ArticlePDF Available

Comprehensive Investigation of Proteoglycan Gene Expression in Breast Cancer: Discovery of a Unique Proteoglycan Gene Signature Linked to the Malignant Phenotype

Wiley
Proteoglycan Research
Authors:

Abstract and Figures

Solid tumors present a formidable challenge in oncology, necessitating innovative approaches to improve therapeutic outcomes. Proteoglycans, multifaceted molecules within the tumor microenvironment, have garnered attention due to their diverse roles in cancer progression. Their unique ability to interact with specific membrane receptors, growth factors, and cytokines provides a promising avenue for the development of recombinant proteoglycan‐based therapies that could enhance the precision and efficacy of cancer treatment. In this study, we performed a comprehensive analysis of the proteoglycan gene landscape in human breast carcinomas. Leveraging the available wealth of genomic and clinical data regarding gene expression in breast carcinoma and using a machine learning model, we identified a unique gene expression signature composed of five proteoglycans differentially modulated in the tumor tissue: Syndecan‐1 and asporin (upregulated) and decorin, PRELP and podocan (downregulated). Additional query of the breast carcinoma data revealed that serglycin, previously shown to be increased in breast carcinoma patients and mouse models and to correlate with a poor prognosis, was indeed decreased in the vast majority of breast cancer patients and its levels inversely correlated with tumor progression and invasion. This proteoglycan gene signature could provide novel diagnostic capabilities in breast cancer biology and highlights the need for further utilization of publicly available datasets for the clinical validation of preclinical experimental results.
This content is subject to copyright. Terms and conditions apply.
Proteoglycan Research
ORIGINAL RESEARCH
Comprehensive Investigation of Proteoglycan Gene
Expression in Breast Cancer: Discovery of a Unique
Proteoglycan Gene Signature Linked to the Malignant
Phenotype
Simone Buraschi
1
| Gabriel Pascal
1
| Federico Liberatore
2
| Renato V. Iozzo
1
1
Department of Pathology and Genomic Medicine, and the Translational Cellular Oncology Program, Sidney Kimmel Cancer Center, Sidney Kimmel Medical
College at Thomas Jefferson University, Philadelphia, Pennsylvania, USA |
2
School of Computer Science and Informatics, Cardiff University, Cardiff, UK
Correspondence: Simone Buraschi (simone.buraschi@jefferson.edu) | Renato V. Iozzo (renato.iozzo@jefferson.edu)
Received: 23 August 2024 | Revised: 26 November 2024 | Accepted: 6 December 2024
Funding: The original research was supported, in part by National Institutes of Health Grants RO1 CA245311 and RO3 CA270830, grant 3R01CA24531104S1.
Keywords: asporin | decorin | podocan | PRELP | proteoglycan gene signature | syndecan1
ABSTRACT
Solid tumors present a formidable challenge in oncology, necessitating innovative approaches to improve therapeutic outcomes.
Proteoglycans, multifaceted molecules within the tumor microenvironment, have garnered attention due to their diverse roles
in cancer progression. Their unique ability to interact with specific membrane receptors, growth factors, and cytokines provides
a promising avenue for the development of recombinant proteoglycanbased therapies that could enhance the precision and
efficacy of cancer treatment. In this study, we performed a comprehensive analysis of the proteoglycan gene landscape in
human breast carcinomas. Leveraging the available wealth of genomic and clinical data regarding gene expression in breast
carcinoma and using a machine learning model, we identified a unique gene expression signature composed of five pro-
teoglycans differentially modulated in the tumor tissue: Syndecan1 and asporin (upregulated) and decorin, PRELP and po-
docan (downregulated). Additional query of the breast carcinoma data revealed that serglycin, previously shown to be increased
in breast carcinoma patients and mouse models and to correlate with a poor prognosis, was indeed decreased in the vast
majority of breast cancer patients and its levels inversely correlated with tumor progression and invasion. This proteoglycan
gene signature could provide novel diagnostic capabilities in breast cancer biology and highlights the need for further utilization
of publicly available datasets for the clinical validation of preclinical experimental results.
1 | Introduction
Proteoglycans are a highly heterogeneous class of glycanated
proteins primarily found to be secreted in the extracellular
matrix (ECM) of the connective tissues and in the pericellular
and basement membrane spaces [14]. Fortythree proteogly-
cans have been described so far, of which 33 appear to be
glycanated with one or more glycosaminoglycan (GAG) chain
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly
cited.
© 2025 The Author(s). Proteoglycan Research published by Wiley Periodicals LLC.
Abbreviations: AUROC, area under the receiver operating characteristic curve; CAS, ECM, extracellular matrix; CRISPR/Cas9, clustered regularly interspaced short palindromic repeats; EGF,
epidermal growth factor; EGFR, EGF receptor; EMT, epithelial mesenchymal transition; GTEx, genotypetissue expression; HGF, hepatocyte growth factor; HGFR, HGF receptor also known as Met
receptor; IGFIR, insulinlike growth factor receptor I; LOOCV, leaveoneout crossvalidation; PROTAC, proteolysis targeting chimera; RFE, recursive feature elimination procedure; RTK, receptor
tyrosine kinase; SLRP, small leucinerich proteoglycan; TARGET, therapeutically applicable research to generate effective treatments; TCGA, The Cancer Genome Atlas; TGFβ, transforming growth
factor β; VEGF, vascular endothelial growth factor; VEGFR, VEGF receptor.
Simone Buraschi and Gabriel Pascal contributed equally to this study.
1of18Proteoglycan Research, 2025; 3:e70014
https://doi.org/10.1002/pgr2.70014
covalently linked to their individual protein cores. The GAG
chains are long, linear polysaccharides composed of repeating
disaccharide units such as chondroitin sulfate, dermatan sul-
fate, heparan sulfate, keratan sulfate, or the unsulfated hya-
luronan [57]. Proteoglycans play crucial roles in the ECM,
contributing to various biological functions such as structural
support, cell adhesion and migration, tissue osmotic balance,
and regulation of diverse biological processes through inter-
action with growth factors, cytokines, and membrane recep-
tors [810]. Moreover, proteolytic fragments of proteoglycan
protein cores affect tumor angiogenesis [1114] by interacting
with and suppressing the action of vascular receptors [1517].
Because of their critical roles in regulating cellular behavior
andinteractionsacrossECMmoleculesandcells,proteogly-
cans have been described as master regulators of cancer pro-
gression, invasion, and metastatic spreading [18]. Changes in
the expression levels of some proteoglycans are associated
with clinical outcomes of various cancers. High expression of
specific proteoglycans often correlates with poor prognosis,
increased metastatic potential, and resistance to therapy.
Therefore, proteoglycans represent potential biomarkers for
cancer diagnosis, prognosis, and therapeutic targeting [5]. In
this study we discovered a novel proteoglycan gene signature
in breast cancer leveraging available data from the Cancer
Genome Atlas (TCGA) and the GenotypeTissue Expression
(GTEx) databases.
The main goal of our research was to delineate the overall
impact of 43 proteoglycan gene expression profiles in breast
cancer diagnosis and prognosis, focusing on statistically signif-
icant changes in their mRNA expression levels in the breast
tumor tissue compared with the normal, noncancerous, tissue.
The applicability of our analysis lies in the generation of the
scientific basis for future studies focused on proteoglycan roles
in breast cancer pathophysiology including initiation, progres-
sion, and metastasis.
The utility of this study is twofold, such that our analysis of
a large, publicly available data set provides unbiased vali-
dation or criticism of experimentally determined trends of
expression for several proteoglycans, as well as providing
insights into the expression of several proteoglycans in
breast cancer that have previously been unexplored. TCGA
is a cancer genomics program led by the National Cancer
Institute which includes an immense repository of tran-
scriptomic data compiled from several thousand patient
tissue samples collected since 2006. This vast array of
patient data collected over nearly two decades serves as a
unique and evergrowing public resource to members of the
cancer research community by which we may glean novel
insights into the genomic origins of cancer pathophysiology.
Searching through this robust and unbiased body of data has
provided us with a broader perspective of overarching
trends of expression for many proteoglycans within the
context of breast carcinoma, and illuminated opportunities
for future expansion of investigative focus within the field of
proteoglycan science.
The technological advancements of biomedical analysis tools in
the past two decades have produced exciting new ways for us to
understand these genomic origins and the evolution of machine
learning and artificial intelligence is at the forefront of the
contemporary technological revolution. We have sought to
compound the potential of TCGA's robust body of data with
cuttingedge machine learning techniques to drive the
advancement of proteoglycanbased diagnostic knowledge so
that we may continue to advance the field of proteoglycan sci-
ence and provide new directions for future investigation and
ultimately develop innovative proteoglycan centered therapeu-
tic approaches in the future. This analysis aids in prioritizing
proteoglycan candidates for further investigation with a data
driven approach to identify promising undiscovered therapeutic
avenues. The integration of the proteoglycan expression pro-
files, their biological functions, and clinical outcomes has the
potential to guide the development of precision therapies that
may enhance the prognosis of solid tumor patients.
2 | Methods
2.1 | Data Set
The data set employed in this study consists of a combined
cohort derived from TCGA, Therapeutically Applicable
Research to Generate Effective Treatments (TARGET), and
GenotypeTissue Expression (GTEx) projects, accessible via the
UCSC Xena Project [19], and publicly available at https://
xenabrowser.net/datapages/?cohort=TCGA%20TARGET%
20GTEx. Specifically, this research utilizes the gene expression
RNASeq data, encompassing a total of 19,120 samples and
providing the logtransformed expression values for 58,581
genes.
2.2 | Preprocessing
The data set is refined based on phenotype information sourced
from the UCSC Xena Project. A filter is implemented, which
keeps only the samples identified with a primary site of Breast
and categorized under the sample types Normal Tissueor
Primary Tumor.Additionally, the analysis is restricted to the
gene expression profiles of proteoglycans. Consequently, the
refined data set encompasses expression data for 43 genes
across 1271 samples. The data set is divided into 179 normal
tissue samples, all sourced from GTEx, and 1092 tumor tissue
samples, all derived from TCGA. Furthermore, the tumor
samples are augmented with clinical data detailing the patho-
logic stage of the tumor tissue samples. Table 1displays the
distribution of samples across different stages.
TABLE 1 | Number of samples per pathological stage.
NA Stage I Stage II Stage III Stage IV Stage X
12 182 617 248 20 13
2of18 Proteoglycan Research, 2025
2.3 | Gene Expression Analysis
Here, we present an analysis that involves detailed gene ex-
pression profiling and logistic regression modeling to identify
and validate a gene signature capable of distinguishing between
normal and tumor breast tissues, using advanced statistical
methods and machine learning techniques. This analysis was
developed and performed with initial criteria in place to select
only the most pertinent genes from the data set. Genes were
only considered for downstream analysis if (1) the log
2
fold
change of the gene in the breast carcinoma group was > 2 or
<2, and (2) the pvalue of this comparison was > 0.001.
For each gene, the log
2
fold change is calculated as the logarith-
mically normalized difference between the mean expression values
of cases (TCGA) and controls (GTEx). Following the DESeq. 2
analytical framework [20], pvalues were derived using a Wald test
[21] applied within a simple linear regression model. Then, these p
values were adjusted to control for multiple testing errors using the
BenjaminiHochberg procedure [22]. We note that while DESeq. 2
was used as an analytical framework, this package was not used to
perform the analysis. Data retrieved from the UCSD Xena Project is
logtransformed upon download and reverting these transformed
data to raw counts required DESeq. 2 compatibility, thereby intro-
ducing error and consequently decrease the precision of down-
stream analysis. The code was developed in R (https://www.r-
project.org/). We used the Wald test function, waldtest(),providedin
the package lmtest [23]. pvalues adjustment was carried out
through the function p.adjust(),includedinthebasepackagestats,
specifying the parameter method =BHin the function call.
Of the 43 proteoglycans evaluated in this analysis, only 11
proteoglycans meet these stringent criteria, namely GPC3, DCN,
PTPRZ1, EPYC, ACAN, PODN, SDC1, TGFBR3, OGN, ASPN,
and PRELP. Notably, ACAN, which is predominantly expressed
in cartilaginous tissues and plays a key role in the structural
integrity of cartilage, also appears in this list. Given its usual
tissuespecific expression and the focus of this study on breast
cancer, this finding was unexpected. To refine our analysis
further, we applied a meanbased filter [24] excluding genes
with a mean logcounts across all samples of 4.45435 or less.
This threshold, considering the data is logtransformed as
X' = log
2
(X+1), corresponds to a minimum average gene count
of approximately 85, effectively excluding ACAN from the data
set. After applying this filter, we narrowed the focus of our
study to the following proteoglycans: GPC3, DCN, PODN, SDC1,
TGFBR3, OGN, ASPN, PRELP, which became the subject of
further investigation.
2.4 | Gene Signature for Breast Cancer Detection
The objective of this analysis was to identify a gene signa-
ture for breast cancer that would effectively differentiate
between normal (class 0) and tumor (class 1) tissues. To
achieve this, we implemented a Recursive Feature Elim-
ination (RFE) procedure [25] to isolate the minimal set of
proteoglycans that yields the highest classification accuracy.
This approach involved evaluating all possible combinations
of the eight previously identified proteoglycans. For each
combination, we used a Logistic Regression Model, a binary
classification method that models the probability of an
outcome using predictor variables, producing interpretable
estimates for risk prediction. This model was then fitted for
each combination and the accuracy of each model was
assessed using LeaveOneOut CrossValidation (LOOCV).
LOOCV is a model evaluation method where each data point
is used once for testing, with the rest used for training. This
maximizes data use and provides an unbiased performance
estimate. The optimal model demonstrated a validation
accuracy of 0.9984 and a training accuracy of 1.0. This
model includes the following five proteoglycans: ASPN,
PODN, SDC1, PRELP, and DCN. The coefficients of these
proteoglycans, which contribute to their discriminative
power in the logistic regression model, are detailed in
Table 2.
The exceptionally high accuracy achieved underscores the
efficacy of these five proteoglycans in distinguishing between
normal and tumor tissues. Consequently, this gene signature
holds promise as a reliable tool for breast cancer detection. The
positive value of the intercepts indicates a bias of the model
towards class 1 (tumor tissue). This bias is anticipated, given the
unbalanced nature of the data set, where class 1 is approxi-
mately six times more prevalent than class 0.
2.5 | Cancer Stage Analysis
Given the promising findings from the previous analysis, it
has been extended to explore the influence of cancer stage
on proteoglycan expression. Unlike the initial analysis
which focused on distinguishing between normal and tumor
tissues, here the cancer stage serves as the dependent vari-
able. Therefore, samples from normal tissues and those
lacking a stage assignment were excluded. Additionally,
owing to the scant number of samples in Stage IV and Stage
X(Table1), we have arbitrarily pooled these samples into
Stage III group. Despite the success of previous models, the
Ordered Logistic models fitted through the RFE only
achieved an Area Under the Receiver Operating Charac-
teristic curve (AUROC) of 0.5035587 during training, sug-
gesting performance akin to a trivial classifier that assigns
all samples to the most prevalent class. Consequently, this
analysis suggests that there are no significant variations in
the expression of proteoglycans across different tumor
stages and a gene signature for the severity of breast cancer
could not be derived.
TABLE 2 | Coefficients of logistic regression for selected proteoglycans. Negative coefficients indicate downregulation of the corresponding gene
in tumor tissues relative to normal tissues, while positive coefficients indicate upregulation.
Intercept ASPN PODN SDC1 PRELP DCN
121.42 59.27 49.12 57.50 22.01 23.27
3of18
2.6 | Cancer Type Analysis
Given the heterogeneity of breast cancer, tumor type must be a
primary consideration in studies of gene expression and disease
progression. We therefore conducted a stratified analysis of
gene expression profiles, gene signatures, and cancer stage
across the most prevalent breast tumor subtypes. In our data
set, the most represented tumor types were infiltrating duct
carcinoma (774 samples) and lobular carcinoma (201 samples);
the remaining types presented limited sample sizes (with the
largest being 28 and most others fewer than 9), rendering them
unsuitable for robust analysis.
2.6.1 | Infiltrating Duct Carcinoma
The data set includes a total of 953 samples for this subtype.
Applying initial filtering criteria based on log
2
foldchange and
the pvalue thresholds, we retained the following genes: GPC3,
DCN, PODN, SDC1, TGFBR3, OGN, ASPN, and PRELPwhich
is identical to the gene set identified when analyzing all tumor
types collectively. This overlap probably is because the in-
filtrating duct carcinoma is the most prevalent type of tumor in
the data set.
For infiltrating duct carcinoma, a refined gene signature com-
prising ASPN, PODN, and DCN was identified, achieving both
validation and training accuracies of 1.0. This result indicates
an optimal separation of tumor and normal tissues, suggesting
these genes may serve as reliable markers for this carcinoma
subtype. Interestingly, this signature is a subset of the broader
data set's signature, suggesting that a minimal yet distinct gene
set can retain high discriminatory power within this subtype.
Regression coefficients for this signature (Table 3) exhibit the
same directional influence (positive or negative) as those from
the full data set signature, highlighting regulatory consistency
across tumor contexts.
Stage analysis, however, reveals that proteoglycan expression
alone does not effectively distinguish between cancer stages for
this subtype; the AUROC for stage classification remains low at
0.502, indicating performance no better than random chance.
This finding suggests that while these genes are effective for
tumor detection, they are limited in capturing the complexity of
tumor progression in infiltrating duct carcinoma.
2.6.2 | Lobular Carcinoma
For lobular carcinoma, the data set contains 380 samples. Fol-
lowing similar filtering criteria, we retained a gene set com-
prising GPC3, EPYC, CHAD, SDC1, TGFBR3, ASPN, PRELP,
and BGN. This set introduces three additional genesEPYC,
CHAD, and BGNwhile excluding DCN, PODN, and OGN.
These differences highlight potential subtypespecific expres-
sion patterns, reflecting the molecular diversity within breast
cancer and possibly indicating unique pathobiological roles for
these genes in lobular carcinoma.
The gene signature specific to lobular carcinoma includes
GPC3, ASPN, PRELP, SDC1, and EPYC, achieving validation
and training accuracies of 0.982 and 1.0, respectively. This high
performance suggests a robust separation capability between
tumor and normal tissues, supporting the potential utility of
these genes as subtypespecific biomarkers. Of note, three
genesASPN, PRELP, and SDC1appear in both the lobular
carcinoma and full data set signatures, with consistent regres-
sion coefficient signs (Table 3), reinforcing their importance
across multiple breast cancer contexts (Table 4).
Stage classification for lobular carcinoma achieved an AUROC
of 0.587, which, while superior to that of the full data set and
infiltrating duct carcinoma, remains insufficient as a reliable
stage classifier.
The limited predictive capacity of proteoglycans for stage,
despite high accuracy in tumor detection, confirms that addi-
tional factors or markers beyond proteoglycan expression may
be needed to capture the intricacies of tumor progression in
lobular carcinoma, as well as in infiltrating duct carcinoma.
Further studies incorporating larger data sets and additional
biomarkers could improve stage differentiation for this subtype.
2.7 | Protein Crystal Structure Prediction and
Generation
To provide visual characterization of several proteoglycans
included in our breast carcinoma signature, we utilized previ-
ously published knowledge and stateoftheart artificial intel-
ligence prediction tools to create crystal structures for the
SLRPs. While the structure of several SLRPs have been previ-
ously determined experimentally, it was necessary to predict the
structures of others, which we accomplished using the Alpha-
Fold Database driven by Google's DeepMind Artificial Intelli-
gence. This tool was created as part of the European Molecular
Biology Laboratory's European Bioinformatics Institute [26, 27].
We were able to generate crystal structures highlighting pri-
mary and secondary structures of our SLRPs of interest using
the computerized crystal structure generation software PyMol
(The PyMOL Molecular Graphics System, Version 1.2r3pre,
Schrödinger LLC).
3 | Results and Discussion
3.1 | Gene Expression Analysis
Using the previously described data from TCGA and GTEx,
facilitated by the UCSB Xena platform, mRNA levels of 11
proteoglycans were found to be differentially expressed by a log
2
foldchange factor > 2 in breast carcinoma when compared with
TABLE 3 | Coefficients of logistic regression for Infiltrating duct
carcinoma.Negative coefficients indicate downregulation of the cor-
responding gene in tumor tissues relative to normal tissues, while
positive coefficients indicate upregulation.
Intercept ASPN PODN DCN
148.78 126.15 120.97 69.84
4of18 Proteoglycan Research, 2025
healthy controls. We determined that PTPRZ1, OGN, DCN,
PODN, GPC3, TGFBR3, and PRELP were all significantly
downregulated in breast carcinoma, whereas ASPN, ACAN,
EPYC, and SDC1 were all found to be upregulated (Figure 1).
The 11 proteoglycans identified in our differential expression
analysis were further scrutinized to determine their feasibility
and possible utility as prognostic markers for breast carcinoma.
Through comparison of basal mRNA expression levels of each
of these proteoglycans with their expression levels in breast
carcinoma, we determined that 10 of the identified proteogly-
cans were fit for further evaluation. ACAN was excluded from
downstream analysis on account of its extremely low levels of
expression in both healthy breast tissue, and breast carcinoma,
which cast doubt on both its biological validity and reliability as
a prognostic marker. From the 10 remaining proteoglycans, a
novel machine learning model generated a genetic signature
from a combination of 5 of these which was able to accurately
predict the presence or absence of breast carcinoma. Prediction
of carcinoma, which has a success rate greater than 99%, is
based on the combinatorial upregulation or downregulation of
ASPN, PODN, SDC1, DCN, and PRELP as described in Table 2.
Interestingly, the trends of upregulation and downregulation
associated with each of these respective proteoglycans align
with overarching trends observed when investigating each of
them individually, with upregulation of ASPN and SDC1
(Figure 2A) and downregulation of PODN, PRELP, and DCN
(Figure 2B) helping to predict disease. We must note, however,
that this may not always be the case, and we cannot conclude it
to be the case for other proteoglycans not named in this model.
4 | Machine Learning Model and Differentiation
of Tumor Stages
While the machine learning model that was employed to con-
struct the combinatorial proteoglycan signature of SDC1, ASPN,
DCN, PRELP, and PODN to identify difference between normal
breast tissue and cancerous tissue was able to do so with an
accuracy of over 99%, this model struggled to make distinctions
between samples of breast carcinoma when they were evaluated
on the basis of diagnostic grade of disease. With an AUROC of
0.5035587 during training, the model was unable to differentiate
between stages of disease with much more accuracy than flip-
ping a coin. Therefore, we interrogated the TCGA data set to
explore the possibilities of why the model could be so effective
in preliminary diagnostic determination, but ineffective with a
more granular task. We evaluated breast carcinoma samples
that had been labeled within the TCGA data set as stage I, stage
II, or stage III. We found that there was a significant difference
in mRNA expression of all five proteoglycans used by
the machine learning model compared with normal breast tis-
sue samples.
We further discovered that within the breast carcinoma samples
available through TCGA, the expression of our five proteogly-
cans of interest was not significantly different between stages
of disease for four out of our five candidates. SDC1 and ASPN
mRNA were significantly upregulated in each stage of disease
in samples of breast carcinoma graded as stages IIII, however
no differences were present between different stages of disease
(Figure 3A). While DCN and PRELP mRNA were significantly
downregulated in each stage of breast carcinoma when com-
pared with normal tissue samples, no differences were detected
between stages of disease (Figure 3B). PODN expression was
significantly different between stage I and stage II, and stage II
and stage III breast carcinoma respectively, however, we did not
determine a significant difference in PODN expression between
stage I and stage III within the data set (Figure 3B). This sug-
gests that the model's inability to differentiate disease stages
could be due to the lack of perceptible differences in mRNA
expression of the proteoglycans used. Therefore, it would be
beneficial to return to this model as the TCGA repository of
breast carcinoma samples continues to grow, and the robust-
ness of the data set continues to increase.
5 | Syndecan1
Syndecan1(SDC1) with its ~33 kDa protein core, is the first of
its name amongst the syndecan family of four homologous type
I transmembrane proteoglycans. Its structure is characterized
TABLE 4 | Coefficients of logistic regression for lobular carcinoma. Negative coefficients indicate downregulation of the corresponding gene in
tumor tissues relative to normal tissues, while positive coefficients indicate upregulation.
Intercept GPC3 ASPN PRELP SDC1 EPYC
2.56 3.50 4.76 3.51 6.84 3.31
FIGURE 1 | Differential expression of proteoglycans in breast car-
cinoma. Gene expression data for 42 proteoglycans was compared
between normal breast tissue samples obtained from GTEx and breast
carcinoma samples obtained from TCGA. Proteoglycans with a log
2
foldchange > 2 and a p< 0.001 are named in the volcano plot, with
downregulated genes shown in blue, upregulated genes shown in red,
and all genes that did not meet our criteria for significance going
unnamed with gray icons. Dashed lines represent intersecting limits of
log
2
foldchange = |2| and p= 0.001.
5of18
by an ectodomain with distal heparan sulfate (HS) chains and
chondroitin sulfate (CS) chains located nearer to the plasma
membrane [28]. It has a hydrophobic transmembrane domain
and a Cterminal cytoplasmic domain containing two conserved
domains interspersed by a variable region. SDC1 has been
identified as having various roles in cellcell and cellmatrix
interactions, and within the context of breast cancer, its com-
plex amalgam of effects have been shown to influence cell
adhesion and migration, tumor growth and progression, tumor
angiogenesis, and prognosis [5, 29, 30].
The ectodomain of SDC1 regulates matrixdependent signaling
in breast carcinoma through interaction with α
v
β
3
integrin,
such that an increased presence of cellsurface SDC1 is associ-
ated with increased epithelial cell invasion during early
tumorigenesis [3133]. SDC1 has been implicated in multiple
signaling pathways related to cell adhesion and migration,
including the Wnt/FGF pathway. In multiple breast cancer cell
types, SDC1 knockdown significantly inhibited various gene
products of Wnt and Fibroblast Growth Factor (FGF) signaling,
suggesting that SDC1 may mediate tumor cell migration
through these pathways [34]. The PI3K/Akt pathway has also
been implicated in SDC1 behavior, whereby cleavage of SDC1
was shown to enhance breast carcinoma cell proliferation via
increased SUMOylation of Akt [35]. Acting as a membrane
bound coreceptor for FGF Receptor, SDC1 activates FGF2
mediated MAPK signaling through induction of MAPK phos-
phorylation, ultimately leading to breast cancer cell prolifera-
tion. Unsurprisingly, SDC1 knockdown inhibits MAPK
signaling [36]. Additionally, SDC1 plays a role in IL6/JAK/
STAT3 signaling, such that SDC1 silencing inhibits IL6 sig-
naling, suggesting that overexpression of SDC1 activates this
important pathway for breast cancer cell proliferation and
migration [37].
Despite the evergrowing arsenal of knowledge regarding the
role of SDC1 in breast cancer, it is not unilaterally agreed upon
whether increased or decreased SDC1 expression indicates a
more positive prognosis. Some studies have shown increased
SDC1 protein expression to be associated with worse patient
outcomes and a more aggressive disease phenotype [3840].
Others indicate the opposite, stating that decreased SDC1 ex-
pression is associated with poorer outcomes [37, 41]. However,
more recent large scale, bioinformatic analyses of largescale
public datasets in triplenegative breast cancer has given sup-
port for the former assertion that high SDC1 expression is
FIGURE 2 | Differential expression of machine learning identified proteoglycans in breast carcinoma. (A) ASPN and SDC1 were found to be
upregulated in breast carcinoma samples obtained from TCGA compared with normal tissue samples obtained from GTEx. (B) PODN, PRELP, and
DCN were found to be downregulated in breast carcinoma samples obtained from TCGA compared with normal breast tissue samples from GTEx.
6of18 Proteoglycan Research, 2025
associated with poorer rates of both overall survival and disease
free survival [42, 43]. These recent findings are corroborated in
this study, in which we also propose a novel prognostic utility
for SDC1.
6 | Asporin
Asporin, originally purified from human articular cartilage [44],
is a class I SLRP [45, 46] as it is closely related to decorin and
biglycan. In addition of containing the conserved Nterminal
CX
3
CXCX
6
C pattern typical of decorin and biglycan [47],
asporin harbors a unique Nterminal stretch of aspartic acid
residues thus its eponym [44, 48] (Figure 4). Unlike decorin and
biglycan, asporin contains no consensus glycosaminoglycan
attachment sites (SerGly) at its amino terminus. However, it is
possible that the polyaspartate region due to its polyanionic
nature would modulate the interaction of asporin with growth
factors and collagen as shown for other SLRPs, with a function
analogous to that of the glycosaminoglycan chains of decorin
and biglycan, the closest family members [44]. Indeed, asporin
competes with decorin in collagen binding and promotes oste-
oblast collagen mineralization [49]. Due to its ubiquitous ex-
pression, asporin has been implicated as an activator of invasion
in schirrhous gastric carcinomas [50], as well as growth and
migration of gastric cancers cells [51, 52]. Moreover, stromal
derived asporin is a biomarker associated with prostate cancer
progression [53], bladder cancer where asporin levels correlate
with the degree of malignancy [54]. Prostate tumor allografts in
Aspn
/
mice show a decreased propensity toward pulmonary
metastases [55] suggesting that asporin could promote meta-
static progression. A recent report has convincingly corrobo-
rated these findings by discovering a reciprocal interplay
between asporin and decorin, with asporin acting as a gastric
cancer promoting gene and decorin as a tumor repressor [56].
The proposed mechanism of action involves and asporin/TGFβ
interaction suggesting that preventing this interaction, together
with overexpression of decorin would counteract the gastric
cancer growth and invasion [56]. In line with these results, is
the finding that asporin can repress gastric cancer apoptosis by
activating LEF1mediated gene transcription independent of β
catenin [57], suggesting that asporin could act as a prognostic
marker in this malignancy [56]. Additional evidence for broad
protumorigenic activity of asporin derives from its ability to
reprogram gastric cancer cells to acquire resistance to oxidative
stress [58], and its interaction with HER2 in promoting thyroid
cancer metastases [59].
The multifaceted roles of asporin are challenging and somewhat
controversial as asporin levels have also been linked to anti-
tumorigenic activity [60, 61]. While asporin acts as an oncogene
in pancreatic, colorectal, gastric, and prostate cancers, and some
types of breast cancer, there are reports that asporin can act
as a tumor suppressor gene in triplenegative breast cancer [62].
Intriguingly, high asporin expression associates with a
significantly better relapse free survival rate in patients with
FIGURE 3 | Evaluation of differential expression of proteoglycans in breast carcinoma by stage. (A) SDC1 and ASPN mRNA were found to be
upregulated in all stages of breast carcinoma samples obtained from TCGA compared with healthy breast tissue samples. There were, however, no
significant differences in expression of either of these mRNA transcripts between samples graded by stage of disease. (B) DCN, PRELP, and PODN
mRNA were found to be downregulated in all stages of breast carcinoma samples obtained from TCGA compared with healthy breast tissue samples.
When breast carcinoma samples were evaluated by stage of disease, there was no differential expression determined for DCN or PRELP mRNA,
however, we identified differences in PODN expression between stage I and stage II carcinoma and between stage II and stage III carcinoma.
*Significance is defined by p< 0.05. Furthermore, **significance is defined as p< 0.01 and ***significance is defined by p< 0.001.
7of18
lowgrade tumors but is significantly worse in breast cancer
patients with stage 3 tumors [63]. Our results show in an
unbiased way that asporin gene expression correlates with a
promalignant role in breast cancer independently of the stage
of the tumor (Figure 3A). If confirmed at the protein level,
upregulated syndecan1 and asporin protein levels could be
therapeutically challenged by proteolysis targeting chimera
(PROTAC) protein degraders [64]. This novel approach in
cancer treatment harnesses the body's natural protein degra-
dation machinery to target specific proteins for destruction. As
heterobifunctional molecules, PROTACs function by recruiting
and binding the protein of interest and E3 ubiquitin ligase, hence
inducing ubiquitylation of the targeted protein and its subsequent
degradation by the ubiquitinproteasome system [64]. Syndecan1
and asporin, as a transmembrane and secreted proteoglycan
respectively, could represent ideal candidates for PROTAC ther-
apy as exposed amino acidic domains could be easily identified as
targets of this innovative therapeutic modality.
7 | Decorin
Decorin is the prototype member of Class I SLRP and is composed
of 12 LRRs and a single glycosaminoglycan chain at the
Nterminus [65, 66](Figure4). Abundantly found in multiple
tissues throughout the body, decorin was named after its unique
ability to interact with and decoratecollagen fibrils to maintain,
regulate and organize their spacing and structure [67]. This
interplay with collagen is crucial for the tensile strength and
elasticity of tissues, ensuring proper tissue architecture and
homeostasis, and structural integrity of the ECM [68, 69]. The role
of decorin in cancer has been the subject of several studies in the
past 3 decades [7072], starting with colon carcinoma [7375], and
then extending to several solid tumors with diverse histogenetic
backgrounds [7695]. Decorin was subsequently described as an
active modulator of cell behavior and function through its inter-
actions with different growth factors and its role as a panreceptor
tyrosine kinase (RTK) inhibitor [9698]. Indeed, decorin defi-
ciency promotes epithelialmesenchymal transition (EMT) and
colon cancer metastasis [75]
Decorin binds to the transforming growth factor β(TGFβ) and
vascular endothelial growth factor (VEGF), sequestering them
from their respective receptors and modulating their bio-
availability in the ECM. This regulation is crucial for main-
taining tissue homeostasis and preventing excessive growth
factor signaling [99103]. One of the most studied roles of
decorin is its ability to interact with TGFβ[104, 105] and to
block TGFβ/Smad downstream signaling [106, 107], which in
general leads to suppression of fibrosis [78, 108, 109]. Decorin
has also been involved in inhibiting angiogenesis [99, 110120],
and also in promoting angiogenesis in a mouse model of corneal
injury [100]. Moreover, decorin is involved in regulating
inflammation [121, 122], bone homeostasis [123], bio-
mechanical properties of cartilage [124], mitophagy [111,
125135], and can signal through modulation of intracellular
Ca
2+
[136, 137].
Decorin interactome has expanded significantly with the dis-
covery of several RTKs that specifically bind to decorin with
nanomolar affinity, including the epidermal growth factor
receptor (EGFR), insulinlike growth factor receptor (IGFIR),
and Met receptor (HGFR), as well as the vascular endothelial
growth factor receptors 2 and 3 (VEGFR2, VEGFR3) [138141].
By binding to these receptors, decorin interferes with their
activation and downstream signaling pathways overall off-
setting their ability to sustain growth of cancer cells and stim-
ulate angiogenesis in the tumor stroma [30].
By interfering with VEGF signaling and disrupt endothelial cell
interactions with ECM components, decorin reduces en-
dothelial cell proliferation, a critical step in angiogenesis [117,
142], and induces autophagy in endothelial cells. Autophagy, a
highly conserved process involving the degradation and re-
cycling of cellular components [143, 144], is tightly regulated by
decorin. The mechanism involves decorinmediated inhibition
of Akt/mTOR signaling pathway, leading to activation of
autophagyrelated genes and formation of autophagosomes
[130, 132, 145, 146]. Overall, decorin exemplifies the complexity
of ECM proteins in regulating cellular processes through in-
teractions with RTKs, modulation of autophagy, and influence
FIGURE 4 | Structure of the four SLRPs linked to the breast carcinoma signature. Using PyMol, published knowledge regarding proteoglycan
structure, and the AlphaFold structure predictor driven by Google's DeepMind artificial intelligence tool, we created crystal structures for each of the
four SLRPs included in our machine learning breast carcinoma signature. The primary structure of each SLRP is shown in blue, with secondary
structures delineated by alternate colors: αhelices are marked in red and yellow arrows are used to identify βpleated sheets.
8of18 Proteoglycan Research, 2025
on angiogenesis. Its diverse interactions with the ECM, growth
factors, and immune mediators highlight its importance in
normal physiological processes beyond its structural role.
Decorin affects the growth of various tumors [76, 77], and we
further discovered that there is cooperative action of germline
mutations in Dcn
/
and the tumor suppressor p53
/
that leads
to an acceleration of lymphoma tumorigenesis [147]. Adminis-
tration of recombinant decorin to these lymphoma cells derived
from the doublymutant Dcn
/
;p53
/
mice significantly retarded
their growth further, supporting a potential therapeutic role for
decorinintumorsuppression.Inaninvivotriplenegative or-
thotopic breast cancer model, we found that systemic adminis-
tration of recombinant decorin modulated the differential
expression of 374 genes within the stromal compartment of the
tumor xenograft associated with immunomodulatory responses,
cellular adhesion and tumor suppressive gene properties [148]. In
inflammatory breast cancer (IBC) cells overexpression of DCN
markedly decreased migration and invasion, and inhibited tumor
growth and metastasis in IBC xenograft mouse models through
inhibition of the EGFR/ERK signaling pathway and decorin
mediated autophagic degradation of Ecadherin [79]. Moreover,
the inhibition of colon carcinoma growth and migration by dec-
orin also involves modulation of Ecadherin levels [83]. The
therapeutic benefits of systemic delivery of decorin to different
in vitro and in vivo tumor models dictate the need for a deeper
investigation of its the expression level as they could be associated
with prognosis and survival rate estimate of cancer patients
[149, 150].
However, when examining decorin expression in breast cancer
tissue compared with normal breast tissue, studies have shown
contrasting results depending on the specific subtype of breast
cancer and the stage of the disease [88]. In normal breast tissue,
decorin expression is typically wellmaintained and plays a
crucial role in regulating the ECM and growth factor signaling,
as described earlier [151]. Decreased expression of decorin in
breast cancer has been associated with poorer prognosis,
increased tumor growth, and enhanced metastatic potential.
8 | PRELP
A member of the Class II SLRP genes is PRELP (Proline/
arginineRich End Leucinerich repeat Protein), also known as
prolargin, and was originally cloned from human articular
chondrocytes [152] (Figure 4). Its eponym is based on its unique
Nterminal domain which, in contrast to asporin, harbors sev-
eral basic amino acid residues such as proline, arginine and
leucine [152]. The Nterminus domain of PRELP binds heparin
and heparan sulfate; as fibroblasts interact with PRELP in an
heparindependent manner, it has been proposed that PRELP
can function as linker also between the ECM and cell surface
proteoglycans [153], acting as an important regulator of cell
adhesion [154]. Indeed the Nterminus of PRELP is involved
binding to perlecan and collagen [155], potentially functioning
as a basement membrane anchor [156].
An interesting biological role of PRELP is its ability to directly
inhibit all complement pathways by binding C9 and thereby
preventing the formation of the complement membrane attack
complex [157]. Notably, the Nterminal basic domain of PRELP
functions as a cell specific inhibitor of NFκB signaling and
impairs osteoclastogenesis [158]. Specifically, a recombinant
form of the basic Nterminal region of PRELP reduces osteoclast
number and activity in ovariectomized mice, underlying its role
in skeletal remodeling [158]. Additional functions of PRELP
include its role as a natural TGFβantagonist and inhibitor of
fibrosis [159], and an interaction with IGFIR and lowaffinity
nerve growth factor receptor (p75NTR) which results in growth
inhibition of A549 lung carcinoma cells [160]. A recent report
utilizing Prelp
/
mice has shown that PRELP secreted by
mural cells protects the blood brain barrier by promoting the
integrity of endothelial cells [161].
There are only few reports of PRELP association with various
cancer types, although the majority of independent studies
point to a tumor suppressor role in ovarian cancer [162], oral
squamous cell carcinomas [163] and colon cancer cells [164].
Moreover, there is a positive correlation between PRELP ex-
pression and survival in hepatocellular [165] and pancreatic
carcinoma patients [166]. A variant of PRELP was found to be
uniquely expressed in chronic lymphocytic leukemia cells [167],
suggesting that there might be an organand tissuespecific
bioactivity. PRELP is an endogenous inhibitor of bladder cancer
initiation and progression [168], and by regulating cellcell
adhesion and EMT inhibits retinoblastoma progression [169].
Loss of PRELP expression in melanomas correlates with tumor
escape and enhanced aggressiveness [170]. There is also evi-
dence that PRELP inhibits colon cancer progression by sup-
pressing EMT and angiogenesis [171]. In contrast to these
studies, PRELP has been shown to promote EMT in colon
cancer and to stimulate growth and invasion of colon cancer
cells [172]. So, much more research needs to be done to clarify
these contrasting results.
9 | Podocan
Podocan is a class V noncanonical SLRP that was originally
identified from sclerotic glomerular lesions in experimental
human immunodeficiency virusassociated nephropathy [173]
(Figure 4). This SLRP has been implicated in the suppression of
endothelial cell growth, proliferation, and migration, particu-
larly in the kidney, smooth muscle, and adipose tissue. Func-
tional analysis of podocan suggests that his proteoglycan
contains Nlinked oligosaccharides and that it binds to Type I
collagen [174]. It has been shown that suppression of cell
growth and migration by podocan is associated with changes in
p21 and Rho activity [174]. Additionally, inhibition of podocan
expression by MiR31805p promotes human bladder smooth
muscle cell proliferation [175]. Unsurprisingly, podocan is
highly expressed in vascular smooth muscle cells after injury,
and it has been shown that podocandeficient mice demonstrate
increased arterial lesion formation compared with their wild
type counterparts, identifying podocan as a negative regulator
of smooth muscle cell proliferation [176]. Both in vivo and
in vitro, podocandeficient mice demonstrate increased smooth
muscle cell proliferation through activation of the Wnt/β
catenin pathway [177]. Podocan modulation of smooth muscle
cell proliferation and differentiation has been shown to be a
direct result of podocan interacting with Wnt and enhancing
9of18
Wnt/bcatenin signaling [178]. Most recently, podocan has been
shown to play a role in postinjury smooth muscle cell differ-
entiation by binding with TGFβ1 thereby inhibiting its down-
stream signaling pathway [179].
Podocan has been characterized in pathological processes such
as kidney diseases and recently colorectal cancer [180, 181].
Podocan was shown through microarray analysis to be a sig-
nificant biomarker for transplant glomerulopathy and have
been previously shown to be significant to processes involving
the glomerular basement membrane [173, 180]. It was also
found to be correlated with diabetic nephropathy, as podocan
mRNA levels negatively correlated with the urinary albuminto
creatine ratio, a marker for glomerular injury in models of
diabetic nephropathy [182]. In colorectal cancer, podocan has
been identified as having locationdependent differential ex-
pression with qRTPCR revealing decreased podocan mRNA
expression in tumors compared with normal colon tissue in
patient samples of leftsided colon cancer, but not rightsided
colon cancer [181]. Podocan has not yet been implicated as a
potential biomarker in any other cancers, and presently, no
studies exist which explore podocan within the context of a
transcriptomic analysis derived from a largescale publicly
available data set. This study is the first to identify a potential
role of podocan as a prognostic marker in breast carcinoma.
10 | Serglycin
Serglycin (SRGN), initially discovered as a small chondroitin
sulfate proteoglycan secreted by a rat yolk sac tumor, was
shown to be highly enriched in Ser and Gly amino acids and to
be associated with the cell surface [183]. Subsequent cloning of
its protein core [184] and gene [185] showed a protein
of ~10 kDa with a central region of ~50 amino acids composed
of alternating Ser/Gly residues, hence its eponym [184]
(Figure 5A). A number of subsequent studies have clearly
shown that serglycin is the main carrier of heparin and
is directly involved in packaging proteases in mast cell granules
[186, 187] and regulating the maturation of mast cell granules
[188, 189]. Serglycin is also expressed in embryonic stem
cells [190], uterine decidua [191], neutrophils [192], myeloma
cells [193, 194], and has been proposed to be a potential marker
for acute myeloid leukemia [195]. Indeed, serglycin plays a crit-
ical role in the protein cargo loading of exosomes produced by
multiple myeloma cells [196].
An interesting discovery was the identification of serglycin as a
main product of endothelial cells where it was localized within
intracellular vesicular compartments codistributing with tis-
sue plasminogen activator [197]. It was subsequently shown
that serglycin is secreted in polarized human endothelial cells
mainly in the apical region and is associated with chemokine
secretion [198] and directly involved in the inflammatory
response [199]. Indeed, serglycin is a major proteoglycan
secreted by macrophages and is involved in regulating TNFα
secretion in response to proinflammatory stimuli [200].
Recently, serglycin has been implicated as a protumorigenic
factor in various solid tumors including: head and neck carci-
noma [201], colorectal cancer [202], hepatocellular carcinoma
[203], lung adenocarcinoma [204], and glioblastomas [205, 206],
including a role as a potential biomarker for glioblastoma pro-
gression [207]. Moreover, serglycin produced by gastric carci-
noma cells promotes secretion of the proinflammatory
interleukin 8 in cancer associated fibroblasts [208].
Regarding the role of serglycin in breast cancer, there are sev-
eral recent scientific reports underlying and corroborating its
role as a protumorigenic and prometastatic factor [209]. An
elegant study using genetic ablation of Srgn gene in a MMTV
PyMTdriven mouse breast carcinoma model has shown a
complete suppression of lung metastases without affecting the
growth of the primary tumors [210].
Serglycin is also highly expressed by infiltrating immune cells in
breast cancer microenvironment [211], and is involved in
epithelialmesenchymal transition (EMT) evoked by TGFβ
[211] as well as promoting a more aggressive phenotype in
breast cancer [212, 213]. Mechanistically, serglycin may protect
breast cancer cells from complement attack thereby supporting
their survival and expansion [212]. Collectively, these data
suggest that serglycindependent mediators may represent
potential drug targets to prevent pulmonary metastases in pa-
tients with breast cancer.
Utilizing the online data exploration tool, Xena, developed by the
University of California Santa Cruz, we analyzed serglycin (SRGN)
mRNA expression in available breast carcinoma data from TCGA
and GTEx. The data available to the public suggests a different
conclusion than the one reported by the majority of experimental
findings concerning serglycin expression on the mRNA level.
These public data revealed that SRGN expression is lower in breast
carcinoma compared with normal tissue. Additionally, SRGN
levels decline as stage of disease progresses. Specific evaluation of
infiltrating ductal carcinoma revealed decreased SRGN expression
as stage of disease advances (Figure 5BD).
There are at least two scientific reports corroborating our
findings, showing lower levels of serglycin expression in breast
carcinoma visànormal tissue [211, 214]. Using the Tissue
Immune Estimation Resource to explore TCGA breast carci-
noma data showed lower SRGN levels in breast tumor tissue
compared with normal tissue [214]. While this result is in
opposition to the body of in vitro evidence reported elsewhere,
separate analysis of breast cancer subtype data from TCGA and
the OSLO2 study support the finding that SRGN expression is
lower in tumor than normal breast tissue [211]. Furthermore, it
has been shown that trends in SRGN expression vary between
different subtypes of breast cancer based on public data and
even amongst different breast cancer cell lines of variable
aggressiveness studied in vitro [211, 213, 215]. Indeed, qPCR of
SRGN expression in basallike subtypes of variable phenotype,
from epithelial to mesenchymal, revealed higher SRGN ex-
pression in mesenchymal subtypes compared with epithelial or
partial epithelial cancer cell lines, however, single cell RNA
sequencing (scRNAseq) of 18 patient samples of variable sub-
types revealed very low SRGN expression specifically in cancer
cells, but much higher expression shown in Tcells, dendritic
cells, and mast cells [211]. Interestingly, SRGN immunogeni-
cally evokes IL8/CXCR2 signaling to promote epithelial mes-
enchymal transition in breast cancer cells of less aggressive
phenotype, to then have more SRGN expression in aggressive
10 of 18 Proteoglycan Research, 2025
breast cancer celltypes [213]. Also interesting is im-
munostaining of aggressive breast tumor sections showing
granular cytoplasmic localization of SRGN, suggesting local-
ization of SRGN into secretory granules [215]. Additionally,
KaplanMeyer analysis suggests a differential, timebased
association of SRGN expression with overall survival, such
that survival is lower in patients with low tumor SRGN ex-
pression up to a time point of about 3500 days, at which point,
survival becomes lower for patients with high SRGN expression
[211, 214]. This scRNAseq data along with the aforementioned
in vitro findings invite questions of whether SRGN is truly
contributing to survival through its effect on cancer cells
themselves, or whether outsidein signaling in the tumor
microenvironment ought to be a greater focus of its influence
on disease outcomes. While largescale datasets offer the sta-
tistical advantage of reducing bias, it is possible that the
heterogeneous nature of bulk sequencing models may muddle
our understanding of SRGN. Additional scRNAseq experi-
mentation focused on SRGN in various breastcancer subtypes
may elucidate discrepancies between bioinformatic and in vitro
approaches, since these analyses provide celltypespecific an-
swers that bulk RNA sequencing cannot accommodate. The
variability of these results suggests that further exploration of
the data directed toward disease progression, aggressiveness,
and tumor microenvironment are necessary to reconcile con-
flicting results regarding serglycin in breast cancer. With careful
consideration given to the body of experimental findings sug-
gesting serglycin itself, or serglycin dependent mediators, to be
protumorigenic, no study has thus far evaluated the publicly
available data to validate these findings. The integration of
proteoglycan expression profiles, their biological functions, and
clinical outcomes has the potential to guide the development of
FIGURE 5 | Structure and under expression of SRGN in breast carcinoma malignancy. (A) Crystal structure of serglycin rendered using
AlphaFold and PyMol. Primary structure is shown in blue, while helices are shown in red and βpleated sheets are shown with yellow arrows. (B)
Breast carcinoma samples obtained from TCGA data demonstrate lower levels of SRGN mRNA expression compared with normal tissue samples
obtained from GTEx. (C) TCGA samples categorized by diagnostic stage of disease demonstrate lower SRGN expression with disease progression. (D)
Specifically within TCGA samples of infiltrating ductal carcinoma, relative SRGN levels decrease with disease progression. *** significance is defined
by p< 0.001.
11 of 18
precision diagnostic and potentially therapeutic tools that may
enhance the prognosis of not only breast cancer but also of
other solid tumors.
11 | Conclusions
Analyzing publicly available data sets, we discovered a unique
gene expression signature in breast cancer composed of five dif-
ferentially modulated proteoglycans: Syndecan1andasporin
(upregulated) and decorin, PRELP and podocan (downregulated).
Although there are 4550 proteoglycan encoding genes, our
machinelearning analysis has revealed that only this small cohort
or proteoglycans is meaningfully regulated in breast cancer.
Notably, four out of five are SLRPs, primarily stromal derived
constituents, with syndecan1 being the only cell associated
proteoglycan among the cohort. We should point out that
syndecan1 can be shed and its ectodomain is also found in the
stroma [216221]. We propose the five proteoglycan group as a
new biomarker set for the diagnosis, prognosis and potential
treatment of breast cancer patients.
We also performed a model literature review regarding the five
proteoglycans included in the breast carcinoma gene expression
signature and serglycin, a small chondroitin sulfate proteoglycan
that has also been studied in the context of breast cancer. We
found significant discrepancies when comparing published ex-
perimental results regarding protein and gene expression with
TCGA data. While most sources in the literature have published
experimentally determined findings suggesting that increased
SRGN expression is associated with breast cancer development,
analysis of TCGA data suggests the opposite association. This
striking difference demonstrates the value of public information in
efforts to validate experimentally determined data. Additionally,
these results highlight the need to return to validation analyses in
the future as public data sets like TCGA continue to include new
patient samples and grow the statistical power of the data set. As
these public datasets grow to include single cell, and eventually
even single nucleus sequencing, the opportunity to validate bio-
chemical experimental data will only grow. Great contributions
canbemadetoourfieldbycomparingpublishedliteratureto
public data for the rest of the proteoglycans that have not been
assessed in the present study.
Through a therapeutic lens, our study could hold significant
translational significance. For example, a potential approach would
be to target the two upregulated proteoglycans, syndecan1and
asporin, using gene editing technology such as CRIPR/Cas9.
Another would be to target the protein cores using PROTACs,
heterobifunctional molecules harboring two active domains and an
inker that can lead to a removal specific unwanted protein. On the
other hand, systemic delivery of recombinant proteoglycans or
protein cores such as decorin, PRELP or podocan, could pave the
way for proteoglycanbased therapies that, together with conven-
tional chemotherapy, targeted therapies, or immunotherapy could
enhance treatment efficacy and overcome drug resistance.
The continuously increasing statistical power of public data-
sets should not be overlooked as a tool for cost and time effi-
cient analyses in the future. Especially as great strides are
made in the field of artificial intelligence, and novel analytical
methods like the machine learning model used in this project
become increasingly refined, the proteoglycan research com-
munity is privy to exciting avenues for further investigation.
The opportunity exists within the marriage of data and novel
technology to drive inquiry towards discovering clinically
relevant phenomena and developing innovative therapeutic
options.
Author Contributions
Simone Buraschi: conceptualization, writingoriginal draft,
writingreview and editing. Gabriel Pascal: conceptualization,
writingoriginal draft, writingreview and editing, methodology, visu-
alization. Federico Liberatore: writingoriginal draft, writingreview
and editing, methodology. Renato V. Iozzo: conceptualization, funding
acquisition, writingoriginal draft, writingreview and editing.
Acknowledgments
The original research was supported, in part by National Institutes of
Health Grants RO1 CA245311 and RO3 CA270830 (to R.V.I.).
Gabriel Pascal was supported by NIH grant 3R01CA24531104S1.
Ethics Statement
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Data Availability Statement
The authors have nothing to report.
References
1. R. V. Iozzo and L. Schaefer, Proteoglycan Form and Function: A
Comprehensive Nomenclature of Proteoglycans,Matrix Biology 42
(2015): 1155.
2. N. K. Karamanos, Z. Piperigkou, A. D. Theocharis, et al., Proteo-
glycan Chemical Diversity Drives Multifunctional Cell Regulation and
Therapeutics,Chemical Reviews 118 (2018): 91529232.
3. I. Caon, B. Bartolini, P. Moretto, et al., Sirtuin 1 Reduces Hyalur-
onan Synthase 2 Expression by Inhibiting Nuclear Translocation of NF
ΰB and Expression of the LongNoncoding RNA HAS2AS1,Journal of
Biological Chemistry 295 (2020): 34853496.
4. S. RicardBlum, R. R. Vivès, L. Schaefer, et al., A Biological Guide to
Glycosaminoglycans: Current Perspectives and Pending Questions,
FEBS Journal 291 (2024): 33313366.
5. R. V. Iozzo and R. D. Sanderson, Proteoglycans in Cancer Biology,
Tumour Microenvironment and Angiogenesis,Journal of Cellular and
Molecular Medicine 15 (2011): 10131031.
6. S. Garantziotis and R. C. Savani, Hyaluronan Biology: A Complex
Balancing Act of Structure, Function, Location and Context,Matrix
Biology 7879 (2019): 110.
7. M. K. Cowman and E. A. Turley, Functional Organization of Ex-
tracellular Hyaluronan, CD44, and RHAMM,Proteoglycan Research 1
(2023): e4.
8. E. Ruoslahti and Y. Yamaguchi, Proteoglycans as Modulators of
Growth Factor Activities,Cell 64 (1991): 867869.
9. R. V. Iozzo and L. Schaefer, Proteoglycans in Health and Disease:
Novel Regulatory Signaling Mechanisms Evoked by the Small Leucine
Rich Proteoglycans,FEBS Journal 277 (2010): 38643875.
12 of 18 Proteoglycan Research, 2025
10. L. Schaefer, Proteoglycans, Key Regulators of CellMatrix Dynam-
ics,Matrix Biology 35 (2014): 12.
11. M. Mongiat, S. M. Sweeney, J. D. San Antonio, J. Fu, and
R. V. Iozzo, Endorepellin, a Novel Inhibitor of Angiogenesis Derived
From the C Terminus of Perlecan,Journal of Biological Chemistry 278
(2003): 42384249.
12. J.F. Cailhier, I. Sirois, P. Laplante, et al., Caspase3 Activation
Triggers Extracellular Cathepsin L Release and Endorepellin Proteoly-
sis,Journal of Biological Chemistry 283 (2008): 2722027229.
13. B. P. Woodall, A. Nyström, R. A. Iozzo, et al., Integrin α2β1 Is the
Required Receptor for Endorepellin Angiostatic Activity,Journal of
Biological Chemistry 283 (2008): 23352343.
14. A. Nyström, Z. P. Shaik, D. Gullberg, et al., Role of Tyrosine
Phosphatase SHP1 in the Mechanism of Endorepellin Angiostatic
Activity,Blood 114 (2009): 48974906.
15. G. Bix, R. Castello, M. Burrows, et al., Endorepellin in Vivo: Targeting
the Tumor Vasculature and Retarding Cancer Growth and Metabolism,
JNCI: Journal of the National Cancer Institute 98 (2006): 16341646.
16. A. Goyal, N. Pal, M. Concannon, et al., Endorepellin, the Angio-
static Module of Perlecan, Interacts With Both the α2β1 Integrin and
Vascular Endothelial Growth Factor Receptor 2 (VEGFR2),Journal of
Biological Chemistry 286 (2011): 2594725962.
17. A. Goyal, M. A. Gubbiotti, D. R. Chery, L. Han, and R. V. Iozzo,
EndorepellinEvoked Autophagy Contributes to Angiostasis,Journal
of Biological Chemistry 291 (2016): 1924519256.
18. A. D. Theocharis and N. K. Karamanos, Proteoglycans Remodeling
in Cancer: Underlying Molecular Mechanisms,Matrix Biology 7576
(2019): 220259.
19. M. R. Bond and J. A. Hanover, OGlcNAc Cycling: A Link Between
Metabolism and Chronic Disease,Annual Review of Nutrition 33
(2013): 205229.
20. M. I. Love, W. Huber, and S. Anders, Moderated Estimation of Fold
Change and Dispersion for RNASeq Data With DESeq. 2,Genome
Biology 15 (2014): 550.
21. A. Wald, Tests of Statistical Hypotheses Concerning Several
Parameters When the Number of Observations Is Large,Transactions
of the American Mathematical Society 54 (1943): 426482.
22. Y. Benjamini and Y. Hochberg, Controlling the False Discovery
Rate: A Practical and Powerful Approach to Multiple Testing,Journal
of the Royal Statistical Society Series B: Statistical Methodology 57 (1995):
289300.
23. A. Zeileis and T. Hothorn, Diagnostic Checking in Regression
Relationships,R News 2/3 (2002): 710.
24. A. Rau, M. Gallopin, G. Celeux, and F. Jaffrézic, DataBased Fil-
tering for Replicated HighThroughput Transcriptome Sequencing Ex-
periments,Bioinformatics 29 (2013): 21462152.
25. I. Guyon, J. Weston, S. Barnhill, and V. Vapnik, Gene Selection for
Cancer Classification Using Support Vector Machines,Machine
Learning 46 (2002): 389422.
26. J. Jumper, R. Evans, A. Pritzel, et al., Highly Accurate Protein
Structure Prediction With Alphafold,Nature 596 (2021): 583589.
27. M. Varadi, D. Bertoni, P. Magana, et al., Alphafold Protein Struc-
ture Database in 2024: Providing Structure Coverage for Over 214
Million Protein Sequences,Nucleic Acids Research 52 (2024):
D368D375.
28. S. Gopal, S. Arokiasamy,C.Pataki,J.R.Whiteford,andJ.R.Couchman,
Syndecan Receptors: Pericellular Regulators in Development and Inflam-
matory Disease,Open Biology 11 (2021): 200377.
29. N. A. Afratis, D. Nikitovic, H. A. B. Multhaupt, A. D. Theocharis,
J. R. Couchman, and N. K. Karamanos, Syndecans: Key Regulators of
Cell Signaling and Biological Functions,FEBS Journal 284 (2016):
2741.
30. C. Xie, L. Schaefer, and R. V. Iozzo, Global Impact of Proteoglycan
Science on Human Diseases,iScience. 26 (2023): 108095.
31. B. J. Burbach, Y. Ji, and A. C. Rapraeger, Syndecan1 Ectodomain
Regulates MatrixDependent Signaling in Human Breast Carcinoma
Cells,Experimental Cell Research 300 (2004): 234247.
32. D. M. Beauvais, B. J. Burbach, and A. C. Rapraeger, The
Syndecan1 Ectodomain Regulates αvβ3 Integrin Activity in Human
Mammary Carcinoma Cells,Journal of Cell Biology 167 (2004):
171181.
33. D. M. Beauvais and A. C. Rapraeger, Syndecan1Mediated Cell
Spreading Requires Signaling by αvβ3 Integrins in Human Breast Car-
cinoma Cells,Experimental Cell Research 286 (2003): 219232.
34. S. H. Pham, K. Pratt, R. K. Okolicsanyi, et al., Syndecan1 and 4
Influence Wnt Signaling and Cell Migration in Human Breast Cancers,
Biochimie 198 (2022): 6075.
35. S. Nadanaka, Y. Bai, and H. Kitagawa, Cleavage of Syndecan1
Promotes the Proliferation of the BasalLike Breast Cancer Cell Line
BT549 via Akt SUMOylation,Frontiers in Cell and Developmental
Biology 9 (2021): 659428.
36. N. Hassan, N. Bückreiß, J. Efing, et al., The Heparan Sulfate
Proteoglycan Syndecan1 Triggers Breast Cancer CellInduced Coagu-
lability by Induced Expression of Tissue Factor,Cells 12 (2023): 910.
37. M. W. Conklin, R. E. Gangnon, B. L. Sprague, et al., Collagen
Alignment as a Predictor of Recurrence After Ductal Carcinoma
in Situ,Cancer Epidemiology, Biomarkers & Prevention 27 (2018):
138145.
38. F. Baba, K. Swartz, R. van Buren, et al., Syndecan1 and Syndecan4
Are Overexpressed in an Estrogen ReceptorNegative, Highly Prolifer-
ative Breast Carcinoma Subtype,Breast Cancer Research and
Treatment 98 (2006): 9198.
39. M. Barbareschi, P. Maisonneuve, D. Aldovini, et al., High
Syndecan1 Expression in Breast Carcinoma Is Related to an Aggressive
Phenotype and to Poorer Prognosis,Cancer 98 (2003): 474483.
40. W. Qiao, H. Liu, W. Guo, P. Li, and M. Deng, Prognostic and
Clinical Significance of Syndecan1 Expression in Breast Cancer: A
Systematic Review and MetaAnalysis,European Journal of Surgical
Oncology 45 (2019): 11321137.
41. D. Loussouarn, L. Campion, C. Sagan, et al., Prognostic Impact of
Syndecan1 Expression in Invasive Ductal Breast Carcinomas,British
Journal of Cancer 98 (2008): 19931998.
42. X. Cui, X. Jing, Q. Yi, C. Long, J. Tian, and J. Zhu, Clinico-
pathological and Prognostic Significance of SDC1 Overexpression in
Breast Cancer,Oncotarget 8 (2017): 111444111455.
43. Y. Zhong, F. Li, S. Zhang, et al., Syndecan1 as an Immunogene in
TripleNegative Breast Cancer: Regulation TumorInfiltrating Lym-
phocyte in the Tumor Microenviroment and EMT by TGFb1/Smad
Pathway,Cancer Cell International 23 (2023): 76.
44. P. Lorenzo, A. Aspberg, P. Önnerfjord, M. T. Bayliss, P. J. Neame,
and D. Heinegård, Identification and Characterization of Asporin,
Journal of Biological Chemistry 276 (2001): 1220112211.
45. R. V. Iozzo, The Family of the Small LeucineRich Proteogly-
cans: Key Regulators of Matrix Assembly and Cellular Growth,
Critical Reviews in Biochemistry and Molecular Biology 32 (1997):
141174.
46. R. V. Iozzo, The Biology of the Small LeucineRich Proteoglycans,
Journal of Biological Chemistry 274 (1999): 1884318846.
47. L. Schaefer and R. V. Iozzo, Biological Functions of the Small
LeucineRich Proteoglycans: From Genetics to Signal Transduction,
Journal of Biological Chemistry 283 (2008): 2130521309.
13 of 18
48. S. P. Henry, M. Takanosu, T. C. Boyd, et al., Expression Pattern and
Gene Characterization of Asporin,Journal of Biological Chemistry 276
(2001): 1221212221.
49. S. Kalamajski, A. Aspberg, K. Lindblom, D. Heinegård, and
Å. Oldberg, Asporin Competes With Decorin for Collagen Binding,
Binds Calcium and Promotes Osteoblast Collagen Mineralization,
Biochemical Journal 423 (2009): 5359.
50. R. Satoyoshi, S. Kuriyama, N. Aiba, M. Yashiro, and M. Tanaka,
Asporin Activates Coordinated Invasion of Scirrhous Gastric Cancer
and CancerAssociated Fibroblasts,Oncogene 34 (2015): 650660.
51. Q. Ding, M. Zhang, and C. Liu, Asporin Participates in Gastric
Cancer Cell Growth and Migration by Influencing EGF Receptor Sig-
naling,Oncology Reports 33 (2015): 17831790.
52. P. Li, Asporin Promotes Cell Proliferation via Interacting With
PSMD2 in Gastric Cancer,Frontiers in Bioscience 24 (2019): 11781189.
53. A. Rochette, N. Boufaied, E. Scarlata, et al., Asporin Is a Stromally
Expressed Marker Associated With Prostate Cancer Progression,
British Journal of Cancer 116 (2017): 775784.
54. F. Wei, T. Li, J. Li, et al., Prognostic and Immunological Role of
Asporin Across Cancers and Exploration in Bladder Cancer,Gene 878
(2023): 147573.
55. R. M. Hughes, B. W. Simons, H. Khan, et al., Asporin Restricts
Mesenchymal Stromal Cell Differentiation, Alters the Tumor Micro-
environment, and Drives Metastatic Progression,Cancer Research 79
(2019): 36363650.
56. D. Basak, Z. Jamal, A. Ghosh, et al., Reciprocal Interplay Between
Asporin and Decorin: Implications in Gastric Cancer Prognosis,PLoS
One 16 (2021): e0255915.
57. Z. Zhang, L. Min, H. Li, et al., Asporin Represses Gastric Cancer
Apoptosis via Activating LEF1Mediated Gene Transcription Indepen-
dent of βCatenin,Oncogene 40 (2021): 45524566.
58. Y. Sasaki, K. Takagane, T. Konno, et al., Expression of Asporin
Reprograms Cancer Cells to Acquire Resistance to Oxidative Stress,
Cancer Science 112 (2021): 12511261.
59. S. Zhan, T. Wang, J. Li, H. Zhu, W. Ge, and J. Li, Asporin Interacts
With HER2 to Promote Thyroid Cancer Metastasis via the MAPK/EMT
Signaling Pathway,Frontiers in Oncology 12 (2022): 762180.
60. S. Zhan, J. Li, and W. Ge, Multifaceted Roles of Asporin in Cancer:
Current Understanding,Frontiers in Oncology 9 (2019): 948.
61. S. P. Lall, Z. W. Alsafwani, S. K. Batra, and P. Seshacharyulu, As-
porin: A Root of the Matter in Tumors and Their Host Environment,
Biochimica et Biophysica Acta (BBA)Reviews on Cancer 1879 (2024):
189029.
62. P. Maris, A. Blomme, A. P. Palacios, et al., Asporin Is a Fibroblast
Derived TGFβ1 Inhibitor and a Tumor Suppressor Associated With
Good Prognosis in Breast Cancer,PLoS Medicine 12 (2015): e1001871.
63. D. Simkova, G. Kharaishvili, G. Korinkova, et al., The Dual Role of
Asporin in Breast Cancer Progression,Oncotarget 7 (2016): 52045
52060.
64. M. Békés, D. R. Langley, and C. M. Crews, PROTAC Targeted
Protein Degraders: The Past Is Prologue,Nature Reviews Drug
Discovery 21 (2022): 181200.
65. D. Sui and J. E. Wilson, Interaction of InsulinLike Growth Factor
Binding Protein4, Miz1, Leptin, LipocalinType Prostaglandin D Synthase,
and Granulin Precursor With the NTerminal Half of Type III Hexokinase,
Archives of Biochemistry and Biophysics 382 (2000): 262274.
66. R. V. Iozzo, Proteoglycans: Structure, Biology and Molecular Inter-
actions (New York, New York: Marcel Dekker, Inc, 2000).
67. K. G. Danielson, H. Baribault, D. F. Holmes, H. Graham,
K. E. Kadler, and R. V. Iozzo, Targeted Disruption of Decorin Leads to
Abnormal Collagen Fibril Morphology and Skin Fragility,Journal of
Cell Biology 136 (1997): 729743.
68. N. K. Karamanos, A. D. Theocharis, T. Neill, and R. V. Iozzo,
Matrix Modeling and Remodeling: A Biological Interplay Regu-
lating Tissue Homeostasis and Diseases,Matrix Biology 7576
(2019): 111.
69. R. V. Iozzo and M. A. Gubbiotti, Extracellular Matrix: The Driving
Force of Mammalian Diseases,Matrix Biology 7172 (2018): 19.
70. R. V. Iozzo, Proteoglycans and Neoplasia,Cancer and Metastasis
Review 7 (1988): 3950.
71. R. V. Iozzo and I. Cohen, Altered Proteoglycan Gene Expression
and the Tumor Stroma,Experientia 49 (1993): 447455.
72. C. Xie, D. K. Mondal, M. Ulas, T. Neill, and R. V. Iozzo, Onco-
suppressive Roles of Decorin Through Regulation of Multiple Receptors
and Diverse Signaling Pathways,American Journal of PhysiologyCell
Physiology 322 (2022): C554C566.
73. R. V. Iozzo, R. P. Bolender, and T. N. Wight, Proteoglycan Changes
in the Intercellular Matrix of Human Colon Carcinoma,Laboratory
investigation; A Journal of Technical Methods and Pathology 47 (1982):
124138.
74. R. V. Iozzo and T. N. Wight, Isolation and Characterization of
Proteoglycans Synthesized by Human Colon and Colon Carcinoma,
Journal of Biological Chemistry 257 (1982): 1113511144.
75. L. Mao, J. Yang, J. Yue, et al., Decorin Deficiency Promotes
EpithelialMesenchymal Transition and Colon Cancer Metastasis,
Matrix Biology 95 (2021): 114.
76. M. Santra, T. Skorski, B. Calabretta, E. C. Lattime, and R. V. Iozzo,
De novo Decorin Gene Expression Suppresses the Malignant Pheno-
type in Human Colon Cancer Cells,Proceedings of the National
Academy of Sciences 92 (1995): 70167020.
77. M. Santra, D. M. Mann, E. W. Mercer, T. Skorski, B. Calabretta, and
R. V. Iozzo, Ectopic Expression of Decorin Protein Core Causes a
Generalized Growth Suppression in Neoplastic Cells of Various Histo-
genetic Origin and Requires Endogenous p21, an Inhibitor of Cyclin
Dependent Kinases,Journal of Clinical Investigation 100 (1997):
149157.
78. I. J. Edwards, Proteoglycans in Prostate Cancer,Nature Reviews
Urology 9 (2012): 196206.
79. X. Hu, E. S. Villodre, R. Larson, et al., DecorinMediated Sup-
pression of Tumorigenesis, Invasion, and Metastasis in Inflammatory
Breast Cancer,Communications Biology 4 (2021): 72.
80. X. Bi, C. Tong, A. Dockendorff, et al., Genetic Deficiency of Decorin
Causes Intestinal Tumor Formation Through Disruption of Intestinal
Cell Maturation,Carcinogenesis 29 (2008): 14351440.
81. J. Köninger, N. A. Giese, F. F. di Mola, et al., Overexpressed Dec-
orin in Pancreatic Cancer: Potential Tumor Growth Inhibition and
Attenuation of Chemotherapeutic Action,Clinical Cancer Research 10
(2004): 47764783.
82. J. G. Tralhão, L. Schaefer, M. Micegova, et al., In Vivo Selective and
Distant Killing of Cancer Cells Using AdenovirusMediated Decorin
Gene Transfer,FASEB Journal: Official Publication of the Federation of
American Societies for Experimental Biology 17 (2003): 464466.
83. X. Bi, N. M. Pohl, Z. Qian, et al., DecorinMediated Inhibition of
Colorectal Cancer Growth and Migration Is Associated With
ECadherin In Vitro and in Mice,Carcinogenesis 33 (2012): 326330.
84. T. Y. Eshchenko, V. I. Rykova, A. E. Chernakov, S. V. Sidorov, and
E. V. Grigorieva, Expression of Different Proteoglycans in Human
Breast Tumors,Biochemistry (Moscow) 72 (2007): 10161020.
85. S. Goldoni, D. G. Seidler, J. Heath, et al., An AntiMetastatic Role
for Decorin in Breast Cancer,American Journal of Pathology 173
(2008): 844855.
14 of 18 Proteoglycan Research, 2025
86. X. Bi, X. Xia, D. Fan, et al., Oncogenic Activin C Interacts With
Decorin in Colorectal Cancer In Vivo and In Vitro,Molecular
Carcinogenesis 55 (2015): 17861795.
87. Y. Hu, H. Sun, R. T. Owens, et al., Decorin Suppresses Prostate
Tumor Growth Through Inhibition of Epidermal Growth Factor and
Androgen Receptor Pathways,Neoplasia 11 (2009): 10421053.
88. A. D. Theocharis, S. S. Skandalis, T. Neill, et al., Insights Into the
Key Roles of Proteoglycans in Breast Cancer Biology and Translational
Medicine,Biochimica et Biophysica Acta (BBA)Reviews on Cancer
1855 (2015): 276300.
89. C. Ji, H. Liu, M. Xiang, et al., Deregulation of Decorin and FHL1
Are Associated With Esophageal Squamous Cell Carcinoma Progression
and Poor Prognosis,International Journal of Clinical and Experimental
Medicine 8 (2015): 2096520970.
90. Y. Gao, H. Y. Ma, Q. Y. Xu, et al., Mechanism of Decorin Protein
Inhibiting Invasion and Metastasis of NonSmall Cell Lung Cancer,
European Review for Medical and Pharmacological Sciences 23 (2019):
15201527.
91. X. Zheng, P. Wang, L. Li, et al., CancerAssociated Fibroblasts
Promote Vascular Invasion of Hepatocellular Carcinoma via Down-
regulating DecorinIntegrin β1 Signaling,Frontiers in Cell and
Developmental Biology 9 (2021): 678670.
92. Y. Li, L. Gan, M. Lu, et al., HBx Downregulated Decorin and
DecorinDerived Peptides Inhibit the Proliferation and Tumorigenicity
of Hepatocellular Carcinoma Cells,FASEB Journal 37 (2023): e22871.
93. A. Berdiaki, E. M. Giatagana, G. Tzanakakis, G. Tzanakakis, and
D. Nikitovic, The Landscape of Small LeucineRich Proteoglycan
Impact on Cancer Pathogenesis With a Focus on Biglycan and Lumi-
can,Cancers 15 (2023): 3549.
94. R. A. Assal, R. B. E. Abd ElBary, R. A. Youness, et al., OncomiR
181a Promotes Carcinogenesis by Repressing the Extracellular Matrix
Proteoglycan Decorin in Hepatocellular Carcinoma,BMC
Gastroenterology 24 (2024): 337.
95. L. Rudnicka, J. Varga, A. M. Christiano, R. V. Iozzo, S. A. Jimenez,
and J. Uitto, Elevated Expression of Type VII Collagen in the Skin of
Patients With Systemic Sclerosis,Journal of Clinical Investigation 93
(1994): 17091715.
96. D. G. Seidler, S. Goldoni, C. Agnew, et al., Decorin Protein Core
Inhibits in vivo Cancer Growth and Metabolism by Hindering Epi-
dermal Growth Factor Receptor Function and Triggering Apoptosis via
Caspase3 Activation,Journal of Biological Chemistry 281 (2006):
2640826418.
97. R. V. Iozzo, D. K. Moscatello, D. J. McQuillan, and I. Eichstetter,
Decorin Is a Biological Ligand for the Epidermal Growth Factor
Receptor,Journal of Biological Chemistry 274 (1999): 44894492.
98. M. A. Gubbiotti, S. D. Vallet, S. RicardBlum, and R. V. Iozzo,
Decorin Interacting Network: A Comprehensive Analysis of Decorin
Binding Partners and Their Versatile Functions,Matrix Biology 55
(2016): 721.
99. D. S. Grant, C. Yenisey, R. W. Rose, M. Tootell, M. Santra, and
R. V. Iozzo, Decorin Suppresses Tumor CellMediated Angiogenesis,
Oncogene 21 (2002): 47654777.
100. E. Schönherr, C. Sunderkötter, L. Schaefer, et al., Decorin Defi-
ciency Leads to Impaired Angiogenesis in Injured Mouse Cornea,
Journal of Vascular Research 41 (2004): 499508.
101. E. Schönherr, M. Broszat, E. Brandan, et al., Decorin Core Protein
Fragment Leu 155Val260 Interacts With TGFB but Does Not Compete
for Decorin Binding to Type I Collagen,Arch.Biochem.Biophys. 35
(1998): 241248.
102. K. Baghy, R. V. Iozzo, and I. Kovalszky, DecorinTGFβAxis in
Hepatic Fibrosis and Cirrhosis,Journal of Histochemistry &
Cytochemistry 60 (2012): 262268.
103. C. Bocian, A. K. Urbanowitz, R. T. Owens, R. V. Iozzo, M. Götte,
and D. G. Seidler, Decorin Potentiates InterferonγActivity in a Model
of Allergic Inflammation,Journal of Biological Chemistry 288 (2013):
1269912711.
104. Y. Yamaguchi, D. M. Mann, and E. Ruoslahti, Negative Regula-
tion of Transforming Growth Factorβby the Proteoglycan Decorin,
Nature 346 (1990): 281284.
105. Z. Ferdous, V. M. Wei, R. Iozzo, M. Höök, and K. J. GrandeAllen,
DecorinTransforming Growth Factorß Interaction Regulates Matrix
Organization and Mechanical Characteristics of ThreeDimensional Colla-
gen Matrices,Journal of Biological Chemistry 282 (2007): 3588735898.
106. J. Cui, S. Zhang, K. Acharya, et al., Decorin Attenuates Hyper-
trophic Scar Fibrosis via TGFB/Smad Signalling,Experimental
Dermatology 33 (2024): e15133.
107. V. Zinchuk, Y. Wu, O. GrossenbacherZinchuk, and E. Stefani,
Quantifying Spatial Correlations of Fluorescent Markers Using Enhanced
Background Reduction With Protein Proximity Index and Correlation
Coefficient Estimations,Nature Protocols 6 (2011): 15541567.
108. K. Baghy, K. Dezső, V. László, et al., Ablation of the Decorin Gene
Enhances Experimental Hepatic Fibrosis and Impairs Hepatic Healing
in Mice,Laboratory Investigation 91 (2011): 439451.
109. R. V. Iozzo, Matrix Proteoglycans: From Molecular Design to
Cellular Function,Annual Review of Biochemistry 67 (1998): 609652.
110. H. Järveläinen, P. Puolakkainen, S. Pakkanen, et al., A Role for
Decorin in Cutaneous Wound Healing and Angiogenesis,Wound
Repair and Regeneration 14 (2006): 443452.
111. A. Torres, M. A. Gubbiotti, and R. V. Iozzo, DecorinInducible Peg3
Evokes Beclin 1Mediated Autophagy and Thrombospondin 1Mediated
Angiostasis,Journal of Biological Chemistry 292 (2017): 50555069.
112. K. N. Sulochana, H. Fan, S. Jois, et al., Peptides Derived From
Human Decorin LeucineRich Repeat 5 Inhibit Angiogenesis,Journal
of Biological Chemistry 280 (2005): 2793527948.
113. H. Fan, K. N. Sulochana, Y. S. Chong, and R. Ge, Decorin Derived
Antiangiogenic Peptide LRR5 Inhibits Endothelial Cell Migration by
Interfering With VEGFStimulated NO Release,International Journal
of Biochemistry & Cell Biology 40 (2008): 21202128.
114. H. Järveläinen, A. Sainio, and T. N. Wight, Pivotal Role for Dec-
orin in Angiogenesis,Matrix Biology 43 (2015): 15