High Expression of FGD3, a Putative Regulator of Cell Morphology and Motility, Is Prognostic of Favorable Outcome in Multiple Cancers

Purpose Identification of single-gene biomarkers that are prognostic of outcome can shed new insights on the molecular mechanisms that drive breast cancer and other cancers. Methods Exploratory analysis of 20,464 single-gene messenger RNAs (mRNAs) in the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) discovery cohort indicates that low expression of FGD3 mRNA is prognostic for poor outcome. Prognostic significance of faciogenital dysplasia 3 (FGD3), SUSD3, and other single-gene proliferation markers was evaluated in breast cancer and The Cancer Genome Atlas (TCGA) cohorts. Results A meta-analysis of Cox regression of FGD3 mRNA as a continuous variable for overall survival of estrogen receptor (ER)–positive samples in METABRIC discovery, METABRIC validation, TCGA breast cancer, and Combination Chemotherapy in Treating Women With Breast Cancer (E2197) cohorts resulted in a combined hazard ratio (HR) of 0.69 (95% CI, 0.63 to 0.75), indicating better outcome with high expression. In the ER-negative samples, the combined meta-analysis HR was 0.72 (95% CI, 0.63 to 0.82), suggesting that FGD3 is prognostic regardless of ER status. The potential of FGD3 as a biomarker for freedom from recurrence was evaluated in the Breast International Group 1-98 (BIG 1-98; Letrozole or Tamoxifen in Treating Postmenopausal Women With Breast Cancer) study (HR, 0.85; 95% CI, 0.76 to 0.93) for breast cancer–free interval. In the Hungarian Academy of Science (HAS) breast cancer cohort, splitting on the median had an HR of 0.49 (95% CI, 0.42 to 0.58) for recurrence-free survival. A comparison of the Stouffer P value in five ER-positive cohorts showed that FGD3 (P = 3.8E-14) outperformed MKI67 (P = 1.06E-8) and AURKA (P = 2.61E-5). A comparison of the Stouffer P value in four ER-negative cohorts showed that FGD3 (P = 3.88E-5) outperformed MKI67 (P = .477) and AURKA (P = .820). Conclusion FGD3 was previously shown to inhibit cell migration. FGD3 mRNA is regulated by ESR1 and is associated with favorable outcome in six distinct breast cancer cohorts and four TCGA cancer cohorts. This suggests that FGD3 is an important clinical biomarker.


INTRODUCTION
An increasing collection of breast cancer cohorts have been molecularly profiled on Affymetrix and Illumina platforms, so it is now feasible to conduct an exploratory analysis to identify single-gene biomarkers in which messenger RNA (mRNA) expression is prognostic of outcome. 1 By limiting the exploratory analysis to a single gene, we intended to identify novel gene(s) that might provide insight into biological mechanisms that drive breast cancer metastasis. 2 The starting point for this analysis was the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) data set, which contains clinical traits, expression data, copy number variation profiles, and single nucleotide polymorphism genotypes derived from breast tumors collected from participants in the METABRIC trial. 3 We also used the Genomic Data Commons, which incorporates The Cancer Genome Atlas (TCGA) cancer cohorts and currently spans 21 cancer types suitable for survival analysis. This type of analysis could lead to the discovery of prognostic biomarkers that behave as oncogenic drivers and could potentially be novel therapeutic targets.
In this study, we identified FGD3 mRNA expression as a putative biomarker prognostic of outcome in the METABRIC cohort. FGD3 and SUSD3 4 cell motility genes were compared as a single-gene biomarker with proliferation genes MKI67, 5,6 AURKA, 7-10 and PCNA 11,12 in six distinct breast cancer cohorts and as a pan-cancer biomarker in TCGA cancer cohorts. FGD3 protein expression was evaluated in breast cancer tissue microarrays (TMAs) as an indicator of regional lymph node status. FGD3 mRNA expression as a biomarker for an immune response was evaluated by using tumor-infiltrating lymphocyte cells in TCGA breast cancer and Breast International Group 1-98 (BIG 1-98; Letrozole or Tamoxifen in Treating Postmenopausal Women With Breast Cancer) cohorts. ESR1 transcriptional regulation of FGD3 mRNA expression in the breast cancer cell line ZR-75-1 was confirmed.

Cohorts
Detailed description of each cohort is provided in the Data Supplement.

Survival Analysis
To illustrate the outcome benefit of low versus high expression, the Kaplan-Meier method 13 was used to estimate the distribution of timeto-disease outcome end points by gene expression status bifurcated on the cohort mean and median (Data Supplement). An overall inference measure was determined by using a Stouffer weighted Z-test P value 14,15 to combine probabilities from the cohorts. The Wald-test P values from Cox proportional hazards models for the association of cancer outcomes with gene expression as a continuous variable for each cohort were used to determine the Stouffer weighted Z-test P value. BioJava was used to implement the R survival package 16 for analysis. Meta-analysis data for combined hazard ratio (HR) and forest plot figures were generated by using MetaXL 17 (http://www.epigear.com/).

FGD3 Expected Tissue Cell Profile and Tumor-Infiltrating Lymphocyte Association
Faciogenital dysplasia 3 (FGD3) protein is found to have high expression in immune response cells according to curated data in 79 human and 61 mouse tissues from the GeneAtlas 18 using BioGPS. 19 Tumors with high expression of FGD3 and favorable outcome could indicate an immune response. Tumor-infiltrating lymphocytes (TILs) were called in 389 samples from TCGA breast cancer high-resolution slide images using methods previously defined by the International TIL Working Group 2014. 20 TILs were called using a similar methodology in 725 samples from the BIG 1-98 DASL cohort, and correlation to FGD3 mRNA expression was determined by using Pearson's correlation coefficient (r). 21

FGD3 Protein Expression Analysis
FGD3 protein expression in tumor samples was determined by immunohistochemical (IHC) staining. Breast cancer samples were provided by Avera Cancer Institute, and breast cancer TMAs (BR1504a, BR1505b, HBre-Duc068Bch-01, and BR20837) were purchased from US Biomax (Rockville, MD).
FGD3 protein expression was quantitatively determined in the range of 0 to 4. The US Biomax metadata indicates whether the patient had no regional lymph node metastasis (N0) or degrees of metastasis in regional lymph nodes (N1 to N3). FGD3 expression levels for samples with N0 designation (n = 135) and samples with N1 to N3 designations in metastatic tissue (n = 98) were compared by using an unpaired t test. Tissues for matched lymph node metastasis (n = 103) were compared with primary tumor tissues using unpaired t test, and figures were generated using GraphPad Prism 7.0. Additional details on analysis can be found in the Data Supplement.

FGD3 mRNA Expression Regulated by Estrogen Receptors
Breast cancer cell line ZR-75-1 was grown in RPMI-1640 medium with 10% fetal bovine serum. For the treatment of estrogen, cells were deprived of hormone for 3 days in phenol-free RPMI-1640 medium with 5% charcoal-stripped fetal bovine serum and then treated with either ethanol (vehicle) or 1 nM 17b-estradiol (estrogen) for 16 hours. Reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR) was performed and is described in the Data Supplement.

mRNA Expression of Genes of Interest With Published Data
The Expression Atlas Web site 22 was used to query RNA sequencing (RNA-seq) expression levels in breast cancer cell lines. In addition, the Web site was used to search for cancer cell line experiments with a 2.0 or greater fold change for FGD3, and downloaded experimental data were used to illustrate differences in differential expression of the proliferation genes using Prism 7.0.

Discovery of FGD3
We undertook an exploratory analysis of 20,464 possible single-gene biomarkers as categorical variables split on the mean in the METABRIC discovery cohort and identified FGD3 mRNA expression as the highest ranked prognostic gene based on the P value for overall survival (OS), which we subsequently verified as being prognostic in the METABRIC validation cohort (data not shown). A detailed summary of the genes in this article that were used for Cox models as continuous and categorical variables, Kaplan-Meier figures split on cohort means, and expression profiles can be found in the Data Supplement.

FGD3 Breast Cancer Prognostic Marker
IHC estrogen receptor-positive cohorts. The results of the Cox regression analysis of FGD3 mRNA expression as a continuous variable in five distinct estrogen receptor (ER)-positive breast cancer cohorts are shown in Figure 1A ( 1.20) and was not prognostic in any individual cohort (Data Supplement). The Stouffer P value was used to rank the prognostic significance of each gene in four distinct ER-negative cohorts: FGD3 (P = 3.88 E-5 ), PCNA (P = .063), SUSD3 (P = .319), MKI67 (P = .477), and PCNA (P = .820).
Hungarian Academy of Science breast cancer cohort. The Hungarian Academy of Science (HAS) breast cancer cohort represents a collection of all publicly available breast cancer cohorts profiled on the Affymetrix platform. The Kaplan-Meier Plotter Web interface allows selection of HAS cohorts on the basis of ER status using the mRNA expression level. FGD3 is prognostic in the HAS ER-positive cohort (HR, 0.61; 95% CI, 0.5 to 0.74; P = 6.4 E-7 ) and in the ER-negative cohort (HR, 0.4; 95% CI, 0.3 to 0.53; P = 1.8 E-11 ) when split on the median (Fig 2A-   To further investigate whether FGD3 mRNA expression is a feature of the tumor, breast cancer TMAs were purchased from US Biomax, and IHC was used to determine FGD3 protein expression levels (scored from 0 to 4). FGD3 protein expression was determined to be a feature of the tumor and was not associated with the presence of lymphocytes. By using metadata provided by US Biomax, FGD3 protein levels in tumors (n = 135) with no regional lymph node metastasis (N0) were compared with tumors with lymph node metastasis (N1 to N3) and corresponding matched metastatic tissues (Fig 4). An unpaired t test comparing N0 with N1 to N3 suggests that lymph node metastasis is associated with lower FGD3 protein levels (P , 1 E-4 ). An unpaired t test comparing FGD3 protein expression in N1 to N3 primary tumor samples with lymph node metastatic tissue samples suggests that lower FGD3 protein level indicates metastasis (P = .142). Representative images of FGD3 IHC staining are shown in Figure 5 and the Data Supplement. Benign tumors ( Fig 5A) and breast adenocarcinomas in lower stages (Fig 5B-C) showed strong expression of FGD3, whereas late-stage breast adenocarcinomas in higher stages (Fig 5D-F) 29 cBioPortal was used to query co-expression relationships in the META-BRIC cohort, which showed a tendency toward cooccurrence with P , .001 and a log odds ratio of 1.52. The TCGA breast cohort showed a tendency toward co-occurrence with P , .001 and a log odds ratio of 2.3 in TMA expression data. These data suggest that ESR1 plays a regulatory role in FGD3 mRNA expression. Surprisingly, FGD3 mRNA expression in BT-20, a triple-negative breast cancer cell line, was upregulated to 1.99 and ESR1 was upregulated to 0.47 on a log2 scale when treated with an EGFR inhibitor (Data Supplement). 30

DISCUSSION
The exploratory analysis of 20,464 possible singlegene biomarkers in the METABRIC discovery cohort identified FGD3 as a highly prognostic biomarker. Minimal research has focused on it as a possible driver of metastasis in breast cancer, and a pan-cancer analysis in TCGA cohorts found that FGD3 mRNA was a putative prognostic biomarker in other cancers. This is a significant finding in the TCGA cohorts, considering the median survival time in these cohorts was typically less than 2 years.
FGD3 has a putative guanine nucleotide exchange factor that targets cell division control protein 42 homolog (CDC42) 31 and shares high sequence similarity with FGD1 in their Dbl homology (70%) and pleckstrin homology (60.6%) domains; however, FGD3 lacks the N-terminal proline-rich domain found in FGD1. The proline-rich domain plays a crucial role in the response to extracellular signal-responsive translocation of FGD1 to the leading-edge membrane in cells facing toward a wound during the wound-healing process. 32 Through a highly conserved recognized destruction motif (DSGIDS) in both FGD3 and FGD1 as well as in other unrelated proteins including IkBs and b-catenin, downregulation occurs through the ubiquitin/proteasome system via phosphorylation by GSK-3b of the serine residues in the DSGIDS motif. 33 Thus, both FGD3 and FGD1 intracellular levels are tightly regulated by the same destruction pathway. Functionally, FGD1 is involved in the regulation of cellular structures required for invadopodia biogenesis and extracellular matrix degradation in an invasive cell model by modulating Cdc42 activation. [34][35][36][37] In addition, mutations in FGD1 are responsible for the X-linked disorder known as faciogenital dysplasia, and FGD1 is highly expressed during bone growth and mineralization. 36 Using the HeLa Tet-Off cell system, Hayakawa et al 31 showed that notwithstanding their similarity, FGD3 and FGD1 played quite different roles in regulating cellular functions. They found that full-length FGD3 could induce and activate Cdc42. Furthermore, inducible expression of FGD3 resulted in significant morphologic changes that included the formation of broad sheet-like protrusions known as lamellipodia when GTPbound Cdc42 levels were significantly increased by the inducible expression of FGD3, whereas high expression of FGD1 led to the formation of filopodia, which are found in migrating cells. 31 Thus, cell motility seemed to be inversely regulated by FGD3 and FGD1: FGD3 inhibited cell migration and FGD1 stimulated it.
The FGD3-SUSD3 metagene (these genes share the same chromosomal location directly adjacent to each other at Chr9q22.31) was ranked with the highest concordance index 38 in the Sage Bionetworks-DREAM Breast Cancer Prognosis Challenge and was a key biomarker presented by N0, primary tumor with no indication of metastasis to regional lymph nodes (n = 135); N1 to N3, metastatic primary tumor with one to three positive regional lymph nodes (n = 98); and tumor tissue for matched lymph node metastasis (n = 103).
the group submitting the winning model. 39 Using the METABRIC data set, 3 they determined that the expression value of two genes, FGD3 and SUSD3, was the most prognostic molecular metagene marker associated with a good prognosis, and they referred to it as a protective metagene because it displayed a CI that was significantly less than 0.5. They also validated the poor prognosis associated with low expression of the FGD3-SUSD3 metagene in the OsloVal data set (described in Liu et al 40 ). The prognostic significance of FGD3 and SUSD3 as single gene prognostic biomarker using Cox regression models in a large collection of breast cancer cohorts and TCGA cohorts has not been published.
In a follow-up study, the group developed a breast cancer prognostic test, Breast Cancer Attractor Metagenes (BCAM), which had several molecular features, including the breast cancer-specific FGD3-SUSD3 metagene, as well as tumor size, and number of positive lymph nodes. 41 Notably, Ou Yang et al 41 went on to suggest that breast cancer subtype classification as well as hormone receptor and HER2 status did not add additional prognostic information when expression levels of the FGD3-SUSD3 metagene and the attractor metagenes were known and taken into consideration.
In a similar manner, SUSD3 expression was found to be regulated by ESR1 in MCF-7 cells through direct interaction of E2 with its regulatory region. Experiments in MCF-7 cells further showed that SUSD3 was implicated in E2-mediated cell proliferation, adhesion, and migration. 4 However, as with FGD3, the role that SUSD3 plays in ER-positive breast cancer has not been fully established.
On the basis of normal tissue expression profiles, FGD3 is highly expressed in T cells, natural killer cells, monocytes associated with immune response, and myeloid whole blood cells. A characteristic of these cell types is that they are mobile, and evidence from the HeLa Tet-Off wound healing assay suggests that high expression of FGD3 limits cell mobility. 31  wide range of treatment options. It has been implicated as a gene that regulates cell migration in the progression of cancer. Comparing the prognostic value of FGD3 in breast cancer with the prognostic value of genes associated with proliferation such as MKI67, PCNA, and AURKA indicates that FGD3 may offer superior disease progression metrics in all clinically relevant breast cancer subtypes (Fig 6A). Overall, AURKA is the most prognostic gene in the TCGA cancer cohorts in which the median time of 2 years suggests that it is an indicator of early relapse as measured by OS (Fig 6B). FGD3 is prognostic in six TCGA cohorts, and AURKA is prognostic in nine TCGA cohorts. MKI67, PCNA, and AURKA are largely prognostic in the same cohorts with renal papillary cell carcinoma, lower-grade brain glioma, renal clear cell carcinoma, pancreatic adenocarcinoma, and lung adenocarcinoma. FGD3 is uniquely highly prognostic in head and neck squamous cell carcinoma and lung adenocarcinoma. FGD3 is prognostic in breast invasive carcinoma, cervical squamous cell carcinoma, sarcoma, and bladder urothelial carcinoma. A PubMed search for FGD3 and cancer mentioned in the abstract resulted in one publication. 41 Repeating the PubMed search for the other proliferation genes resulted in the following publication metrics (MKI67, 92; KI67, 3,104; PCNA, 2,741; and AURKA, 284).
The key differentiator of FGD3 as a putative biomarker is that high expression indicates favorable outcome; for other established proliferation biomarkers, high expression of MKI67, KI67, PCNA, and AURKA are prognostic of poor outcome.
The availability of a growing collection of cancer cohorts with outcome data has allowed for the confirmation of the clinical importance of FGD3 as a prognostic biomarker and implications that can guide treatment. Given the obvious cohort differences in treatment conditions, patient populations, formalin-fixed paraffin-embedded frozen tissue, and platform differences in Illumina and RNA-seq, FGD3 represents a robust indicator of outcome in breast cancer as well as other cancers and requires further study.