Gene Expression Profiling of Triple-Negative Breast Tumors with Different Expression of Secreted Protein Acidic And Cysteine Rich (SPARC)

Abstract

Aim:

To determine the expression signature of triple-negative breast cancer (TNBC) with differences of secreted protein acidic and rich in cysteine expression and clinical behavior.

Patients, materials & methods:

cDNA microarray analysis was performed to determine the expression profiling of TNBC, characterized regarding secreted protein acidic and rich in cysteine expression status. Immunohistochemistry analysis on tissue microarrays containing an independent cohort of TNBC was performed for validation.

Results:

Negative staining of SOHLH2 and positive staining of DNAJC12 and LIM1 was correlated with a poor outcome of the patients.

Conclusion:

Our findings provide new information on transcriptome changes associated with the clinical behavior of TNBC that may serve as a potential tool for the identification and characterization of new candidate biomarkers.

Keywords::

• expression profiling
• immunohistochemistry
• protein expression
• SPARC
• tissue microarray
• triple-negative breast cancer

Breast cancer is the second most common type of cancer worldwide and by far the most frequent among women, with an estimated 1.67 million new cases diagnosed in 2012 – the equivalent of 25% of all cancers [1]. Breast cancer is a complex and heterogeneous disease that is currently classified into four intrinsic subtypes: Luminal-A, Luminal-B, Triple-Negative and HER-2-overexpressed breast cancer [2–6].

Triple-negative breast cancer (TNBC) subtype constitutes a distinct group of breast tumors characterized by the absence of estrogen receptor (ER) and progesterone receptor expression, and absence of amplification/overexpression of the HER2 oncoprotein [7]. TNBC is a highly lethal molecular intrinsic subtype corresponding to 12–17% of all breast tumors [8–10]. TNBC often affects African–American patients and young premenopausal women [11,12]. There is a well-established association between deleterious BRCA1 mutations and the risk of developing TNBC subtype, with lifetime risks reaching as high as 50–85% [13–15]. In addition to inactivation of BRCA1, mutations in TP53 and loss of RB1 function are also frequently observed in TNBC [11,16]. Studies using next-generation DNA sequencing and the integration of genomic, epigenomic and transcriptome platforms have revealed a complex profile of TNBC molecular changes, including amplifications, deletions and mutations, which characterize tumors with a high degree of genomic instability [17,18]. The high genetic heterogeneity of these tumors represents a challenge for the development of targeted therapies targeting this tumor subtype. Currently, there is no clinically approved target therapy, and treatment options are reduced, making chemotherapy and surgery still the conventional approach. Neoadjuvant chemotherapy has proven efficacy in the treatment of TNBC, and the regimens include combinations of anthracyclines, alkylating agents, taxanes and platinum salts [19–22].

The secreted protein acidic and cysteine rich (SPARC) gene (also known as BM-40 or osteonectin) encodes a 32-kDa matricellular glycoprotein involved in several biological processes, such as extracellular matrix organization (ECM), focal adhesion, cell proliferation, cell death, angiogenesis and cell differentiation [23–25]. Due to its matricellular protein characteristics, SPARC has multiple functional properties that make it difficult to establish its role in tumor progression, which depending on the context, cell type and microenvironment can act as an oncogene or tumor-suppressor gene [26]. Changes in SPARC expression profile are observed in different tumors. However, the role of SPARC protein and its prognostic value in different types of tumors is divergent and reflects the complexity of actions of this matricellular protein [27–30]. Also, there is evidence that increased SPARC expression confers increased sensitivity to chemotherapy and radiotherapy in colon carcinoma [31]. Furthermore, it has been shown to bind with high affinity to albumin, which raised some interest in its possible role in improving the delivery of albumin-bound cytotoxic drugs to the tumor [32].

A previous study from our group has associated low SPARC expression with worse prognosis and more aggressive phenotypes of breast cancer, including TNBC [33]. In the present study, using cDNA microarray analysis, we evaluated the expression signature of TNBC with differences of SPARC expression. Several genes were identified when comparing the gene expression profile of TNBCs with positive expression of SPARC and good prognosis in comparison with TNBC with negative SPARC expression and poor prognosis. Three of the genes identified as differentially expressed, SOHLH2, DNAJC12 and LIM1 were further validated using immunohistochemistry (IHC) on tissue microarrays (TMAs) containing TNBC. Our study allowed the identification of new biomarker candidates with potential for a better classification and prognosis prediction of TNBC.

Materials & methods

Study design, patients & tumor specimens

The study was designed to evaluate the expression profiling of TNBC with differences in SPARC expression. For this retrospective study, we used two independent cohorts composed of TNBC patients. The discovery cohort was composed of four samples of primary TNBC samples that were selected based on the tumor status of SPARC expression and patient’s outcome previously determined by our group [33]. Tumor samples were dissected to remove residual normal tissue before being frozen and stored in liquid nitrogen until the time of extraction of RNA. From this cohort, two TNBC samples with SPARC expression under median value (66.7–89.9) and short overall survival (OS; 12.5–32 months), and other two samples with SPARC expression above median value (127.6–169.4) and long OS (51.9–57.9 months) were used for gene expression cDNA microarray analysis. The validation cohort composed of 51 specimens of well-characterized TNBC was retrospectively obtained from A. C. Camargo Cancer Center, São Paulo, Brazil archives and evaluated as duplicate by IHC in TMA slides. All 51 patients were diagnosed with invasive ductal carcinomas, the majority of the patients had tumors with histological and nuclear Grade 3, and clinical stage was homogeneous between Stages II and III at diagnosis (Table 1). The age of the patients ranged from 24 to 77 years (median 48 years). The follow-up period ranged from 18.57 to 243.67 months (median 80.2 months). The Institutional Ethics Committee approved this study and all subjects provided informed consent.

Table 1.  Clinicopathological features of triple-negative breast cancer patients.

Variable	Characteristic	n (%)
Age (years)	Median 48 (range 24–77)
TNM stage	Early	20 (39.2)
	Late	20 (39.2)
	Missing	11 (21.6)
Tumor size	pT1 + pT2	41 (80.4)
	pT3 + pT4	10 (19.6)
Lymph node status	pN0	29 (56.9)
	pN^+	21 (41.2)
	Missing	01 (2.0)
Histological grade	I + II	07 (13.7)
	III	44 (86.3)
Nuclear grade	II	03 (5.9)
	III	47 (92.2)
	Missing	01 (2.0)

Tumor-node-metastasis (TNM) stage: Patients’ stage determined according to the 5th Edition of the Union for International Cancer Control (UICC) TNM classification of malignant tumors.

Extraction of RNA

Tumor samples were sprayed with the aid of a Termovac, and total RNA was extracted using the acid guanidinium thiocyanate-phenol-chloroform method [34], and high-quality RNA was purified using the RNeasy Mini Kit (Qiagen, MD, USA) as previously described [35]. The sample integrity was also determined by RNA nanochips using the Agilent 2100 Bioanalyzer platform (Agilent Technologies, CA, USA), using samples that had an RNA integrity number (RIN) ≥ 8.

cDNA microarray analysis

The cDNA microarray analysis was performed using the low-input RNA Linear Amplification Kit 2-color (GE Healthcare Bio-Sciences Corp., NJ, USA) as previously described [35]. The data were quantified, and the quality control performed using the Agilent FE Software (Agilent Technologies). The extracted data were imported into Agilent Gene Spring 12.5 – GX Analyze Program (Agilent Technologies). The Ingenuity Pathway Knowledge Base (IPA) system (Qiagen) was used to identify and construct gene networks.

The raw data from the present study have been deposited in NCBI′s (National Center for Biotechnology Information) Gene Expression Omnibus (GEO) [36] and are accessible through GEO Series accession number GSE98931 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE98931).

TMAs construction & IHC

Breast TMAs were constructed and immunostained as previously described [33]. The primary antibodies were rabbit polyclonal anti-SOHLH2 (Abcam, MA, USA; catalog ab101402, 1:50), rabbit polyclonal anti-SPARC (Santa Cruz Biotechnology, Inc., CA, USA; catalog sc-25574; clone H90, 1:50), goat polyclonal anti-DNAJC12 (Santa Cruz Biotechnology, Inc.; catalog ta-803415; clone OTI1C2, 1:200), mouse monoclonal anti-LIM1 (Abcam; all according to the manufacturers’ recommendations). The optimal dilution was defined using well-characterized positive samples and tested before staining of the TMA slides.

Evaluation of IHC

Immunohistochemical evaluation was semiquantitative for each marker, considering the percentage of stained cells and intensity of the reaction. For the most intense markers (SPARC and SOHLH2), the immunostainings were graded according to the intensity of staining: 0 (no signal), 1 (weak), 2 (moderate) and 3 (strong), and to the percentage of stained cells: 0 (no cell), 1 (<10%), 2 (10–50%), 3 (51–90%) and 4 (>90% of cells). The core was classified according to the most representative staining in tumor cells, and for statistical analysis, cases were grouped as negative (no signal/weak) or positive (moderate/strong). The lower intensity markers (DNAJC12 and LIM1) were categorized as positive when more than 10% of the cells were labeled, and negative when the labeling was absent, or present in less than 10% of the tumor cells.

Statistical analysis

Statistical analysis was performed using demographic, clinical and pathological information of the patients, and patient follow-up tabulated in the Statistical Package for Social Science for Windows program, version 20.0. The correlations between the data obtained in the immunohistochemical analysis and the demographic, clinical and anatomopathological variables were calculated by the Fisher exact test. The analysis of the prognostic factors was done through the Kaplan–Meier product-limit estimator and Cox proportional hazard models. To compare the cumulative survival probability curves, we used the log-rank test. All differences with a p-value of less than 0.05 were considered significant.

Results

In this study, we sought to determine the expression profile of TNBC tumors with differences in SPARC expression pattern and biological behavior to identify potential new biomarkers. cDNA microarray assays (Agilent Technologies, 4X44K) were performed to identify differentially expressed genes in a small discovery cohort composed by TNBC with SPARC-positive expression and good prognosis (TNBC/SPARC⁺) and TNBC with SPARC-negative expression and poor prognosis (TNBC/SPARC^-). Using a fold change of more than three as a cutoff, we identified a significant number of differentially expressed genes (Supplementary Tables 1 and 2). In the comparison between TNBC/SPARC⁺ versus TNBC/SPARC^-, 338 transcripts were upregulated and 437 transcripts were downregulated. The lists containing the top 50 differentially expressed genes (up- and downregulated) and its association to cancer are presented in Tables 2 and 3, respectively. Figure 1 shows the heat map of the top 50 differentially expressed genes.

Table 2.  50 top genes differentially expressed identified in transcriptome analysis: SPARC positive versus SPARC negative – positively regulated.

Gene symbol	Description	Fold change	Associated with cancer
SOHLH2	Spermatogenesis and oogenesis-specific basic helix-loop-helix 2	24.20	Yes^†
KLHL13	Kelch-like family member 13	22.09	No
HPGD	Hydroxyprostaglandin dehydrogenase 15-(NAD)	20.01	Yes^‡
GPR158	G-protein-coupled receptor 158	18.73	Yes
FABP7	Fatty acid-binding protein 7, brain	18.30	Yes^‡
CTAG1A	Cancer/testis antigen 1A	18.09	Yes
ZNF556	Zinc finger protein 556	17.34	No
SLC26A4	Solute carrier family 26 (anion exchanger), member 4	15.66	Yes^†
MAGEA1	Melanoma antigen family A, 1	14.49	Yes^‡
UGT2B11	UDP glucuronosyltransferase 2 family, polypeptide B11	14.13	Yes^†
SLAIN1	SLAIN motif family, member 1	13.44	Yes
C6orf105	Androgen-dependent TFPI-regulating protein	13.05	No
GSTM2P1	Glutathione S-transferase mu 2 (muscle) pseudogene 1	13.04	No
HS3ST4	Heparan sulfate (glucosamine) 3-O-sulfotransferase 4	12.58	No
CYP26A1	Cytochrome P450, family 26, subfamily A, polypeptide 1	12.56	Yes^†
TOX3	TOX high-mobility group box family member 3	12.52	Yes^†
FBN2	Fibrillin 2	11.53	Yes^†
FSIP1	Fibrous sheath-interacting protein 1	11.29	Yes^†
CDH1	Cadherin 1, type 1, E-cadherin (epithelial)	10.64	Yes^‡
ZNF385B	Zinc finger protein 385B	10.52	Yes^†
TDRD9	Tudor domain containing 9	9.71	No
CHRDL2	Chordin-like 2	9.67	Yes^†
CELF4	CUGBP, Elav-like family member 4	9.10	Yes
TNFRSF17	Tumor necrosis factor receptor superfamily, member 17	8.63	Yes^†
DDC	Dopa decarboxylase (aromatic L-amino acid decarboxylase)	8.53	Yes
KIRREL3	Kin of IRRE like 3 (Drosophila)	8.31	Yes
FBXW10	F-box and WD repeat domain containing 10	8.20	Yes
GLYATL2	Glycine-N-acyltransferase-like 2	8.13	Yes
TRIM36	Tripartite motif containing 36	7.99	Yes^†
MSMB	Microseminoprotein, β-	7.97	Yes^†
ZNF204P	Zinc finger protein 204, pseudogene	7.63	No
SEMA3D	Sema domain, immunoglobulin domain (Ig), short basic domain, secreted, (semaphorin) 3D	7.47	Yes
TSPAN8	Tetraspanin 8	7.41	Yes
MUC20	Mucin 20, cell surface associated	7.37	Yes
APOD	Apolipoprotein D	7.33	Yes^†
C8orf46	Chromosome 8 open-reading frame 46	7.31	No
ATP1A2	ATPase, Na^+/K^+ transporting, α 2 polypeptide	7.26	Yes^†
MYOM2	Myomesin 2	7.02	Yes^†
FAM65B	Family with sequence similarity 65, member B	6.78	Yes
CLDN1	Claudin 1	6.69	Yes^‡
RIMS2	Regulating synaptic membrane exocytosis 2	6.67	Yes
SC5DL	Sterol-C5-desaturase	6.56	No
PPM1H	Protein phosphatase, Mg^2+/Mn^2 ^+ dependent, 1H	6.51	Yes^†
MAGEC1	Melanoma antigen family C, 1	6.26	Yes^†
MSLN	Mesothelin	6.26	Yes^‡
EAF2	ELL-associated factor 2	6.14	Yes
AMTN	Amelotin	6.14	Yes
IGLL5	Immunoglobulin lambda-like polypeptide 5	6.07	Yes^†
FAM70A	Transmembrane protein 255A	5.97	No
EGF	Epidermal growth factor	5.87	Yes^‡

^†Associated with breast cancer.

^‡Associated with TNBC.

TNBC: Triple-negative breast cancer.

Figure 1.  Heat map depicting the relative fold change in expression of differentially expressed genes in the comparison of TNBC/SPARC+ vs TNBC/SPARC-.

The hierarchical cluster shows the fold change inexpression of the 50 TOP upregulated genes and 50 TOP downregulated genes, using the cut off foldchange ≥ 3. Columns in heat map correspond to the samples categorized as TNBC with SPARC positive expression and good prognosis (in grey) and TNBC with SPARC negative expression and poor prognosis (in black). Transcript enrichment is encoded in the heat map in green when showing downregulation; or red when showing upregulation in the comparison of TNBC/SPARC+ vs TNBC/SPARC.

TNBC: Triple negative breast cancer; TOP: Gene expression profiling of TNBC with differences in SPARC expression.

The differentially expressed genes were categorized by the gene ontology analysis regarding of the cellular processes they participate. They are involved in different biological processes, such as immune system processes, metabolic processes, cellular response to chemical stimuli, angiogenesis, apoptosis and cell adhesion. However, there was no increment in specific biological processes in the comparison between up- or downregulated genes. Furthermore, using the IPA Ingenuity^® program, we evaluated the possible occurrence of interactions between the differentially expressed genes in the TNBC/SPARC⁺ versus TNBC/SPARC^- comparison. Different gene networks were generated. Among these networks, we highlight one that shows the negative modulation of transcripts of members of the CEACAM family (CEACAM-3, 5, 6 and 7) and the LHX1 (LIM1) gene transcripts (Figure 2).

Figure 2.  Molecular interaction network of the downregulated genes in the comparison of TNBC/SPARC+ vs. TNBC/SPARC-.

Genes or gene products are represented as nodes, and the biological relationship betweentwo nodes is represented as an edge. The intensity of the node color indicates the degree of upregulation (red) or downregulation (green). Edges are displayed with various labels that describe the nature of the relationship between the nodes: — binding only, → acts on, Θ auto-regulation.

The selection of the genes for validation was done using gene annotation in public data base and literature mining. Several genes identified as differentially expressed in the comparison between TNBC/SPARC⁺ versus TNBC/SPARC^- seem to be good candidates for biomarkers with potential prognostic value for TNBC. Two of the selected genes showed high fold change in the comparison between TNBC/SPARC⁺ versus TNBC/SPARC^-, one upregulated (SOHLH2; fold change: 24.20) and one downregulated (LHX1; fold change: -22.51), and are contained in the table of the top 50 genes (Table 2 & Table 3). Also, we selected for validation an upregulated gene with showed a smaller fold change (DNAJC12; fold change: 3.5; Supplementary Table 1), which is already part of the gene pool studied by our laboratory [37]. For the validation of the selected genes, we used two independent cohorts. First, we performed an exploratory analysis using the KM Plotter database (www.kmplot.com [38]) to evaluate Kaplan–Meyer curves for disease-free survival (DFS) and OS for all breast cancer subtypes and for Basal-like/TNBC stratified by the transcript expression of SOHLH2, DNAJC12 and LIM1. This analysis was performed using the best cutoff and validated probes (SPARC – 212667_at, SOHLH2 – 1555401_at, DNAJC2 – 223722_at and LIM1 – 206230_at). Second, we sought to access the prognostic value of the proteins encoded by the selected genes by IHC on TMAs containing 51 TNBC samples (Figure 3).

Table 3.  50 top genes differentially expressed identified in transcriptome analysis: SPARC positive versus SPARC negative – negatively regulated.

Gene symbol	Description	Fold change	Associated with cancer
KRT1	Keratin 1	-57.75	Yes^‡
CAPN8	Calpain 8	-33.00	No
IVL	Involucrin	-32.16	Yes^†
SPRR3	Small proline-rich protein 3	-31.54	Yes^†
SCGB2A2	Secretoglobin, family 2A, member 2	-30.474	Yes^†
SCGB2A1	Secretoglobin, family 2A, member 1	-25.594	Yes^†
LHX1	LIM homeobox 1	-22.51	Yes
CEACAM7	Carcinoembryonic antigen-related cell adhesion molecule 7	-21.611	Yes^†
DDTL	D-dopachrome tautomerase-like	-21.601	No
MesP1	Mesoderm posterior basic helix-loop-helix transcription factor 1	-21.20	No
CEACAM6	Carcinoembryonic antigen-related cell adhesion molecule 6 (nonspecific cross-reacting antigen)	-19.95	Yes^‡
FAM25A	Family with sequence similarity 25, member A	-19.13	No
PKNOX2	PBX/knotted 1 homeobox 2	-18.79	No
CHP2	Calcineurin-like EF-hand protein 2	-17.99	Yes
CAMP	Cathelicidin antimicrobial peptide	-17.759	Yes^†
SERPINB4	Serpin peptidase inhibitor, clade B (ovalbumin), member 4	-17.099	Yes
SERPINB3	Serpin peptidase inhibitor, clade B (ovalbumin), member 3	-16.55	Yes^†
CAPS	Calcyphosine	-14.89	Yes^†
SCGB1D2	Secretoglobin, family 1D, member 2	-14.83	Yes^†
HS6ST3	Heparan sulfate 6-O-sulfotransferase 3	-13.98	Yes^†
PLA2G4D	Phospholipase A2, group IVD (cytosolic)	-13.79	Yes
KRT32	Keratin 32	-13.72	No
SPINK6	Serine peptidase inhibitor, Kazal type 6	-12.23	Yes
HMP19	HMP19 protein	-12.20	Yes
KRTDAP	Keratinocyte differentiation-associated protein	-11.97	Yes
HIST2H3A	Histone cluster 2, H3a	-11.97	Yes
C9orf169	Cysteine-rich tail protein 1	-11.867	No
SIM2	Single-minded family bHLH transcription factor 2	-11.857	Yes^†
DAPL1	Death-associated protein-like 1	-11.237	Yes
FABP6	Fatty acid binding protein 6, ileal	-11.17	Yes
SNORA74A	Small nucleolar RNA, H/ACA box 74A	-11.16	Yes
SNORA53	Small nucleolar RNA, H/ACA box 53	-11.12	No
CEACAM3	Carcinoembryonic antigen-related cell adhesion molecule 3	-11.06	Yes
DEFB1	Defensin, β1	-11.03	Yes
SNORA28	Small nucleolar RNA, H/ACA box 28	-10.45	No
S100A14	S100 calcium-binding protein A14	-10.39	Yes^‡
SLURP1	Secreted LY6/PLAUR domain containing 1	-10.31	Yes^‡
RGS20	Regulator of G-protein signaling 20	-10.23	Yes^‡
CEACAM5	Carcinoembryonic antigen-related cell adhesion molecule 5	-10.21	Yes^‡
HHIPL2	HHIP-like 2	-10.19	Yes
TMEM45A	Transmembrane protein 45A	-10.14	Yes^†
RN7SK	RNA, 7SK small nuclear	-10.11	No
DEFB4A	Defensin, β4A	-9.81	Yes^†
SNORD3B-1	Small nucleolar RNA, C/D box 3B-1	-9.76	No
CDSN	Corneodesmosin	-9.56	No
RAET1E	Retinoic acid early transcript 1E	-9.54	Yes^†
PVALB	Parvalbumin	-9.49	Yes
DMKN	Dermokine	-9.44	Yes
ACOX2	Acyl-CoA oxidase 2, branched chain	-9.35	Yes^†
PCDHB5	Protocadherin β 5	-9.305	Yes

^†Associated with breast cancer.

^‡Associated with triple-negative breast cancer.

Figure 3.  Protein expression of SOHLH2, DNAJC12, and LIM1 studied by immnohistochemistry on tissue microarrays containing TNBC samples.

SOHLH2 protein showed nuclear and cytoplasmic staining: (A) tumor cells with cytoplasmic weak staining. (B) Nuclear and cytoplasmic positive staining of tumor cells with a discreet stromal staining. DNAJC12 protein showed cytoplasmic staining. (C) Negative staining. (D) Cytoplasmic positive staining. LIM1 protein showed cytoplasmic staining. (E) Negative staining. (F) Positive staining.

Magnification: 20x.

The Kaplan–Meier curves generated using the KM Plotter tool for all breast cancer subtypes, and Basal-like/TNBC subtype stratified by the transcript expression of selected genes are presented in Supplementary Figures 1, 2, and 3. Low mRNA levels of SOHLH2 were associated with short DFS and a trend to be associated with OS for all breast cancer subtypes and Basal-like/TNBC subtype patients (Supplementary Figure 1). Interestingly, low mRNA levels of DNAJC12 were found to be correlated to worsen DFS and OS for breast cancer patients of all subtypes, whereas the high transcriptional expression of DNAJC12 was associated with shorter DFS and OS in Basal-like/TNBC subtype (Supplementary Figure 2). Low mRNA levels of LIM1 expression were associated with shorter DFS for both breast cancer patients of all subtypes and Basal-like/TNBC patients (Supplementary Figure 3).

To access the contribution of the potential new biomarkers to the development and progression of TNBC, the demographic and clinical–pathological characteristics of the patients were compared to their status of immunostaining. SPARC (neg. vs pos.: 47.5 ± 11.9 vs 51.6 ± 13.3; p = 0.183), SOHLH2 (neg. vs pos.: 34.2 ± 9.4 vs 51.8 ± 11.3; p = 0.005) and LIM1 (neg. vs pos.: 54.1 ± 10.4 vs 44.9 ± 11.2; p = 0.023) immunostaining showed age differences (Mann–Whitney test). No age differences were observed for the DNAJC12 immunostaining status. There was no significant correlation between SPARC, SOHLH2, DNAJC12 and LIM1 staining status and tumor size, lymph nodal status, clinical Stage and histological. LIM1 was the only to present a correlation with nuclear grade (p = 0.018).

To access the individual prognostic value of the potential new biomarkers, we evaluated the association between the IHC expression and patient’s outcomes by Kaplan–Meier analysis for the validation cohort. The immunohistochemical expression of the SPARC protein was determined for the evaluation in combination with the expression of the potential new biomarkers. Although it did not reach statistical significance, the negative staining of the SPARC protein showed a tendency to be associated with worse prognosis. Patients with tumors with reduced expression of SPARC had a shorter DFS time and a shorter OS time when compared with patients with tumors with SPARC-positive staining (Figure 4A, DFS; p = 0.154; Figure 4B, OS; p = 0.163). Here, due to the limited number of recruited patients, multivariate analyses or binary logistic regression of gene expression status with all other clinical–pathological variables were not applicable.

Figure 4.  Downregulation of SPARC is associated with poor prognosis of TNBC.

Kaplan-Meier curves for disease-free (A) and overall (B) survival of triple-negative breast cancer patients, stratified according to SPARC protein expression. Patients were categorized as positive (moderate or intense) or negative (negative or weak) according to SPARC protein immunostaining. LogRank test was performed for curves comparison.

Negative SOHLH2 immunohistochemical expression showed a trend to be correlated with a poor prognosis for both DFS (Figure 5A; p = 0.134) and OS (Figure 5B; p = 0.182). The evaluation of the new potential biomarkers in association with the expression of SPARC was only possible in the analyzable cases for SPARC, which further limited our casuistic. Interestingly, when the cohort was split into four subgroups for the evaluation of SOHLH2 expression in combination with SPARC expression, only TBNC patients with tumors with SPARC-negative staining and SOHLH2-positive staining showed a trend to unfavorable prognosis (Figure 5C, DFS; p = 0.412; Figure 5D, OS; p = not reached).

Figure 5.  Downregulation of SOHLH2 is associated with a worse clinical evolution of triple-negative breast cancer.

Kaplan-Meier curves for disease-free (A and C) and overall (B and D) survival of triple-negative breast cancer patients stratified according to SOHLH2 protein expression. Patients were categorized as positive (moderate orintense) or negative (negative or weak) according to SOHLH2 protein immunostaining. In (C) and (D), patients were classified in four categories according to the protein immunostaining of SPARC and SOHLH2 status. LogRank test was performed for curves comparison.

The DFS and OS curves for patients with TNBC stratified by the immunohistochemical expression of DNAJC12 alone and in combination with SPARC expression are shown in Figure 6. Positive staining of DNAJC12 protein showed a significant association with a worse prognosis for both DFS (Figure 6A; p = 0.005) and OS (Figure 6B; p < 0.001) in patients with TNBC. In the evaluation of the expression of DNAJC12 in association with the expression of SPARC, we observed that the combination of positive expression of DNAJC12 and negative expression of SPARC confers increased tumor aggressiveness and confers the worse prognosis for patients with TNBC (Figure 6C & D). Although not statistically significant, patients with tumors with negative expression of SPARC and positive expression of DNAJC12 (TNBC SPARC^-/DNAJC12⁺) showed a trend to have the shorter time of DFS (Figure 6C; p = 0.331) and OS (Figure 6D; p = 0.273) compared with the other subgroups.

Figure 6.  Upregulation of DNAJC12 is associated with poor prognosis of triple-negative breast cancer.

Kaplan-Meier curves for disease-free (A and C) and overall (B and D) survival of triple-negative breast cancer patients stratified according to DNAJC12 protein expression. Patients were categorized as positive (>10%) ornegative (<10%) according to the percentage of tumor cells positive to DNAJC12. In (C) and (D), patients were classified in four categories according to the protein immunostaining of SPARC and DNAJC12 status. LogRank test was performed for curves comparison.

Although without statistical significance, positive staining of LIM1 was associated with unfavorable prognosis. Patients with tumors with positive immunostaining for LIM1 showed a shorter DFS (Figure 7A; p = 0.119) and OS (Figure 7B; p = 0.103) than patients with tumors with negative immunostaining. When stratifying patients for the combination of LIM1 and SPARC expression status, positive staining of LIM1 in the presence of negative staining for SPARC conferred the worse prognosis as compared with all other SPARC and LIM1 combinations (Figure 7C & D).

Figure 7.  LIM1 overexpression tends to be associated with more aggressive biological behavior in triple-negative breast cancer.

Kaplan-Meier curves for disease-free (A and C) and overall (B and D) survival of triple-negative breast cancer patients stratified according to LIM1 protein expression. Patients were categorized as positive (>10%) or negative (<10%) according to the percentage of tumor cells positive to LIM1. In (C) and (D), patients were classified in four categories according to the protein immunostaining of SPARC and LIM1 status. LogRank test was performed for curves comparison.

Discussion

Triple-negative tumors represent a unique subtype of breast cancer and have been indicative of poorer prognosis since its onset [10,19,39,40]. We have previously shown that SPARC downregulation is associated with poor prognosis in TNBC patients [33]. In the present study, we used a cDNA microarray to identify transcriptional changes that could better characterize differences in the clinical behavior of TNBC with negative SPARC expression and positive SPARC expression. Currently, we were able to identify a large number of up- and downregulated genes with the potential to be associated with the clinical behavior of TNBC. Furthermore, using an independent validation cohort, we examine the protein expression pattern of three of the differentially expressed genes, SOHLH2, DNAJC12 and LIM1, and their relationship with TNBC patients’ survival data.

Within this scenario, the gene SOHLH2 (spermatogenesis and oogenesis basic helix-loop-helix transcription factor 2) was highlighted as one of the top upregulated transcripts among the differentially expressed genes. SOHLH2 encodes for a protein that belongs to the BHLH superfamily of transcription factors that play critical roles in many physiological processes, including cell differentiation, cell cycle disruption and apoptosis [41]. SOHLH2 is a transcriptional factor essential for spermatogenesis and oogenesis [42]. There is evidence that SOHLH2 can act as homodimer or heterodimer with SOHLH1 to promote spermatogonial stem cells and spermatogonial cells differentiation by the transcriptional regulation of different target genes [42–44]. Recent data show that high expression of SOHLH2 is observed in many normal tissues, especially in epithelial tissues, while a low level of expression of SOHLH2 has been detected in many cancer cell lines [45]. SOHLH2 are considered a tumor-suppressor gene for ovarian epithelial tumors, where low levels of SOHLH2 expression were associated with increased cyclin D1 activity, and consequently with worse prognosis [46]. We found similar results in TNBC, in which was possible to perceive that when there are low levels of SOHLH2 expression, there is a worse clinical evolution, whereas in tumors that expressed high levels of SOHLH2 were observed a better DFS and OS, respectively, indicating that SOHLH2 may also act as a tumor suppressor in breast cancer. In breast tumors, Ji et al. [47] demonstrated that reduced expression of SOHLH2 in metastatic breast tumors. In the same study, these authors also found that ectopic expression of SOHLH2 in breast cancer cells reduced EMT and inhibited cell migration and invasion in vitro, and metastasis in vivo [47]. Although not statistically significant, in our study, when analyzing SOHLH2 in combination with SPARC expression, it is possible to recognize an opposite trend of SOHLH2 expression regarding of prognostic in patients with TNBC. We found that positive expression of SOHLH2 was related to events of recurrence, compared with the group of patients with TNBC with negative expression. We are aware that the main limitation of our study is the small quantity of TNBC analyzed. However, our results provide new insights on the role played by SOHLH2 in TNBC that warranted further experimental and clinical studies and the potential role of SOHLH2 in breast cancer.

Here, the transcripts of DNAJC12 were found to be upregulated in TNBC with positive expression of SPARC (fold change = 3.5) in comparison with TNBC with negative expression of SPARC. The DNAJC12 gene (also named JDP1) encodes for the J-domain containing protein 1, a member of the small molecular size heat-shock proteins (DNAJ/HSP40) family [48]. The DNAJ/HSP40 family is highly conserved through evolution and comprises 41 members with a molecular size of approximately 40 kDa that has essential cochaperone activity with other heat-shock proteins [48,49]. Experimental and clinical evidences have indicated the involvement of some of the DNAJ/HSP40 family members in different types of cancers [50]. In previous studies, we have identified the transcripts of DNAJC12 as differentially expressed between ER-positive and ER-negative breast tumors [51]. Furthermore, we provide evidence that DNAJC12 is an ER target gene highly expressed in ER-positive tumors [37]. In the present study, the positive immunohistochemical expression of DNAJC12 was associated with poorer prognosis in TNBC patients. When analyzed in association with the SPARC expression, positive expression of DNAJC12 showed a tendency to confer the worst prognosis for patients with TNBC with negative expression of SPARC, indicating that DNAJC12 protein expression has potential to identify TNBC patients with worse prognosis in different contexts. In prostate cancer cells, DNAJC12 has been shown to act as a cofactor of the chaperone Hsc70 and to be upregulated by the stress-inducing drug A23187 and by the stress-regulated transcription factor AIbZIP/CREB3L4 [52]. In cancer of the rectum, high immunohistochemical expression of DNAJC12 was found to be correlated with poor prognosis for both OS and DFS, and multivariate analysis showed DNAJC12 overexpression is an independent prognostic factor [53]. Also, these authors identified that DNAJC12 overexpression acts as a negative predictive factor for the response to neoadjuvant concurrent chemoradiotherapy [53]. Taken together, these findings indicate DNAJC12 as a potential prognostic and predictive factor in cancer and support the need for future research to elucidate the role of DNAJC12 in TNBC.

LIM family proteins (LMO 1–4) are emerging as key molecules in a wide variety of malignant tumors in humans. In particular, all members, which are required for various developmental processes, are involved in the initiation or progression of different types of neoplasms, including T-cell leukemia, breast cancer and neuroblastoma [54]. There are indications that the gene acts on cell motility, but the mechanisms involved in inducing cell motility, tumor growth, angiogenesis and invasion are still poorly understood [55]. However, the high expression of LIM in breast cancer cells appeared to be linked to angiogenesis and increased the presence of metastases, suggesting an oncogene behavior in this type of tumor [56]. And this corroborates our results, showing that in TNBC with low SPARC expression, LIM overexpression correlated with worse clinical evolution and this seems to be supported by its right relation with processes of cell invasion and angiogenesis [57]. Singh et al. [58] demonstrated that the high expression of LIM family genes function as repressors of the ER-α as the genes are not only linked to this receptor but also to its corepressor, the tumor metastasis antigen 1. They showed that LIM silencing resulted in increased expression of the regulatory genes of ER-α. Consequently, since LIM is highly expressed in breast tumors, negative phenotypes for ER-α may be related to progression and increased tumor aggressiveness [59]. This behavior is also observed in DFS, and OS curves are corroborating a trend to more aggressive biological behavior in TNBC with high expression of LIM, especially when related to tumors that also present low expression of SPARC. Regarding recurrence, the high expression of LIM correlated with a greater number of local and systemic events, although not reaching statistical significance, in comparison to the low expression of LIM. In addition, our results showed the worst OS of patients with a high level of LIM expression. Therefore, inhibition of LIM may be useful in the treatment of TNBC, which does not present specific therapeutic options, as in other breast cancer subtypes, especially those with SPARC⁺/LIM^-.

The main limitations of this study rely on our single-institution retrospective study with the inclusion of a small number of TNBC. We are aware that more reliable studies for the validation of the differentially expressed genes identified in the present study with robust method validation using larger and independent cohorts are required. At present, we are conducting in silico data analysis using the TCGA [60] and METRABIC [18] breast cancer datasets to test the biologically relevant hypothesis with different sets of differentially expressed genes generated by our group. The complete list and analysis of these differentially expressed genes will be reported as part of another study. Also, we are recruiting new patients and beginning collaborative studies with researchers from other institutions to conduct a series of large-scale independent validation studies.

Conclusion

In the present study determining the expression profile of TNBC, with differences in the expression of SPARC and clinical behavior, we have identified SOHLH2, DNAJC12 and LIM1 as potential new biomarkers for TNBC prognosis. Moreover, by previous reports in the literature, our findings indicated that SPARC could regulate gene expression in breast tumors. This effect is possibly due to the ability of SPARC to act on extracellular matrix remodeling that modulates several cell-signaling pathways that impact on gene expression [61–63]. Further clinical and experimental studies are required to determine the impact of the SPARC-modulated genes in breast cancer progression and their possible use in target therapy. Additionally, our findings illustrate the importance of studying specific subgroups of very well-characterized TNBC about clinical behavior. Importantly, our findings provide new informative data on the expression profiling potentially associated with TNBC that may serve as a potential tool for the identification and characterization of new candidate biomarkers with potential to be used as prognostic and predictive factors for patients with TNBC in the future.

Summary points

• Triple-negative breast cancer (TNBC) tumors represent a unique subtype of breast cancer and have been indicative of poorer prognosis since its onset.

• cDNA microarray analysis allowed us to determine the expression signature of TNBC with differences of secreted protein acidic and rich in cysteine (SPARC) expression and clinical behavior.

• We have identified transcriptional changes that could better characterize differences in the clinical behavior of TNBC with negative SPARC expression and positive SPARC expression.

• Three of the genes identified as differentially expressed, SOHLH2, DNAJC12 and LIM1, were further validated using immunohistochemistry on tissue microarrays containing TNBC.

• We found that positive expression of SOHLH2 was related to events of recurrence.

• Positive immunohistochemical expression of DNAJC12 was associated with poorer prognosis in TNBC patients.

• A more aggressive biological behavior was observed in TNBC with high expression of LIM1.

• Our study provides valuable information on new biomarker candidates with potential for better classification and prognosis prediction of TNBC.

Financial & competing interests disclosure

This work was supported by Fundação de Amparo a Pesquisa do Estado de São Paulo (FAPESP; 2013/07035–4) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; 303134/2013–5). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

The authors state that for investigations involving human subjects, informed consent has been obtained from the participants involved.

Data deposition

The data generated in the present study have been deposited in NCBI’s gene Expression Omnibus [36] and are accessible through GEO Series accession number GSE98931 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE98931).

Acknowledgements

The authors gratefully acknowledge Yuri Moreira for microarray analysis support.

Supplementary data

To view the supplementary data that accompany this paper please visit the journal website at: www.tandfonline.com/doi/suppl/10.2217/bmt-2017-0019

Additional information

Funding

This work was supported by Fundação de Amparo a Pesquisa do Estado de São Paulo (FAPESP; 2013/07035–4) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; 303134/2013–5). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.