Dynamic contrast-enhanced MRI predicts PTEN protein expression which can function as a prognostic measure of progression-free survival in NPC patients

The objective of our study was to investigate whether a phosphatase and tensin homolog deleted on chromosome 10 (PTEN) expression was associated with dynamic contrast-enhanced MRI (DCE-MRI) parameters and prognosis in nasopharyngeal carcinoma (NPC). Two-hundred-and-forty-five (245) patients with NPC who underwent pretreatment biopsy, expression of PTEN detected by immunohistochemistry of biopsy, and radical intensity-modulated radiation therapy (IMRT) with or without chemotherapy were included. Tumor segmentations were delineated on pretreatment MRI manually. The pharmacokinetic parameters (Ktrans, Kep, Ve, and Vp) derived from dynamic contrast-enhanced MRI (DCE-MRI) using the extended Toft’s model within the tumor segmentations were estimated. The following demographics and clinical features were assessed and correlated against each other: gender, age, TNM stage, clinical-stage, Epstein–Barr virus (EBV), pathological type, progression-free survival (PFS), and prognosis status. DCE parameter evaluation and clinical feature comparison between the PTEN positive and negative groups were performed and correlation between PTEN expression with the PFS and prognosis status using Cox regression for survival analysis were assessed. A significantly lower Ktrans and Kep were found in NPC tumors in PTEN negative patients than in PTEN positive patients. Ktrans performed better than Kep in detecting PTEN expression with the ROC AUC of 0.752. PTEN negative was associated with later TNM stage, later clinical-stage, shorter PFS, and worse prognosis. Moreover, N stage, pathological type, Kep, and prognostic status can be considered as independent variables in discrimination of PTEN negative expression in NPCs. PTEN negative indicated a shorter PFS and worse prognosis than PTEN positive in NPC patients. Ktrans and Kep derived from DCE-MRI, which yielded reliable capability, may be considered as potential imaging markers that are correlated with PTEN expression and could be used to predict PTEN expression noninvasively. Combined radiological and clinical features can improve the performance of the classification of PTEN expression.


Introduction
Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is a tumor suppressor located on chromosome band 10q23 that regulates proliferation and differentiation in a variety of cancer cell types, including nasopharyngeal carcinoma (NPC) cells (Aquila et al. 2020;Rasheed et al. 1997). This locus encodes a dual-specificity phosphatase, with the PTEN protein has both lipid and protein phosphatase activity at its N-terminal domain, and a C-terminal C2 domain which can control cell migration in a calcium Gang Wu, Weiyuan Huang and Junnv Xu contributed  Extended author information available on the last page of the article independent manner facilitated by the C-terminal PDZ-binding motif (Raftopoulou et al. 2004). PTEN contains lipid phosphatase activity that regulates various cellular processes and signaling pathways, such as the induction of apoptosis through phosphatidylinositol 3-kinase (PI3K)/Akt pathway inhibition and control of cell adhesion, migration, and tumor invasion, by downregulating the activity of focal adhesion kinases (FAKs) (Kim et al. 2020;Lyu et al. 2015). PTEN possesses a 50 amino acid C-terminal tail that is nonessential for activity but influences stability and can act to regulate and inhibit PTEN functionality (Vazquez et al. 2000). Inactivated or mutated PTEN also activates AMPK/mTOR/ HIF1 pathway to promote angiogenesis and ultimately to promote the growth of NPC cells (Lyu et al. 2018). Inactivation of PTEN or low PTEN expression is associated with lymph node metastasis, advanced stage, and poor prognosis in NPCs (Cai et al. 2015a, b;Xu et al. 2004). Activation of PTEN may therefore be a promising approach for NPC tumor prevention and treatment target. Further, knowledge of a tumor possessing PTEN negative acquired mutations, being that this is an indicator of disease severity and progression, could help shape treatment regime. Multiparametric MRI plays an important role in the detection, localization, staging, and predictive prognosis of NPC. The potential of various MRI techniques for the assessment of tumor aggressiveness and response to treatment regimes has recently been investigated (Liu et al. 2021;Zhang et al. 2017;Huang et al. 2013). Furthermore, as an integral part of clinical practice, MRI can characterize a diverse spectrum of tumoral phenotypes as potential biomarkers of genetic status. In particular, a sequence such as dynamic contrastenhanced MRI (DCE-MRI) has the potential to provide parameters that may correlate and be predictive of levels of aggressiveness and genetic status of the tumor microenvironment (Huang et al. 2013). Quantitative DCE-MRI parameters can be easily derived for each pixel within the tumor using commercially available software and can be visualized in a separate image for each parameter (Switlyk et al. 2019). This makes DCE-MRI a powerful prognostic tool. However, the genomic profiles underlying these images are not known, and the potential of quantitative DCE-MRI parameters as a surrogate or replacement for genomic markers and genetic assessment has not been extensively investigated with correlative and comparative biopsy tissue analysis (Switlyk et al. 2019). Imaging genomics is a novel field in cancer research that links and connects imaging parameters with genomic profiles (Liu et al. 2021;Meyer et al. 2019). Imaging methods that are found to be associated with gene expression patterns from tumor biopsies can then be used as noninvasive surrogate markers for genomic profiling and indicate potential information on tumor progression and prognosis (Xie et al. 2019). K trans derived from DCE has been shown to discriminate HIF expression level in NPC (Liu et al. 2021;Surov et al. 2019). However, to the best of our knowledge, no study has been reported to focus on whether DCE can detect PTEN protein expression in NPC. Based on a previous knowledge and the following scientific hypothesis, we aim to explore the value of DCE in differentiating positive or negative PTEN expression in NPC patients and evaluate the correlation of PTEN expression status and clinical prognosis further.

Patients
All consecutive newly diagnosed NPC patients referred to the Department of Radiotherapy at Hainan General Hospital (The Affiliated Hainan Hospital of Hainan Medical University) between May 2018 and October 2020 were included in this prospective study approved by the hospital IRB (NO.2018025). All participating subjects were formally informed about the purpose of this study and a letter of consent was signed by every subject involved. The T, N, M, and clinical staging were performed according to the American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) staging system for NPC (8th edition, 2016) (Amin et al. 2016). Patients met the inclusion criteria by (1) having a Karnofsky performance status ≥ 70%, (2) having NPC as confirmed by biopsy and histopathology, (3) performed PTEN immunohistochemical staining before treatment, and (4) having undergone DCE-MRI before treatment in our hospital. Patients were excluded if they (1) had MRI (e.g., artificial cochlea, cardiac pacemaker implantation) or radiotherapy contraindications, (2) failure to complete treatment or loss of follow-up less than 3 months, (3) had previously received radiotherapy in the neck region, (4) had any other malignant tumors in the prior 5 years, or (5) suffered from severe neurological or psychiatric diseases.
Clinical endpoints were progression-free survival (PFS) for follow-up status assessed on 1 Jan 2021, using the last MR follow-up as the endpoint. For PFS, time from diagnosis to disease-related death, the first event of relapse, or clinical follow-up endpoint was used. Patients suffering from disease-related death or relapse were categorized under "poor prognosis", others were categorized under "good prognosis".

Treatment regimen
The choice of treatment regimen was according to the National Comprehensive Cancer Network (NCCN) Guidelines (https:// www. nccn. org/) (Colevas et al. 2018). All patients were treated by radical three-dimensional conformal and intensity-modulated radiation therapy (IMRT). Radiotherapy was completed using the following equipment: three-dimensional treatment planning system (Eclipse, American), 3000A mobile laser positioning system (Gammex A), Trilogy medical linear accelerator 6MVX line (Varian), 16-slice spiral CT (Siemens), supporting vacuum bag fixing device, and three-dimensional fixing frame. The patients were in a supine position and fixed with a thermoplastic memory membrane in the head and neck. Positioning CT was performed using a thin-layer continuous scanning scheme with 3 mm thickness, tube voltage 140 kV, current 280 mAs. The images were input into Eclipse three-dimensional treatment planning system. The range of tumors was determined by MR images, and the target area was delineated and planned by fused MR and positioning CT images. The planning target volume (PTV) formed by expanding 3-5 mm from each target area was administered a prescription dose: nasopharyngeal tumor volume (GTVnx) and cervical lymph node volume (GTVnd) at 66-73 Gy and 63-70 Gy, respectively; high-risk area of primary focus (CTV1) at 62-64 Gy; low-risk area of primary focus and cervical lymph node drainage area (CTV2) at 54-58 Gy; PTV1 and PTV2 at 54 Gy and 50 Gy, respectively. The number of segmentations was 31-35 times, 5 times per week.
Induction chemotherapy consisted of docetaxel and cisplatin and were given at 75 mg/m 2 intravenously on day 1, or 25 mg/m 2 intravenously on days 1-3, once per 3 weeks lasting for 2-4 rounds. Concurrent Chemotherapy consisted of Cisplatin 75 mg/m 2 on day 1 (or 25 mg/m 2 days 1-3) once per 3 weeks.
Follow-up assessments were performed every 3 months for the first year, every 6 months for years 2-5, then every year subsequently. Routine follow-up assessments included assessment of vital signs and any adverse events. MR scans were done at every follow-up time point.

MRI acquisition
MRI examinations were performed on a 3.0-T scanner (Skyra, Siemens Medical Solutions, Erlangen, Germany), equipped with a 20-channel sensitivity-encoding combined head and neck coil. Patients underwent routine MRI, including T1-weighted spin-echo sequence (field of view [FOV], 180 × 180 mm 2 ; section thickness, 4 mm; repetition time [TR]/echo time [TE], 625/9.0 ms), T2-weighted fast spin-echo sequence (FOV, 180 × 180 mm 2 ; section thickness, 4 mm; TR/TE, 4070/30 ms). Baseline T1-mapping was obtained before DCE-MRI with 5 different flip angles, including 3°, 6°, 9°, 12°, and 15°, using a fast low angle shot (FLASH)/vibe sequence. All other parameters were held constant as the following: DCE-MRI was performed with a (FLASH)/vibe sequence, 50 dynamic acquisitions, 4.9 s per dynamic acquisition, with the following parameters: TR/TE, 4.09/1.47 ms; flip angle, 9; matrix, 192 × 144; FOV, 180 × 180 mm 2 ; section thickness, 4 mm; phase, 75%; bandwidth, 400 Hz. Contrast enhancement was performed with a standard 0.1 mmol/kg body weight dose of gadolinium-based contrast agent of Gadodiamide (Omniscan, GE Medical Systems, Amersham, Ireland). The gadoliniumbased contrast agent was administered through a catheter in the antecubital vein by an automatic power injector (Medrad, Pittsburgh, Pennsylvania, USA) at a rate of 2 mL/s and was followed by a double bolus injection of isotonic saline. The dynamic acquisition was performed with a temporal resolution of 4.9 s, and the contrast agent was administered after 5 baseline dynamics. After the DCE-MRI, contrast-enhanced T1-weighted spin-echo sequences were conducted on axial, coronal, and sagittal planes.

MRI post-processing and analysis
DCE-MRI data were analyzed using the OmniKinetics (OK) software (GE Pharmaceutical). The DCE parameters (K trans , reflux rate [K ep ], the volume fraction of the extravascular extracellular matrix [V e ], and blood plasma volume [V p ]) were estimated using the extended Toft's two-compartment model (Tofts et al. 1999) and population-averaged AIF (Huang et al. 2013).
Two independent radiologists, (reader I (Weiyuan) and II (Wenzhu); 12 and 5 years of experience in MRI, respectively), who were blinded to clinical information, performed tumor segmentation to encompass whole tumor volume on anatomic reference images; among axial T1WI, PdWI, and post-contrast T1WI. Reference image and DCE parameter maps were loaded into a multimodality tumor tracking application ITK-SNAP (version 3.4.0, USA, http:// www. itksn ap. org) (Yushkevich et al. 2006). The regions of interest (ROIs) drawn on the anatomic reference images were simultaneously mapped to the corresponding location on the DCE parameter maps. The mean values of K trans , K ep , V e , and V p from the volumetric ROIs were recorded for correlations and analysis.

PTEN protein expression detecting by immunohistochemistry
In every patient, the diagnosis was confirmed by tumor biopsy. The pathological type was classified according to the cell type and degree of differentiation: undifferentiated non-keratinized carcinoma, differentiated non-keratinized carcinoma, or keratinizing squamous cell carcinoma (Yushkevich et al. 2006). Immunohistochemistry for PTEN protein was performed for all obtained biopsies for all cases used in this study, and the loss of protein expression was defined as an absence of cytoplasmic staining around a counterstained nucleus. Positive controls were obtained on a per biopsy basis from surrounding unaffected stroma and vasculature. After tissue sections were deparaffinized and rehydrated, tissue was sectioned into 4 μm thick slices were treated with antigen retrieval buffer (S1699, Dako) in a steamer for 20 min. Anti-PTEN antibody [100 μL (product number 9188, Cell Signaling Technology)] was applied to tissue sections for overnight incubation at 4 °C in a humidity chamber. After a Tris-buffered saline (PBS) wash, tissue sections were incubated with biotinylated monoclonal anti-rabbit IgG [100 μL (product number ZA-0635, ZSGB-BIO)] for 20 min at 37 °C. The antigen-antibody binding was detected by a kit (Vectastain Elite ABC kit, product number PK-6100, Vector Laboratories) and a DAB (3, 3-diaminobenzidine) system (product number K3468, Dako). Tissue sections were briefly immersed in hematoxylin for counterstaining and were covered with cover glasses. PTEN expression was scored by the pathologist (Y.W.) based on the percentage of positive staining tumor cells (Nucleus/cytoplasm). PTEN staining in the adjacent stroma and blood vessels served as a positive internal control, due to heterogeneity within the tumor focus. To account for this heterogeneity, foci were scored as PTENpositive (≥ 95% cancer cells positive), and PTEN-negative (< 5% cancer cells positive) (Fig. 1).

Statistical analysis
Patients' characteristics were compared between the PTEN positive and PTEN negative groups. Continuous variables were analyzed using the independent sample t-test or Mann-Whitney U test, and categorical variables were analyzed using a Chi-square test or Fisher's exact test according to the data distribution, respectively. For continuous variables with a p-value less than 0.20, a receiver operating characteristic (ROC) curve was constructed and the area under the curve (AUC) was calculated. Optimal cutoff points were determined based on the maximum Youden index. Spearman correlation analyses (correlation coefficient, ρ) were performed to assess the correlation between PTEN positive or negative and clinical or radiographic features. Cox survival analyses were performed to evaluate the prognostic significance of PTEN expression in NPC patients. DCE parameters' values, gender, age, EBV, T, N, M stage, PFS, and prognosis status were used as covariates to evaluate the independent detecting value of the classifier using multiple variables logistic regression.
All statistical analyses and graphing were performed using Prism GraphPad Software version 9.0 (GraphPad Software, La Jolla, CA, USA. https:// www. graph pad. com/ scien tific-softw are/ prism/), and a p-value less than 0.05 was considered statistically significant.

Demographics and clinical features
A total of 273 NPC patients were enrolled. After verifying exclusion criteria, 245 patients remained. The mean age was 50.09 (± 11.41) years-old, 65.31% males. The PTEN expression type accounted for 87.35% of patients with positive, whereas 12.65% corresponded to negative (Fig. 1). T, N, M stage and clinical-stage showed a significant difference between the PTEN positive and PTEN negative groups in comparison of demographic and clinical features (Table 1)  almost perfect interobserver agreement, the average of both readers' measurements were used. K trans (p < 0.001) and K ep (p < 0.001) were significantly lower in PTEN positive group than PTEN negative group. V e (p < 0.994) and V P (p < 0.074) could not distinguish PTEN positive from negative. Spearmen correlation analyses also confirmed the same results. PTEN expression type was associated significantly with K trans (ρ = − 0.290, p < 0.001) and K ep (ρ = − 0.241, p < 0.001) (Table 2; Fig. 2). ROC analysis showed K trans distinguishing PTEN positive from negative with an AUC of 0.752 (95% CI 0.674-0.829), an 80.65% sensitivity, a 60.75% specificity using 0.85 as the cutoff. K ep performed worse than K trans for classifying PTEN expressions with an AUC of 0.629 (95% CI 0.524-0.734), a 64.52% sensitivity, a 62.62% specificity using 2.068 as the cutoff (Fig. 3). Representative cases of PTEN positive and PTEN negative groups were shown in Fig. 2.

PTEN positive associated with better prognosis than PTEN negative NPCs
The prognostic significance of the PTEN expression type was evaluated based on the PFS of NPC patients. In the PTEN negative group, a significantly shorter PFS, and lower cumulative incidence of recurrence and death have been observed (Table 1; Fig. 3). Multiple variables logistic regression results are based on multivariate analysis using N stage, pathological type, K ep and prognostic status considered as associated variables from the earlier section analysis in the classification of PTEN expression. The multivariable logistic regression model has a good performance in detecting PTEN expression with an AUC of 0.843 (95% CI 0.762-0.924), a 84.62% negative predictive power, and a 91.38% positive predictive power (Fig. 3).

Discussion
Accumulating evidence suggests that PTEN is an independent prognostic factor in patients with NPC; PTEN negative patients are more likely to have worse prognosis. Considering its prognostic significance, this pilot study implemented a comparison with quantitative parameters derived from DCE-MRI firstly, which accurately correlates with and thus predicts PTEN expressions in NPC. This study further confirms that PTEN negative indicates a shorter PFS and worse prognosis than PTEN positive NPC patients. The multivariable regression model combined clinical and image features could provide more accurate predictability of PTEN expression.
PTEN is one of the tumor suppressors whose positive expressions have been found in head and neck squamous cell carcinoma (HNSCC) commonly, with low negative rates of approximately 10-30% (Pfisterer et al. 2015;Costa et al. 2018). Our dataset showed 87.35% positive and 12.65% negative PTEN in NPCs, with an expression rate close to HNSCC. The reported PTEN positive rates of other types of malignant tumors, such as colorectal cancer (Li et al. 2010) or prostate cancer (Eineluoto et al. 2020), were lower than NPCs. This may be a contributing factor to the longer 5-year survival rate and overall survival of NPC patients than other common types of malignancies.
T, N, M stage, and clinical-stage showed a significant difference between the PTEN positive and negative groups among all demographic and clinical features. We demonstrated that higher TNM or clinical stages have a higher PTEN negative rate in NPC tumors. This suggests that PTEN has a significant contribution to TNM stage and clinical-stage in NPC tumors. Previous studies revealed similar results regarding either NPC (Huang et al. 2013), or colorectal cancer (Kim et al. 2020), gastric cancer (Wen et al. 2010), and hepatocellular carcinoma (Fu et al. 2017). Several studies showed protein expression of PTEN was low in EBV positive NPC (Wu et al. 2018;Peng et al. 2016). Our study shows that expression of PTEN did not show a significant difference between the EBV positive and negative groups of NPC patients, needing further study to confirm. This inconsistent result could be due to the small number of patients in the subgroup of negative EBV and PTEN in NPCs in our study.
Additionally, we explore the value of DCE in differentiating PTEN positive from PTEN negative status in NPC patients. DCE-MRI can quantitatively characterize tumor perfusion and is associated with a microvessel density and permeability. Several studies have reliably estimated  (Liu et al. 2021;Xie et al. 2019). It has been shown that different DCE parameters reflected different aspects of tumor microstructure (Yuan et al. 2012;Nejad-Davarani et al. 2017). As the functionality of PTEN is that of a tumor suppressor which is known to influence cell migration, growth rate regulation, and various other processes that lead to tumorigenesis, it is likely that the pathways that are altered due to PTEN negative status cause definite and drastic tumor microenvironmental changes that are gathered by DCE-MRI and can be predictive of PTEN status. At present, we found that K trans and K ep derived from DCE Toft's model had the discriminative ability to detect PTEN expression status. K trans (min −1 ) reflects contrast agent flow from the vascular space to the extravascular extracellular space (EES), which was the most commonly used parameter in Toft's model, representing vessel permeability (Yuan et al. 2012;Nejad-Davarani et al. 2017). Prior studies have correlated K trans to measures of tumor vessel permeability or oxygenation levels such as HIF-1α (Liu et al. 2021). Referring to our results, we speculate that, without the inhibition of the suppressor gene PTEN, angiogenesis and aggressiveness of tumors increase significantly. The high permeability of immature neovascular probably caused the increase of K trans in NPCs. K ep (min −1 ) is the reverse reflux rate constant between the extracellular space and plasma and is equal to K trans /V e . A recent study by McCann et al. (2016) reported there was a significant negative correlation between K ep and PTEN expression in prostate cancer. However, Switlyk et al. (2019) reported that DCE cannot discriminate PTEN expression in prostate cancer. Significant heterogeneity of prostate cancer and differences in the methods for detection of molecular aberrations may be the reason for this disagreement. Further, advancements in signal acquisition could improve resolution that would provide more sensitive measures to correlate with PTEN expression (Baumgartner et al. 2019). NPCs were a homogeneous undifferentiated small round cell solid tumor, tending to have high cellularity and low possibility of necrosis. DCE-MRI might be more *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 suitable for small round cell solid tumors, such as NPC, to evaluate angiogenesis or oncogene/suppressor gene expression. V e might be strongly associated with cellularity because it reflects the amount of extracellular space, as it was exemplarily shown in a glioma model (Ozturk et al. 2019). V p is defined as the blood per unit volume of tissue recommended as an alternative imaging biomarker for the assessment of blood volume (Artzi et al. 2015). Our results showed that V e and V p having limited discriminative ability in predicting PTEN expression, indicating no significant difference in cellularity among NPC tumors with different degrees of angiogenesis and aggressiveness. The failure of the significant correlation between V p and PTEN may be explained by the fact that high vascular permeability induced by inactivation of PTEN can result in a high interstitial pressure in a tumor, which may make microvasculature distort, deform, collapse, and even divert (DeVries et al. 2003;Di et al. 2019). This will reduce the effective blood flow and limit the volume of contrast agent distribution. Di et al. (2019) also failed to detect VEGF expression using V p derived from DCE-MRI. Our ROC analysis showed K trans has a larger AUC than K ep . However, multivariable logistic regression reveals that only K ep out of DCE parameters acts as an independent detecting factor. This result could be due to underlying distributional differences. K ep was analyzed under assumption of normal distribution, however, K trans is not normally distributed. Future research will perform further statistical tests by employing data normalization techniques and standardization across parameters to derive more accurate parameter relatedness.
Furthermore, we evaluate the correlation of PTEN expression status and prognosis and build multiple variable logistic regression model to assess independent detecting metrics. Several studies have suggested that PTEN mutant/ loss function could serve as a prognostic factor in head and neck malignant tumors (Pfisterer et al. 2015;Costa et al. 2018), including NPC (Hu et al. 2019;Li et al. 2019). Our results were comparable to previous studies, PTEN positive NPCs demonstrated a longer PFS and better prognosis status than PTEN negative NPCs using cox survival analysis. Considering NPC has a much higher 5-year survival rate and a better prognosis than other malignant tumors, we defined the clinical endpoint indicating a poor prognosis as either recurrence, metastasis, or death rather than only death during the follow-up. 32 out of 214 PTEN positive NPC patients suffering from a poor prognosis, while 9 out of 31 PTEN negative NPC patients suffering from a poor prognosis. Multiple variable logistic regression showed the N stage, pathological type, K ep and prognostic status can be considered as independent variables in the classification of PTEN expression. We confirmed that combined clinical and radiological features can improve the performance of discrimination of PTEN expression as positive or negative in NPC patients. N stage is a stronger independent variable than T stage which may be related to PTEN inducing lymph node metastasis (Cai et al. 2015a, b).
There are several limitations in this study. First, due to the inherent rarity of PTEN negative in NPC, there is an imbalance of sample size between the two groups. We used a suitable analysis model and correction according to the data distribution to minimize sample size imbalance. Second, PTEN expressions were determined by immunohistochemical rather than sequencing due to a lack of available data. Further studies are needed to verify our results using PCR sequencing. Following this sequencing, it will be important to understand the underlying mutation in PTEN that leads to a nonfunctional phenotype, as C-terminal tail deletion could actually enhance PTEN activity, while many mutations may reduce or remove activity. Comparisons between nascent RNA, mRNA, and DNA sequence in addition to transcript levels will determine the specific form and role of the mutation. There could be a potential upregulation of PTEN in PTEN positive tumors that could act as protective. Thus comparisons between PTEN expression levels between the tumor tissue and control tissue from the same patient will be illuminating towards potential PTEN upregulation. Genome screening will be implemented in future clinical studies to determine whether inherited PTEN mutations exist in patients or the vast majority, as expected, are acquired mutations. Third, extracellular PTEN uptake is possible and further influenced by tissue order and heterogeneity (Putz et al. 2015). Thus in the case of heterogeneous PTEN expression due to acquired mutation, the localization of PTEN positive cells could act as protective to nearby cells if PTEN is further upregulated. Future biopsy sections will be imaged to detect differences in nuclear and cytoplasmic localization of PTEN, along with sequencing data that will unveil mutation type, as cellular localization changes of PTEN could be another prognostic indicator (Bononi and Pinton 2015). Fourth relatively short follow-up period makes it impossible for us to evaluate the correlation of PTEN expressions and 5-year survival rate or long-term prognosis. Finally, MR imaging protocols of our study did not including diffusionweighted imaging (DWI), which could also be considered as a promising tool to provide functional information for tumor assessment. In our study, we only focus on DCE to evaluate the imaging parameter for detecting PTEN protein expression. Further studies will add DWI in MRI protocols to evaluate the potential of DWI in detecting protein expression. Multiparameter models that incorporate information from various contrasts and quantitative imaging methods could provide very robust results in predicting PTEN expression and this will be a future research direction.
In conclusion, the current study applied DCE-MRI to detecting PTEN expression noninvasively in NPC patients. Among DCE quantitative parameters, K trans and K ep , which yielded reliable capability in predicting PTEN expression, may be considered as potential imaging markers. Furthermore, combined clinical and imaging features can improve the performance of classification of PTEN status. PTEN negative expression predicts a less favorable survival outcome in patients with NPC.