative Commons license and your intended use is not permitted by statutory
69. de Vicente JC, Donate-Perez Del Molino P, Rodrigo JP, Allonca E, Hermida-Prado regulation or exceeds the permitted use, you will need to obtain permission directly
F, Granda-Diaz R, et al. SOX2 expression is an independent predictor of oral from the copyright holder. To view a copy of this license, visit http://
cancer progression. J Clin Med. 2019;8(10). creativecommons.org/licenses/by/4.0/.
70. Liu X, Qiao B, Zhao T, Hu F, Lam AK, Tao Q. Sox2 promotes tumor aggressiveness
and epithelial–mesenchymal transition in tongue squamous cell carcinoma. Int J
Mol Med. 2018;42(3):1418–26. © The Author(s) 2023

Cancer Gene Therapy (2023) 30:752 – 765