**High Risk of Deep Neck Infection in Patients with Type 1 Diabetes Mellitus: A Nationwide Population-Based Cohort Study**

**Geng-He Chang 1,2,3,†, Meng-Chang Ding 1,†, Yao-Hsu Yang 2,4,5, Yung-Hsiang Lin 6, Chia-Yen Liu 2, Meng-Hung Lin 2, Ching-Yuan Wu 4,5, Cheng-Ming Hsu 1,5 and Ming-Shao Tsai 1,2,3,\***


Received: 30 September 2018; Accepted: 23 October 2018; Published: 25 October 2018

**Abstract:** Objective: To investigate the risk of deep neck infection (DNI) in patients with type 1 diabetes mellitus (T1DM). Methods: The database of the Registry for Catastrophic Illness Patients, affiliated to the Taiwan National Health Insurance Research Database, was used to conduct a retrospective cohort study. In total, 5741 patients with T1DM and 22,964 matched patients without diabetes mellitus (DM) were enrolled between 2000 and 2010. The patients were followed up until death or the end of the study period (31 December 2013). The primary outcome was the occurrence of DNI. Results: Patients with T1DM exhibited a significantly higher cumulative incidence of DNI than did those without DM (*p* < 0.001). The Cox proportional hazards model showed that T1DM was significantly associated with a higher incidence of DNI (adjusted hazard ratio, 10.71; 95% confidence interval, 6.02–19.05; *p* < 0.001). The sensitivity test and subgroup analysis revealed a stable effect of T1DM on DNI risk. The therapeutic methods (surgical or nonsurgical) did not differ significantly between the T1DM and non-DM cohorts. Patients with T1DM required significantly longer hospitalization for DNI than did those without DM (9.0 ± 6.2 vs. 4.1 ± 2.0 days, *p* < 0.001). Furthermore, the patients with T1DM were predisposed to DNI at a younger age than were those without DM. Conclusions: T1DM is an independent risk factor for DNI and is associated with a 10-fold increase in DNI risk. The patients with T1DM require longer hospitalizations for DNI and are younger than those without DM.

**Keywords:** cervical; cellulitis; abscess; deep neck infection; diabetes mellitus

### **1. Introduction**

Deep neck infection (DNI) is a common infectious disease involving the deep neck space; DNI usually requires intensive care and aggressive treatment [1]. The easy availability of antibiotics, improvements in diagnostic technology, and the concept of early surgical debridement have significantly reduced the morbidity and mortality of DNI [2,3]. However, DNI remains a potentially life-threatening disease when lethal complications, such as descending necrotizing mediastinitis, develop [4,5].

A study reported that patients with diabetes mellitus (DM) are at a 1.4-fold higher risk of DNI than those without DM [6]. DNI can cause higher morbidity and mortality among patients with systemic diseases such as DM, end-stage renal disease, liver cirrhosis, and autoimmune diseases [1,4,7–9]. However, the pathogenesis of type 1 DM (T1DM) is different from that of type 2 DM (T2DM). T1DM is characterized by an immune-mediated depletion of beta cells, which causes a lifelong dependence on exogenous insulin [10]. Patients with T1DM, considered to have an immunocompromised status, are expected to be more vulnerable to complicated infection and have a higher infection-related mortality risk than patients with T2DM [11]. Studies investigating the effect of T1DM on DNI are not currently available in the literature. This study investigated the effect of T1DM on DNI occurrence, treatment, and prognosis.

### **2. Methods**

### *2.1. Data Source*

The government of Taiwan established the National Health Insurance Research Database (NHIRD), which covered 99.6% of Taiwan's population in 2017 [12,13]. The NHIRD provides all medical claims data of all beneficiaries, including disease diagnoses during clinic visits and hospitalization, prescription drugs and doses, examinations, procedures, surgery, payments, resident locations, and income levels, generated during reimbursement for insurance in an electronic format. The diagnostic codes in the NHIRD are based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). This study was exempted from obtaining informed consent from the participants because the data were deidentified. All information of the insurants was unidentifiable, and this study did not violate their rights or adversely affect their welfare. The study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (IRB Number: 201601249B1).

### *2.2. Study Cohort*

In Taiwan, T1DM is categorized as a "catastrophic illness" in the NHIRD. Patients with T1DM are certified by the government and included in the Registry for Catastrophic Illness Patients (RFCIP). Therefore, they can avail considerable discounts on medical expenses. The certification process requires critical evaluation of medical records, serological, and pathological reports by experts [14]. Therefore, the T1DM diagnosis of the enrolled patients was highly accurate and reliable.

Data regarding patients who received new diagnoses of T1DM between January 2000 and December 2010 in Taiwan were retrieved from the RFCIP (Figure 1). The patients who received T1DM diagnoses in or after 2011 were not included to ensure a follow-up period of at least 3 years. We used the following T1DM-associated ICD-9-CM codes, which were defined for the RFCIP: 250.01, 250.03, 250.11, 250.13, 250.21, 250.23, 250.31, 250.33, 250.41, 250.43, 250.51, 250.53, 250.61, 250.63, 250.71, 250.73, 250.81, 250.83, 250.91, and 250.93 [14]. In addition, patients who received DNI diagnoses before T1DM were excluded. Finally, 6201 patients with T1DM were enrolled in the study cohort.

**Figure 1.** Enrollment schema of the study and comparison cohorts. Abbreviations: RFCIP, Registry for Catastrophic Illness Patients; LHID2000, Longitudinal Health Insurance Database 2000; DM, diabetes mellitus; T1DM, type 1 diabetes mellitus; DNI, deep neck infection.

### *2.3. Comparison Cohort*

The Longitudinal Health Insurance Database 2000 (LHID2000), a subset database of the NHIRD, consists of 1,000,000 insurants who were randomly statistically selected from all insurants in Taiwan in 2000. Age distribution, sex distribution, or health care costs did not differ significantly between the LHID2000 sample group and all enrollees in the NHIRD, according to a report by the National Health Research Institutes [13]. The LHID2000 has been used in several population-based studies [15,16]. We used the LHID2000 to generate a comparison cohort, which consisted of patients without DM.

### *2.4. Matching Process*

For each patient with T1DM, four patients without DM were randomly selected from the LHID2000 database, matched for sex, age, urbanization level, and income level to form a comparison cohort. The index date of the study cohort was the date of registry in the RFCIP for patients with T1DM, and an index date matching that of patients with T1DM was created for the comparison cohort. After the matching process, 5741 T1DM and 22,964 non-DM patients were enrolled in the study.

### *2.5. Main Outcome*

The main outcome of this study was the occurrence of DNI, which is defined as hospitalization with the following ICD-9 codes: 528.3 (cellulitis and abscess of oral soft tissues; Ludwig angina), 478.22 (parapharyngeal abscess), 478.24 (retropharyngeal abscess), and 682.1 (cellulitis and abscess of neck) [1,17]. The follow-up period was from the index date to the diagnosis of DNI, death, or the end of 2013.

### *2.6. Comorbidities*

Comorbidities were defined using ICD-9-CM codes recorded in the claims data: hypertension (HTN) (ICD-9-CM codes: 401–405), cerebrovascular accident (CVA) (ICD-9-CM codes: 430–438), coronary artery disease (CAD) (ICD-9-CM codes: 410–414), chronic kidney disease (CKD) (ICD-9-CM codes: 403, 404, 585, and 586), systemic autoimmune diseases (SADs) (ICD-9-CM codes: 443.1, 446.0, 446.2, 446.4–446.5, 446.7, 696.0–696.1, 710.0–710.4, and 714.0–714.4), and liver cirrhosis (LC) (ICD-9-CM

codes: 571.2, 571.5–571.6) [1,17–19]. Medical comorbidities were included if they appeared at least once in the diagnoses of inpatients or at least thrice in the diagnoses of outpatients.

### *2.7. Treatment Modalities*

The treatment methods were divided into two subgroups: "surgical" and "nonsurgical." The patients who received surgical intervention were included in the "surgical" subgroup, whereas those who received antibiotic or abscess aspiration without surgery were included in the "nonsurgical" subgroup [1].

### *2.8. Prognosis Evaluation*

For evaluating prognosis, we analyzed the duration of hospitalization, care in intensive care units (ICUs), performance of tracheostomy, and mediastinal complications, which were defined according to the receipt of mediastinal surgery during hospitalization or the diagnostic codes of mediastinitis (ICD-9-CM codes: 510, 513, and 519.2) [1]. Mortality and mediastinitis-related mortality were also investigated in both cohorts. Mortality was defined as death occurring during DNI treatment. Mediastinitis-related mortality was defined as death during DNI treatment accompanied by the diagnosis of mediastinitis [1]. In addition, we analyzed the age distribution of the patients with DNI identified in the T1DM and non-DM cohorts.

### *2.9. Statistical Analysis*

The demographic characteristic and comorbidities of the T1DM and non-DM cohorts were compared using the Pearson's chi-square test for categorical variables and the unpaired Student *t*-test for continuous variables. Control variables, such as age, sex, urbanization level, income level, and comorbidities (HTN, CVA, CAD, CKD, SADs, and LC) were included as covariates in the univariate model. Variables in the univariate analysis that showed *p* < 0.1 were included in the multivariate analysis. Kaplan–Meier analysis was used to estimate the cumulative incidence in the two cohorts, and the differences were determined using a two-tailed log-rank test. Multivariable Cox proportional hazard regression models were used to measure the hazard ratio (HR) and 95% confidence interval (CI) of DNI incidence between the T1DM and non-DM cohorts. In addition, the stability of HR was examined using sensitivity testing and subgroup analysis if the interaction effects between the comorbidities and T1DM on DNI were significant. All analyses were performed using SAS software, version 9.4 (SAS Institute, Cary, NC, USA), and the level of statistical significance was set at *p* < 0.05.

### **3. Results**

Table 1 illustrates the distribution of sociodemographic characteristics, DNIs, and comorbidities identified in the T1DM and non-DM cohorts. The T1DM cohort exhibited a significantly higher prevalence of DNI, HTN, CVA, CAD, CKD, SADs, and LC. Among the 5741 patients with T1DM, 42 (0.7%) patients with DNI were identified, and the incidence rate was 92.4 per 100,000 person-years in a mean follow-up period of 7.91 ± 2.41 years. By contrast, among the 22,964 controls, 16 (0.1%) patients with DNI were identified in a mean observation period of 8.08 ± 2.29 years, and the incidence rate was 8.6 per 100,000 person-years. The incidence rate ratio was 10.73 with a 95% CI of 6.03–19.08. The incidence of DNI was significantly higher in the T1DM cohort than in the non-DM cohort (*p* < 0.001).


**Table 1.** Demographic characteristics of the T1DM and non-DM cohorts.

Abbreviations: T1DM, type 1 diabetes mellitus; NTD, New Taiwan dollar; HTN, hypertension; CVA, cerebrovascular accident; CAD, coronary artery disease; CKD, chronic kidney disease; SADs, systemic autoimmune diseases; LC, liver cirrhosis; DNI, deep neck infection. \* Pearson's chi-square test.

Results of the Kaplan–Meier analysis revealed the cumulative incidence of DNI in both the cohorts over a 10-year observation period. The T1DM cohort exhibited a significantly higher incidence of DNI than the non-DM cohort did (log-rank test *p* < 0.001, Figure 2). The Cox proportional hazards model revealed that T1DM was associated with a 10-fold higher risk of DNI (adjusted HR: 10.71, 95% CI: 6.02–19.05, *p* < 0.001, Table 2). In addition, the sensitivity test showed a stable effect of T1DM on DNI risk in the study cohort in the main model with each additional covariate. The results of subgroup analysis showed that T1DM is a risk factor for DNI in all the subgroups.

**Figure 2.** Cumulative incidence of DNI in the T1DM versus non-DM cohorts. Kaplan–Meier analysis demonstrated the cumulative DNI identified in the T1DM and non-DM cohorts during the 10-year follow-up period. The log-rank test revealed a significantly higher cumulative incidence in the T1DM cohort than in the non-DM cohort (*p* < 0.001).

**Table 2.** Multivariable Cox proportional hazards model for associations between DNI and T1DM.


\* Main model was adjusted for sex, age, urbanized level, and income. † The model was adjusted for sex, age, urbanized level, income, and each additional comorbidity. Abbreviations: DNI, deep neck infection; T1DM, type 1 diabetes mellitus; NTD, New Taiwan dollar; HTN, hypertension; CVA, cerebrovascular accident; CAD, coronary artery disease; CKD, chronic kidney disease; SADs, systemic autoimmune diseases; LC, liver cirrhosis.

Table 3 presents the treatment modalities and prognosis of DNI in the patients in both cohorts (Table 3). Although the percentage of patients requiring surgical treatment for DNI was higher in the T1DM cohort than in the non-DM cohort, the difference in the percentages was not significant (T1DM vs. non-DM cohorts = 33.3% vs. 18.8%, *p* = 0.276). DNI in the patients in the T1DM cohort required longer hospitalization durations than did those in the non-DM cohort (T1DM vs. non-DM cohorts: 9.0 ± 6.2 vs. 4.1 ± 2.0 days, *p* < 0.001). Furthermore, care in ICU and mediastinal complications were only identified in patients with T1DM and DNI (ICU: 6/42, 14.3%; mediastinitis: 1/42, 2.4%). DNI-related mortality was observed in the T1DM cohort (mortality: 2/42, 4.8%) but not in the non-DM cohort.


**Table 3.** Treatment modalities, complications, and prognostic outcomes in patients with DNI.

\* Mortality occurrence after DNI. <sup>a</sup> Pearson's chi-square tests. <sup>b</sup> Student *t*-tests. Abbreviations: DNI, deep neck infection; SD, standard deviation; ICU, intensive care unit.

Figure 3 presents the age distribution of DNI identified in the T1DM and non-DM cohorts. We divided the age into the following four groups: <10, 10–20, 21–40, and >40 years. Accordingly, the proportions of DNI in the two cohorts (T1DM vs. non-DM) were 2.38% vs. 18.75% (<10 years), 40.47% vs. 18.75% (10–20 years), 42.86% vs. 50% (21–40 years), and 14.28% vs. 12.5% (>40 years). In this study, the peak age of DNI occurrence in the non-DM cohort was 21–40 years, while the T1DM cohort exhibited two peak ages, namely 10–20 and 21–40 years.

**Figure 3.** Age distribution of DNI in the T1DM and non-DM cohorts. The peak age of DNI occurrence in the non-DM cohort was 21–40 years, while those in the T1DM cohort were 10–20 years and 21–40 years.

### **4. Discussion**

Our nationwide study is the first to examine the influence of T1DM on DNI. Our study demonstrated that T1DM is a definite risk factor for DNI. Our results revealed that patients with T1DM are at a 10-fold higher risk of DNI than were those without DM. The higher frequency of infections in patients with T1DM is attributable to hyperglycemia, which results in immune dysfunction, including disrupted neutrophil function, depression of the antioxidant system and humoral immunity, microand macroangiopathies, neuropathy, decrease in the antibacterial activity of urine, gastrointestinal and urinary dysmotility, and the need for medical intervention in these patients [20].

Patients with T1DM are more likely to have complicated infections, such as pneumonia, septicemia, and osteomyelitis, than are those without DM [21]. Simonsen et al. reported that the incidence of bacterial infections was significantly higher in patients with T1DM than in those without DM [22]. Muller et al. reported that patients with T1DM and T2DM have an increased risk of infections of the lower respiratory tract, urinary tract, and skin and mucous membranes [23]. In addition, an Australian diabetes register-based study revealed that patients with T1DM exhibited significantly higher infection-related mortality (pneumonia, septicemia, and osteomyelitis) than did those with T2DM [11]. Therefore, T1DM is a risk factor for complicated infections, and it might be associated with higher incidence and severity of infection than T2DM.

Previous studies have reported that surgical treatment was used in 55–80% of patients with DNI [4,9,24–28]. In our study, few patients received surgical treatment in both the cohorts (T1DM: 33.3% and non-DM: 18.7%). This difference in percentage may result from previous studies being conducted in medical centers or tertiary hospitals, which receive and treat patients with severe DNI [4,8,9,24–27]. Hence, patients with severe DNI were more likely accept surgical interventions. However, we enrolled patients from all hospitals in our nationwide study. The distribution of patients with DNI was from primary to tertiary hospitals, and patients with low DNI severity were also included; thus, our study provided a complete spectrum of DNI treatment and prognosis [1]. In general, the use of surgical interventions to treat a DNI indicates that the infection is more severe and life-threatening. In our study, the percentage of surgical treatment for DNI was higher in the T1DM cohort than in the non-DM cohort; however, the difference was not statistically significant.

DNIs in patients with DM have been reported to be associated with long hospitalization durations and numerous complications [4,28–30]. In our study, the duration of hospitalization for DNI was significantly higher in the T1DM than in the non-DM cohort, and this result was consistent with previously reported findings. Patients with T1DM and DNI were reported to exhibit a higher frequency of lethal complications, such as mediastinitis (2.7–10.0%), and higher mortality (1.6–7.5%) than those without DM [4,28,29]. A higher rate of ICU care for DNI was noted in patients with T1DM than with those without DM [1]. In our study, the occurrence of ICU care for DNI, mediastinitis, and DNI-related mortality was higher in the T1DM cohort than in the non-DM cohort, and these results were consistent with those of previous studies.

We analyzed the age distribution of DNI in the T1DM and non-DM cohorts in our study. In the non-DM cohort, the peak age of DNI occurrence was 21–40 years, while in the T1DM cohort, the two peak ages of DNI occurrence were 10–20 years and 21–40 years. In addition, in the T1DM cohort, DNI developed at age 10–20 years. In general, the incidence of DNI was higher at age 20–40 years. Patients with diabetes have been reported to have a late onset of DNI [24,25,28,31]. However, T1DM was characterized by diagnosis at a young age; according to Magliano's report, infection (pneumonia, septicemia, and osteomyelitis) at a young age was more likely to occur in patients with T1DM than in patients with T2DM (T1DM vs. T2DM = 26.9 vs. 60.4 years) [11]. In summary, we believe that patients with T1DM tend to develop DNI at younger age (10–40 years) than do patients without DM (21–40 years) and those with T2DM (>40 years).

Our study has several strengths, including a large number of patients with T1DM representing a nationwide population and a 10-year observation period. In addition, the diagnosis of T1DM was based on data from the RFCIP, a highly accurate and reliable database affiliated to the NHIRD. Nevertheless, the study has some limitations. The diagnoses were based on ICD-9-CM codes and not on original medical records; therefore, it lacked blood sugar level, laboratory data, data from imaging studies, surgical records, and pathologic reports, which are necessary for evaluating disease severity. The bacterial spectrum and drug sensitivity of T1DM-DNI and the difference from non-DM-DNI are important information for clinical management and prescription of antibiotics. However, our database did not contain that information. The effects of the factors omitted in this study on T1DM and DNI should be investigated in future studies. In addition, ICU care, mediastinitis, and mortality were observed only in patients with T1DM; however, the number of patients was insufficient to develop a statistical conclusion. Additional studies including detailed medical records and a large sample size of patients with DNI are needed.

### **5. Conclusions**

This nationwide population-based study was the first to investigate the epidemiological data of DNI development and prognosis in patients with T1DM. We concluded that T1DM is a predisposing factor for DNI. The duration of hospitalization for DNI is longer in patients with T1DM than in those without DM. In addition, patients with T1DM are predisposed to developing DNI at a younger age than are those without DM.

**Author Contributions:** Conceptualization, G.-H.C. and M.-S.T.; Methodology, M.-C.D. and Y.-H.Y.; Software, C.-Y.L. and M.-H.L.; Validation, Y.-H.L., C.-Y.W. and C.-M.H.; Formal Analysis, C.-Y.L.; Investigation, G.-H.C. and M.-S.T.; Resources, C.-Y.W. and C.-M.H.; Writing-Original Draft Preparation, G.-H.C. and M.-C.D.; Writing-Review & Editing, M.-S.T.; Supervision, Y.-H.Y. and M.-S.T.

**Funding:** The article did not have any financial support provided by companies toward the completion of the work.

**Acknowledgments:** The authors thank the Health Information and Epidemiology Laboratory (CLRPG6G0041) at the Chiayi Chang Gung Memorial Hospital for the comments and assistance in data analysis. This study was supported by a grant from Chiayi Chang Gung Memorial Hospital (CGRPG6G0011), based on the National Health Insurance Research Database provided by the Central Bureau of National Health Insurance, Department of Health, and managed by the National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health or National Health Research Institutes. This manuscript was edited by Wallace Academic Editing.

**Data and Material Availability:** The datasets generated and/or analyzed during the current study are available in the Taiwan National Health Insurance Research Database repository [32].

**Conflicts of Interest:** The authors declare that they have no competing interests.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **PG-Priming Enhances Doxorubicin Influx to Trigger Necrotic and Autophagic Cell Death in Oral Squamous Cell Carcinoma**

### **Shian-Ren Lin and Ching-Feng Weng \***

Department of Life Science and Institute of Biotechnology, National Dong Hwa University,

Hualien 97401, Taiwan; d9813003@gms.ndhu.edu.tw

**\*** Correspondence: cfweng@gms.ndhu.edu.tw; Tel.: +886-3-8903637

Received: 17 September 2018; Accepted: 18 October 2018; Published: 21 October 2018

**Abstract:** Synergistic effects between natural compounds and chemotherapy drugs are believed to have fewer side effects with equivalent efficacy. However, the synergistic potential of prodigiosin (PG) with doxorubicin (Dox) chemotherapy is still unknown. This study explores the synergistic mechanism of PG and Dox against oral squamous cell carcinoma (OSCC) cells. Three OSCC cell lines were treated with different PG/Dox combinatory schemes for cytotoxicity tests and were further investigated for cell death characteristics by cell cycle flow cytometry and autophagy/apoptosis marker labelling. When OSCC cells were pretreated with PG, the cytotoxicity of the subsequent Dox-treatment was 30% higher than Dox alone. The cytotoxic efficacy of PG-pretreated was found better than those of PG plus Dox co-treatment and Dox-pretreatment. Increase of Sub-G1 phase and caspase-3/LC-3 levels without poly (ADP-ribose) polymeras (PARP) elevation indicated both autophagy and necrosis occurred in OSCC cells. Dox flux after PG-priming was further evaluated by rhodamine-123 accumulation and Dox transporters analysis to elucidate the PG-priming effect. PG-priming autophagy enhanced Dox accumulation according to the increase of rhodamine-123 accumulation without the alterations of Dox transporters. Additionally, the cause of PG-triggered autophagy was determined by co-treatment with endoplasmic reticulum (ER) stress or AMP-activated protein kinase (AMPK) inhibitor. PG-induced autophagy was not related to nutrient deprivation and ER stress was proved by co-treatment with specific inhibitor. Taken together, PG-priming autophagy could sensitize OSCC cells by promoting Dox influx without regulation of Dox transporter. The PG-priming might be a promising adjuvant approach for the chemotherapy of OSCC.

**Keywords:** prodigiosin; doxorubicin; priming; influx; autophagy

### **1. Introduction**

Doxorubicin (Adriamycin, Dox), isolated from soil bacteria *Streptomyces peucetius*, is the first member of anthracyclines (including daunorubicin, epirubicin, and idarubicin) [1,2]. The main action of Dox is intercalating within DNA pairs, which leads to the inhibition of topoisomerase IIβ and results in cell cycle blockage [3]. Rendering non-tissue specific characteristics, Dox gets wide indications including leukemia, neuroblastoma, breast carcinoma, ovarian carcinoma, and most recurrent or metastatic cancer [4]. Apart from these, the non-specific characteristics of Dox cause some severe adverse effects in cancer patients including immunosuppression, bone marrow suppression, hepatotoxicity, cardiotoxicity, and mucositis [5,6]. The cause of side effects is from the inhibition of cell division as well as reactive oxygen species (ROS) by-products (doxorubicin-semiquinone, doxorubicinol, dexrazoxane, and 7-deoxy-doxorubicinone) during metabolism of doxorubicin in mitochondria [2,7–9]. As a result, these adverse effects might limit the applicable dosage and cancer treating efficacy of Dox. Accordingly, the alternative approach or new formulation to attenuate Dox

side effects and enhance Dox efficacy turns out to be a crucial issue for Dox use in the regimen of chemotherapy. Recently, nanocarriers have exerted a favorable theme and some research has focused on this topic to dissolve these obstacles, such as Dox encapsulated in pH-sensitive, ultrasonic-responsive, or co-capsulated with MDR-1 inhibitor in PEGylated, liposome, or PLGA nano-carrier, which also promote Dox uptake [10–14]. However, cytotoxicity of nanoparticle conjugated Dox was 10 times lower than free-form Dox, which also restricted the use in cancer treatment [15].

A long treatment period with a low dose chemotherapeutic drug might induce chemoresistance within cancer cells and subsequently toxicity could affect its use [16]. Notably, prevailing mechanisms of chemoresistance could be classified into the following seven phases: drug flux, DNA damage repair, cell death inhibition, epithelial-mesenchymal transition (EMT), drug target alteration, drug inactivation, and epigenetics [17]. In Dox resistance, dug efflux would be the most concerning phase [16]. Dox can import into cells through solute carrier family 22 member 16 (SLC22A16, also known organic cation transporter 6, oct-6) and export by ATP-binding cassette transporter family members, in which multidrug-related protein 1 (MDR-1 or p-glycoprotein) and breast cancer resistance protein (BCRP or ABCG2) are involved [3]. These proteins will be regulated (upregulation of exporter and downregulation of importer) during long-term exposure to a non-toxic dose of Dox [18]. Thereby, numerous studies have put more attention to reducing MDR over-expression to reverse multidrug resistance. CRISPER/Cas-9 gene editing, ursodeoxycholic acid, or *Zingiber officinale* Roscoe, have been reported to down-regulate *ABCB1* gene expressions in chemo-resistant cancer cells [19–21].

Prodigiosin (PG, PubChem CID: 5351169) is a red prodiginine pigment isolated from various bacteria including *Serratia marcescens, Pseudoalteromonas rubra, Hahella chejuensis,* and actinomycete bacteria [22–25]. Even though the original biological function in producer bacteria remains unclear, PG has been identified with numerous biological activities including antimicrobial [26–29], antimalarial [26,27,30], and antitumor [26,27,31–34] activities. Moreover, PG showed apoptotic inducing property in many cancer types such as lung cancer [35–37], breast cancer [38,39], colorectal cancer [40–42], leukemia [43,44], and hepatocellular carcinoma [45] without normal cell cytotoxicity [41,46]. Recently, PG has also been identified as an autophagy inducer in OSCC cells [47,48]. However, the application of PG as an adjuvant in chemotherapy is still unknown.

### **2. Experimental Section**

### *2.1. Research Aims*

This study was conducted to explore the potential of PG combined with doxorubicin in anti-cancer activity by using oral squamous cell carcinoma (OSCC) cells as a test platform. Next, experiments tested the synergistic effects of PG and Dox against OSCC cells to evaluate the adjuvant potential of PG for cancer therapy. Furthermore, the underlying molecular mechanisms of enhanced doxorubicin cytotoxicity under PG-priming were also investigated.

### *2.2. Reagents*

Cell-cultured medium and reagents were purchased from Thermo-Fisher (Waltham, MA, USA). Prodigiosin was purified by Dr. Yu-Hsin Chen (Department of Life Science, National Dong-Hwa University, Hualien, Taiwan). Liposome-coated doxorubicin (abbreviated as Dox) was obtained from Dr. Ming-Fang Cheng (Division of Histology and Clinical Pathology, Hualian Army Forces General Hospital, Hualien, Taiwan). Inhibitors used in this study were purchased from Santa Cruz Biotechnology (Dallas, TX, USA). General chemicals were purchased from Sigma Aldrich (Merck KGaA, Darmstadt, Germany). Polyvinylidene difluoride (PVDF) membrane used in Western blotting was obtained from GE Healthcare (Chicago, IL, USA). The antibodies used in this study were obtained from Santa Cruz Biotechnology, as shown in Table 1.


**Table 1.** Antibodies used in this study.

MW, molecular weight; RRID, Research Resource Identifiers.

### *2.3. Cell Culture*

Cell lines used in this study were obtained from different sources: Human pharynx squamous carcinoma FaDu from Dr. Chun-Shu Lin (Radiation Oncology Department, Tri-Service General Hospital, Taipei, Taiwan), human oral squamous cell carcinoma cell line OECM1 and tongue carcinoma cell line SAS from Professor Ta-Chun Yuan (Department of Life Science, National Dong Hwa University, Hualien, Taiwan), and human bronchus epithelial cell BEAS-2b from American Type Culture Collection (ATCC). OECM1 and SAS were cultured in Roswell Park Memorial Institute medium 1640 (RPMI 1640), FaDu in minimum essential medium (MEM), and BEAS-2b in Dulbecco's Modified Eagle Medium (DMEM) and medium was changed every 2 days. All culture media were mixed with 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic and cultured within 37 ◦C, 5% CO2 incubator (Thermo-Fisher). Cells were detached by 0.25% trypsin/ethylenediaminetetraacetic acid (EDTA) for further experiments. All experiments were obtained within 20 passages concerning uniformity and reproducibility.

### *2.4. Cytotoxicity Assay*

Cytotoxicity was determined using a colorimetric assay by MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) previously described in the literature [48]. The optical density (OD) alteration of mitochondrial enzymatic activity was converted into the cell numbers according to the cell viability or cytotoxicity. Briefly, 1 × 104 cells per wells were seeded in 96-well plate and incubated in culture conditions overnight. Then, cells were divided into the six following treatment groups: (1) PG-Dox group: treated with PG followed by Dox, (2) Dox-PG group treated with Dox and then PG, and (3) PG + Dox group: treated PG and Dox at the same time, respectively. An additional three groups performed the same treatments with the above-described and replaced Dox with cisplatin. All treatments were carried out in 12 h and 1 mg/mL of MTT solution was added and further incubated for 4 h at 37 ◦C as treatment finished. Finally, liquid in wells was replaced by dimethyl sulfoxide (DMSO), and the absorbance at 570 nm was measured by Multiskan™ FC microplate photometer (Thermo-Fisher). Cytotoxicity of each treatment was represented by cell viability which calculated from the absorbance ratio at 570 nm between treated and untreated groups.

To understand the cause of PG- and Dox-induced cell death, inhibitor recovering assays were also performed following the above protocol. Autophagy inhibitors (bafilomycin A1 and 3-methyladenine), endoplasmic reticulum (ER) stress inhibitors (tauroursodeoxycholic acid, TUDC), and AMPK inhibitors (dorsomorphin, CC) were cotreated with various concentrations of PG or Dox, respectively.

### *2.5. Cell Cycle Analysis*

Cell cycle analysis was carried out by flow cytometry. Firstly, 1 × <sup>10</sup><sup>6</sup> cells/well of OSCC were seeded into 6-well plates and incubated in culture condition overnight. To understand drug-pretreated effect, cells were treated with PG for 12 h, Dox for 12 h, or PG for 12 h followed by Dox for additional 12 h, respectively. After treatment, cells including culture medium were collected using trypsin/EDTA and washed by phosphate buffer saline (PBS) twice before being fixed with pre-cooled 70% ethanol/PBS overnight. After fixation, cells were washed twice by PBS and stained with staining buffer (20 μg/mL of propidium iodide, 0.1% Triton X-100, 0.2 mg/mL RNase A) at 37 ◦C for 1 h. The fluorescent intensity in the cells was measured by a flow cytometer (CytomicsTM FC500, Beckman, Fullerton, CA, USA). Data from 10<sup>4</sup> cells were collected for each data file. Fluorescent intensities for each cell line were acquiesced and plotted by flow cytometer software. The gating of each phase was based on the acquisition histogram of untreated controls. Phases of each group were collected and the average of each phase was calculated within the groups.

### *2.6. Doxorubicin Flux Analysis*

Efflux and influx of Dox was determined by indirect method which used rhodamine 123, a fluorescent Dox transporter substrate, detected as an indicator described previously [49,50]. Briefly, <sup>1</sup> × 104 cells/well of OSCC were seeded in 96-well plate and incubated in culture condition overnight. After cell confluence, cells were divided into four groups and then treated with various regimens, respectively. For Rhodamine-influx: short-term influx assay: 0.5 μM of PG in full-culture medium was added and incubated for 1 h and replaced culture medium with 2 μM of rhodamine 123 in PBS for additional 1 h; long-term influx assay: the same treatment as short-term influx assay instead the incubation duration of PG from 1 h to 12 h. For Rhodamine-efflux: short-term efflux assay: 2 μM of rhodamine 123 in PBS was firstly incubated for 1 h and followed by 0.5 μM of PG in full-culture medium for additional 1 h; long-term efflux assay: the same as treatment as short-term influx assay instead the incubation duration of PG from 1 h to 12 h. After incubation, cells were trebly washed by PBS and lysed with 0.1% triton X-100 and the fluorescent intensity was determined at 485/538 nm by α-screen multi-plate reader (Perkin Elmer, Waltham, MA, USA). These rhodamine flux studies were used to estimate DOX flux.

### *2.7. Western Blotting*

The detail protocol of Western blotting was described in our previous study [51]. In brief, 1 × 106 cells/well of OSCC cells were seeded into a 6-well plate and treated the same as with cell cycle analysis. Cells were washed by PBS and lysed by radioimmunoprecipitation assay buffer (RIPA). Then, 30 μg of total proteins from cell lysates were electrophoretically separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred into polyvinylidene difluoride (PVDF) membrane. Proteins of interest were identified via incubation with appropriate primary followed by horseradish peroxidase (HRP)-conjugated secondary antibodies and exposed to the iBright imaging system (Thermo-Fisher) for monitoring intensity of signals after soaking in enhance chemiluminescent (ECL) reagents. Data acquisition was also performed by the iBright imaging system, and signal intensity was normalized with GAPDH as an internal control.

### *2.8. Statistical Analysis*

All quantified results were shown as mean ± standard deviation (SD) of three independent experiments. Significant analysis used a one-way ANOVA, followed by Dunnett's test. A data histogram was built by GraphPad Prism 7.04 (La Jolla, CA, USA).

### **3. Results**

### *3.1. Cytotoxicity Change of PG/Dox Treated Strategies*

In this study, three combined manners (pretreatment, cotreatment, and posttreatment) of PG/Dox were tested in OSCC. In pretreatment and post-treatment approaches, chemicals were previously treated for 12 h and subsequently washed out, followed by new chemical treatment for additional 12 h. Therefore, the term "PG-pretreatment" would be defined as a "PG-priming" procedure in the subsequent section. The cell viability of all tested OSCC cells were declined in all combined strategies

except cotreatment in SAS (*p* < 0.05). In three combined strategies, PG-pretreatment got the highest reducing levels (as compared with Dox alone) than those of the other two strategies, as shown in Figure 1A. This result posed the potential of PG-pretreatment as PG-priming in OSCC. When doubling the concentrations of PG, cell viability was the same as that of single concentration, revealed 0.5 μM of PG, which was the maximum concentration for PG-priming. Also, extending the PG-priming period up to 24 h, the cytotoxicity of Dox failed to exhibit an additive potentiation. These results indicated that 12 h of PG-priming might reach the maximum effect (data not shown). Moreover, with PG-priming in normal cell lines BEAS-2b, the cell viability of Dox treatment did not show the decrease as much as OSCC, even though the concentrations of PG and Dox were twice higher than that of OSCC, as shown in Figure 1B. This result indicated that PG-priming was more effective and less toxic than that of Dox alone. An additional experiment was to investigate whether the PG-priming effect could also be observed in golden chemotherapy drug cisplatin, however, the cytotoxic enhancement in PG/Dox combination could not be found in PG/cisplatin combination, as shown in Figure 1C. Taking all results together, PG-priming could enhance Dox cytotoxicity in OSCC cells through a Dox-related mechanism. In subsequent experiments, the type of cell death triggered by PG-priming and the underlying mechanism were further investigated.

**Figure 1.** Alteration of cytotoxicity in sequential PG (prodigiosin)/Dox (doxorubicin) and PG/cisplatin combination in oral squamous cell carcinoma (OSCC) and BEAS-2b cells. (**A**) OSCC and (**B**) normal bronchus cells-BEAS-2b were treated with various schemes of PG and Dox, and (**C**) Cisplatin substituted Dox for 12 h and analyzed cell viability by 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) assay. The results were represented as mean ± SD from three individual experiments. \* *p* < 0.05 as compared with Dox alone.

### *3.2. Identification of Cell Death Characteristics*

Numerous types of cell death were found in cells, but the most common types were apoptosis and autophagy, respectively. These two cell-death types could be distinguished by analyzing apoptosis and autophagy-related protein and cell markers, and the most obvious marker would be cell cycle analysis. In cell cycle analysis of SAS, Sub-G1 was significantly increased, but non-cleaved PARP1 and caspase-3 protein levels were not decreased after PG-priming, as shown in Figures 2B and 3B,C. These results revealed that PG-priming prior to Dox treatment would lead to SAS undergoing necrosis. Moreover, while PG combined with autophagic inhibitor (bafilomycin A1 (BA1) and 3-methyladenine (3MA), cell viability of PG-priming could be recovered, as shown in Figure 4. This phenomenon might indicate that both necrosis and autophagy were activated after PG-priming and the autophagy would be a major clue, which was further confirmed by increases in LC3 protein levels, as shown in Figure 3A. In FaDu cells, Sub-G1 phase was not significantly increased after PG/Dox treatment, as shown in Figure 2C. Also, non-cleaved PARP1 and caspase-3 protein levels while PG-priming followed by Dox treatment were decreased when compared with Dox alone, as shown in Figure 3B,C. The results also showed the necrosis activation within FaDu cells similar to SAS cells. Unlike SAS cells, this cell viability could not be recovered by autophagy inhibitor, as shown in Figure 4. In FaDu cells, LC3 protein levels were also significantly increased in the PG/Dox group, as shown in Figure 3A. These results revealed that both necrosis and autophagy were activated in FaDu cells and necrosis would be the main cause of cytotoxicity, whereas OECM1 cells showed different patterns from the above two cells lines. The sub-G1 phase did not significantly increase either in the non-cleaved PARP1 or caspase-3 protein levels, as shown in Figures 2A and 3B,C. Taken together, PG-priming cell death of OECM1 was not related to apoptosis or necrosis. On the other hand, the cell viability of OECM1 after PG-priming could be recovered by autophagy inhibitors, as shown in Figure 4. Also, LC3 protein levels were increased 10-fold, as shown in Figure 3A. These two results gave a clear clue for autophagy in OECM1 by PG-priming. Considering all above results, induced cell death characteristics in three different cell lines by PG-priming illustrated in different configurations. OECM1 showed autophagy, and SAS and FaDu posed both cell death and autophagy. In the subsequent experiments, the potential pathways of PG-enhanced Dox cytotoxicity were under investigation.

**Figure 2.** Alteration of cell cycle in (**A**) OECM1, (**B**) SAS, and (**C**) FaDu cells. OSCC cells were treated with PG/Dox for 12/12 h prior to staining with propidium iodide (PI) and fluorescent intensity was analyzed by flow cytometry. The results were represented as mean ± SD from three individual experiments. \* *p* < 0.05 as compared with DOX alone.

**Figure 3.** Expression of (**A**) LC3, (**B**) PARP1, and (**C**) Caspase3 in OSCCs after PG-priming. OSCC cells were treated with PG/Dox for 12/12 h and then desired protein levels were analyzed by Western blotting. The results were normalized with GAPDH and represented as mean ± SD from three individual experiments. Molecular weight: PARP1, 116 KDa; GAPDH, 37 KDa; Caspase3, 34 KDa; LC3, 18 KDa. # *p* < 0.05 compared with untreated control; \* *p* < 0.05 as compared with Dox alone.

**Figure 4. Alteration of cell viability combined with autophagy inhibitor.** OSCC cells were treated with PG + inhibitor/DOX or PG/Dox + inhibitor and cell viability was analyzed. Table under figure was the scheme of treatment. "X" meant incubated with complete medium without PG or Dox. The results were represented as mean ± SD from three individual experiments. \* *p* < 0.05 as compared with PG/Dox.

### *3.3. Doxorubicin Flux Affected by PG-Induced Autophagy*

To measure the possible action of PG-induced autophagy, we examined the Dox flux in PG-priming OSCC cells. This Dox flux was determined by rhodamine-123 (R123) accumulation. R123 is a green fluorescent dye which acts as a Dox-transporter substrate over decades [52]. Accordingly, PG and Dox are all red fluorescence and PG is stronger fluorescence than that of Dox. After PG/Dox combination treatment, PG will interfere with the measurement of Dox fluorescent-intensity. Also, the less cytotoxic nature of R123 could eliminate the interference of cell death caused by Dox. Therefore, R123 was employed as an indicator to indirectly determine the Dox-flux in this study. In short-term priming (1 h), PG-priming did not enhance R123 accumulation, which revealed PG did not allosterically regulate Dox transporter, as shown in Figure 5A. Subsequently, PG-priming showed additional 50–70% of R123 accumulation for long-term priming (12 h). Moreover, the enhancing R123 could be attenuated by autophagy inhibitor, as shown in Figure 5B. This result indicated that PG-priming either enhanced Dox importer expressions or reduced exporter expressions. When checking Dox transporter levels, however, the importer (Oct-6) was not significantly decreased, and exporters (MDR-1 and ABCG2) slightly increased in OECM1 and decreased in SAS and FaDu, as shown in Figure 6. This result was not associated with previous results. It might be postulated as an indication of an unknown but important mechanism of Dox transport.

**Figure 5. Rhodamine 123 accumulation after PG pretreatment.** OSCC cells were treated with PG/R123 for (**A**) short term (1/1 h) and (**B**) long term (12/1 h) followed by analyzed fluorescent intensity within cells. The results were represented as mean ± SD from three individual experiments. # *p* < 0.05 as compared with R123 alone; \* *p* < 0.05 compared with PG/R123 combination.

**Figure 6. Protein levels of Doxorubicin-related importer and exporter after PG-priming.** OSCC cells were treated with PG/Dox for 12/12 h and then were analyzed for Importer OCT-6 (**A**) and exporter MDR-1 (**B**) and ABCG2 (**C**) protein levels by Western blotting. The results were normalized with GAPDH and represented as mean ± SD from three individual experiments. Molecular weight: MDR-1, 170 KDa; ABCG2, 72 KDa; OCT-6, 58 KDa; GAPDH, 37 KDa. \* *p* < 0.05 compared with Dox alone.

#### *3.4. ER Stress and Energy Deprivation Analysis in PG-Priming OSCC Cells*

PG could activate autophagy of OSCC cells as proven by the previous section and literature [48]. In this final part, the aim was to find the trigger of PG-induced autophagy in OSCC cells. As we noted, the two known triggers of autophagy are ER stress (unfolded protein response) and energy deprivation, which may be involved in the PG-priming reaction. ER stress was determined by adding ER stress inhibitor TUDC while energy deprivation was blocked by addition of AMPK specific inhibitor CC. When we combined TUDC or CC with PG and Dox treatment, cell viability of three OSCC cells lines were not significantly changed, as shown in Figure 7, (data of FaDu not shown). This result further postulated that PG-induced autophagy and Dox flux increase were not caused by ER stress and energy deprivation.

**Figure 7. Cell viability change when PG/Dox combined with ER stress and energy deprivation inhibitors.** OSCC cells were treated with PG + inhibitor/Dox or PG/Dox + inhibitor and were analyzed for cell viability. The inhibitors contained tauroursodeoxycholic acid (TUDC) and dorsomorphin (compound C, CC). The results were represented as mean ± SD from three individual experiments. \* *p* < 0.05 compared with PG/Dox.

### **4. Discussion**

The present study is the first demonstration of PG-enhanced Dox influx by activating autophagy in OSCC cells. Based on the Dox flux experiments, the enhancing mechanism of Dox influx was neither related to known importers nor exporters. PG could be a potential adjuvant for Dox treatment. Also, this study excluded the characteristics of known autophagy-triggers in PG-induced autophagy, which posed a new site for autophagy triggering mechanisms. In this work, we first report about the autophagy-activating property of Dox, as this activity was known to be inhibited by the intrinsic activation of autophagy.

Due to the non-tissue specific nature and high-cardiotoxicity, several studies have tried to discover natural compounds that could synergistically potentiate the efficacy of Dox without elevating normal cell toxicity. In a previous report, gambogic acid, a xanthonoid from *Garcinia hanburyi*, sensitized ovarian cancer cells toward Dox through accumulation of ROS [53]. Nitidine chloride, an alkaloid, synergized Dox cytotoxicity in breast cancer cell through PI3K/Akt signaling pathway [54]. Gingerol synergized Dox against liver cancer cells, leading to G2/M arrest [55]. Not only pure compounds; phenolic extract of flaxseed oil also promoted Dox efficacy against breast cancer cells [56]. Evodiamine, a major element of *Evodiae fructus*, reversed chemoresistance in multi-drug resistant breast cancer cells through the Ras/MEK/ERK signaling pathway [57]. Additionally, neferine could combat Dox resistance through ROS accumulation and Fas signaling pathway in lung cancer [58]. In gastric cancer, curcumin and formononetin posed different mechanisms to enhance Dox cytotoxicity [59,60]. These natural compounds exhibit potentiation for main applications in cancer treatment, and also play a supporting role in Dox regimen to overcome the limitation of Dox usage [61–66].

The first aim of this study was to explore the synergistic effect within PG/Dox regimen. PG combined with current chemotherapy agents was studied in breast cancer and found that PG could facilitate paclitaxel sensitivity in triple-negative human breast carcinoma cells via down-regulating survivin expression, an anti-apoptotic protein that acts as a caspase inhibitor [67,68]. Our results demonstrated new evidence that PG acts as an adjuvant with conventional chemotherapeutic drugs, such as paclitaxel and Dox as well.

The synergism of natural compounds with Dox could be found in priming fashion, nevertheless the co-treatment is addressed in main efforts [53–60]. While priming with CDK inhibitor in triple-negative breast cancer cell MDA-MB-231, Dox-induced DNA double-strand break would be activated and resulted in cytotoxic enhancement [69]. Cyclophosphamide, a conventional chemotherapeutic drug that acts as an intercalator of DNA, could increase HER2-targeted liposomal Dox accumulation in breast cancer cells [70]. An in vivo study focused on Dox efficacy after mitomycin C and carboplatin (two conventional chemotherapeutic drugs) pretreatment in human metastatic breast cancer-bearing mice. The results showed inhibition growth of xenografted tumors and reducing expressions of p-glycoprotein [71]. In our study, we showed that PG potentiated Dox cytotoxicity only in a pretreatment fashion (as a PG-priming effect), which was the first report of natural compound that primed cancer cells to be sensitized with Dox, and posed the potential of PG as an adjuvant using Dox as a chemotherapeutic agent. The clinical application of PG-priming might provide a great clue for reducing the dosage of Dox and dampening the side effects of Dox.

Due to red fluorescent nature [72], we preliminarily examined PG influx into OSCC cells. The data showed that PG could enter OSCC cells within 1 h and reached saturation after 1.5 h exposure (data not shown). When OSCC cells were primed with PG for 1 h, R123 fluorescent intensity did not significantly accumulate. This gave clear insight into PG action that did not allosterically modulate Dox importer or exporter activity. A previous study has also indicated that PG was not the substrate of multidrug resistance-related protein including MDR-1, BCRP (ABCG2), and MRP2 [73]. Again, our study confirmed that Dox efflux protein was not allosterically activated by PG-priming and further exposed that PG did not allosterically mediate Oct-6 activity.

By R123 accumulation assay, PG did not allosterically control Dox transporter activity but affected transporter expression in long-term priming, as shown in Figure 5. To our best knowledge, the Dox uptake of cells via Oct-6 and excretion of Dox by MDR-1 and ABCG2 have been reported [3,74]. In our study, Dox influx significantly increased in PG-priming for 12 h, which hypothesized that Oct-6 might be up-regulated or MDR-1/ABCG2 down-regulated. However, Oct-6 was down-regulated after PG/Dox treatments in Western blotting. Furthermore, expression levels of MDR-1 and ABCG2 did not show significant reductions. These results proposed a new Dox flux mechanism that needs to be further investigated.

PG was known as apoptotic and autophagic inducer in previous studies [47,75]. According to a recent study, PG induced apoptosis via inhibiting Bcl-2, activating Bak/Bax, intercalating DNA leading to suppress the cell cycle [75]. However, the action of PG-induced autophagy has not been fully explored yet. Remarkably, autophagy was triggered by stresses, such as ER stress (unfolded protein response), nutrient deprivation, and oxidative stress [76–79]. Hence, these cellular stresses would be the trigger clue of PG-induced autophagy. However, our test exposed that using ER stress inhibitor (TUDC) and nutrient deprivation inhibitor (CC) could not restore the cell viability of PG/Dox. Also, our data showed that PG did not elevate ROS level within OSCC cells (data not shown). All-known causes of autophagy, including ER stress, nutrient deprivation, and oxidative stress, were excluded from the trigger clue of PG-induced autophagy in this study, which the new potential mechanism of autophagy activation necessitates to be further elucidated.

During the screening of an autophagic marker, we also found that Dox-induced autophagy in both OECM1 and FaDu cells. It is well known that Dox was an apoptotic inducer via inhibiting cell cycle and producing ROS [80]. The potentiated role of autophagy in Dox treatment is focused on the activation of autophagy to ameliorate cardiotoxicity, and consequently inhibiting autophagy could promote Dox sensitivity in cancer cells [81–84]. The autophagy-activating features of Dox suggested the unclear field of Dox action. In the result of autophagy inhibitor recovery assay, autophagic inhibitors could not recover Dox-induced cell death, which implies that Dox-induced autophagy might not be solely involved in Dox-induced cell death, as shown in Figure 3.

Collectively, a model for the mechanical action of PG-priming autophagy potentiated Dox influx is proposed, as shown in Figure 8. When PG entered OSCC cells, autophagy which was irrelevant to nutrient deprivation, ER stress, and ROS, was activated. Subsequently, PG-induced autophagy could up-regulate Dox importer expression and in terms translocated to cell membrane, and consequently led to the enhancement of Dox influx resulting in cell death.

**Figure 8.** Potential mechanism of PG-priming doxorubicin cytotoxicity enhancement. "X" indicates not according to this mechanism. "?" illustrates still unknown. ER, endoplasmic reticulum; ROS, reactive oxygen species.

### **5. Conclusions**

The present study firstly demonstrated the potential of PG as an adjuvant for Dox treatment in OSCC cells. PG-induced autophagy was not associated with ER stress, nutrient deprivation, and oxidative stress. Also, the enhancement of Dox influx triggered by PG-primed autophagy did not induce via Dox transporter, such as MDR-1, ABCG2, and OCT-6. The potential mechanism of PG-priming remains unclear and would be a further challenge for PG and Dox investigation.

**Author Contributions:** Conceptualization, S.-R.L. and C.-F.W.; methodology, S.-R.L.; software, C.-F.W.; validation, C.-F.W.; formal analysis, S.-R.L.; investigation, S.-R.L.; writing—original draft preparation, S.-R.L.; writing—review and editing, C.-F.W.; supervision, C.-F.W.; project administration, C.-F.W.; funding acquisition, C.-F.W.

**Funding:** This research was funded by Ministry of Science and Technology, grant number 107-2320-B-259-003 (C.F. Weng).

**Acknowledgments:** We sincerely thank Yu-Tong Chen from Kaoshiung Medical University who gave valuable help in imaging of chemiluminescent Western blotting.

**Conflicts of Interest:** The authors declare no any conflict of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Predictive Value of the Pretreatment Neutrophil-to-Lymphocyte Ratio in Head and Neck Squamous Cell Carcinoma**

### **Miao-Fen Chen 1,2,\*,†, Ming-Shao Tsai 2,3,†, Wen-Cheng Chen 1,2 and Ping-Tsung Chen 2,4**


Received: 21 August 2018; Accepted: 18 September 2018; Published: 20 September 2018

**Abstract:** This study assessed the significance of the neutrophil-to-lymphocyte ratio (NLR) in head and neck squamous cell carcinoma (HNSCC), and the relationships of the NLR with the aldehyde dehydrogenase 1 (ALDH1) level in tumors and the proportion of myeloid-derived suppressor cells (MDSCs) in the peripheral circulation. In total, 227 HNSCC patients who had received curative treatment at our hospital were enrolled into the present study. The NLR of each HNSCC patient before treatment was calculated. The associations of NLR with various clinicopathological parameters and prognoses were then examined. In addition, correlations between the proportion of MDSCs and level of ALDH1 with the NLR were assessed. Our data revealed that an elevated NLR was significantly correlated with the risk of developing locoregional recurrence and with a reduced overall survival in HNSCC patients. Multivariate analyses revealed that the NLR pretreatment and surgical resection were significantly correlated with the rate of treatment failure and the overall survival rate in HNSCC patients. Furthermore, the levels of ALDH1 in tumors and MDSCs in the peripheral circulation were significantly correlated with the prognosis of HNSCC, and the NLR was positively correlated with MDSC levels in the circulation and ALDH1 staining intensity in tumor specimens. In conclusion, the NLR has power in predicting the expression of ALDH1 in tumors, the circulating level of MDSCs, and the prognosis in HNSCC. We suggest that the NLR is an important biomarker that can assist the clinician and patient in making informed decisions regarding treatment options for HNSCC patients.

**Keywords:** head and neck squamous cell carcinoma (HNSCC); neutrophil-to-lymphocyte ratio (NLR); myeloid-derived suppressor cells (MDSC); aldehyde dehydrogenase 1 (ALDH1); prognosis

### **1. Introduction**

Head and neck squamous cell carcinoma (HNSCC) is a heterogeneous disease occurring in various sites, including the oral cavity, oropharynx, and hypopharynx [1]. Treatment failure and locoregional recurrence are common and account for the majority of deaths [2]. Identification of potential molecular markers predicting aggressive tumor growth and treatment response is important for the effective management and prognosis of HNSCC.

Abundant epidemiological data have revealed a strong correlation between inflammation and cancer incidence. Systemic inflammation is a recognized characteristic of malignancy, and numerous inflammatory markers have been investigated as prognostic indicators for cancer patients [3,4]. Host inflammatory responses were reported to play an important role in tumor development and progression [5]. The neutrophil-to-lymphocyte ratio (NLR) is an inflammatory- and immunologically-based index [6,7]. The NLR may reflect host inflammatory responses and changes in the tumor microenvironment [8]. An elevated NLR in many solid tumors, including HNSCC, has been associated with reduced survival [8–10]. However, the predictive value of the NLR in the immune and treatment responses of HNSCC is still unclear. Myeloid-derived suppressor cells (MDSCs) have been reported to attenuate immune surveillance and induce an immunosuppressive tumor microenvironment to promote cancer metastasis [11,12]. We previously reported that the recruitment of MDSCs is significantly associated with a poor prognosis in patients with HNSCC [13]. Furthermore, one of the main causes of treatment failure is the emergence of resistant cancer cells after therapy, which can be partly explained by cancer stem cells (CSCs) [14,15]. CSCs produce immunosuppressive molecules that attenuate the immune system and recruit or activate cells that suppress the immune system, such as MDSCs [16–18]. Aldehyde dehydrogenase 1 (ALDH1), a novel CSC-like cell marker, was reported to play important roles in the treatment response and tumor-promoting microenvironment in squamous cell carcinomas (SCCs) of the aerodigestive tract [19–21]. Accordingly, in the present study, we examined the predictive role of an elevated NLR in the prognosis and relationships of the NLR with the ALDH1 level in tumors, and the proportion of MDSCs in the peripheral circulation of patients with HNSCC.

### **2. Materials and Methods**

### *2.1. Study Population and Study Design*

The study protocol was approved by the institutional review board of Chang Gung Memorial hospital (No. 1035434B). Written informed consent was obtained from all patients. A total of 227 patients with a histologically confirmed diagnosis of HNSCC who received curative treatment were enrolled in our study. The planned treatments included definitive radiotherapy and chemotherapy (CCRT) or surgery +/− adjuvant treatment for patients with HNSCC, according to the guidelines proposed by the oncology team at our hospital. We included patient demographics (age and sex), diagnosis, clinical stage, and treatment characteristics (Table 1). In addition, the patients enrolled in the study had available data regarding ALDH1 levels in tumor specimens and/or the percentage of MDSCs in the peripheral circulation. The NLR was calculated by dividing the absolute neutrophil count by the absolute lymphocyte count. The mean absolute neutrophil count was 5.77 ± 3.29 × <sup>10</sup>3/μL, and the mean absolute lymphocyte count was 1.55 ± 0.85 × 103/μL. To assess the predictive value of the NLR, NLR was redefined as a binary variable by finding the value from a receiver operating characteristic (ROC) curve that maximized the percentage correctly classified for predicting tumor recurrence after treatment. The optimal cut-offs for NLR was 3. Accordingly, all HNSCC patients were divided into two groups according to the pretreatment NLR: high (NLR ≥ 3) and low (NLR < 3) groups.


**Table 1.** Characteristics of head and neck squamous cell carcinoma (HNSCC) patients with curative-intent treatment.


**Table 1.** *Cont.*

NLR: neutrophil-to-lymphocyte ratio; CCRT: radiotherapy and chemotherapy; LN: lymph node; Tx: treatment; \*: Statistically significant covariate; T: tumor.

### *2.2. Immunohistochemical (IHC) Staining*

Formalin-fixed and paraffin-embedded tissues, collected at diagnosis from 227 patients with HNSCC who had completed curative treatment, were subjected to IHC analysis (Table 2). The IHC data were assessed using the semiquantitative immunoreactive score, and positive staining was defined as an immunoreactive score ≥ 2 [21]. The clinical end points were overall survival (OS), disease-free survival, and failure pattern. Disease failure was defined as documented locoregional recurrence and/or distant metastases.

**Table 2.** Characteristics of head and neck squamous cell carcinoma (HNSCC) patients with curative-intent treatment.



**Table 2.** *Cont.*

ALDH1. aldehyde dehydrogenase 1; \*: Statistically significant covariate.

### *2.3. MDSC Isolation and Flow Cytometry Analysis*

Peripheral blood samples were obtained from 118 patients with pathologically and clinically confirmed HNSCC (Table 3). To assess the proportion of MDSCs among peripheral blood mononuclear cells (PBMCs), multicolor fluorescence-activated cell sorting (FACS) was performed using the FACS Caliber flow cytometer (BD Biosciences, San Jose, CA, USA). Human low-density neutrophils and granulocytic MDSCs are closely related, and presently there is no generally accepted consensus on mutually exclusive definitions for these cell types [22]. In the majority of oncological studies, human granulocytic MDSCs are characterized as CD14<sup>−</sup>CD15+CD11b+HLA-DR<sup>−</sup> cells [7]. Accordingly, the human MDSC subset characterized as CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cells was sorted from the peripheral blood. The leukocytes were separated from the peripheral blood using a Ficoll gradient before analysis or sorting. Multicolor cell analysis was performed using the following antibodies: PerCP-Cy5.5-conjugated CD14, polyethylene (PE)-conjugated CD11b, and fluorescein isothiocyanate-conjugated HLA-DR. The percentage of MDSCs was measured using multicolor flow cytometry, and isotype-specific antibodies were used as negative controls.


**Table 3.** Characteristics of head and neck squamous cell carcinoma (HNSCC) patients with curative-intent treatment.

MDSC: myeloid-derived suppressor cell; \*: Statistically significant covariate.

### *2.4. Statistical Analysis*

The Kaplan–Meier method was used to plot survival curves, and the log-rank test was used to determine differences in the survival curves between the two groups. The Cox proportional hazard model was used to compute hazard ratios with 95% confidence intervals (CI) after adjustment for esophageal cancer treatment and clinical characteristics. All analyses were conducted using SAS statistical software, version 9.2 (SAS Institute, Cary, NC, USA).

### **3. Results**

### *3.1. Correlations Between the Pretreatment NLR and Clinicopathological Characteristics of HNSCC Patients*

A total of 227 patients with HNSCC were enrolled in this study (Table 1). The median follow-up time was 25.6 months (range 1.37–148 months). There were 114 (50%) patients with clinical tumor stage T1/T2 disease and 140 (62%) with clinical lymph node involvement. Of these patients, 162 (71%) received surgery with or without adjuvant treatment, and the others received definitive RT and chemotherapy. The pretreatment NLR was calculated as the ratio of the absolute neutrophil count to the lymphocyte count. The median pretreatment NLR of the overall cohort was 3.12 (range 0.5 to 31.8). At baseline, 119 (52%) patients had a high NLR of three or higher and 108 (48%) a low NLR less than three. The relationships between the clinicopathological variables and the pretreatment NLR values are shown in Table 1 and Figure 1a. A high NLR at baseline was significantly associated with locoregional recurrence (*p* < 0.001) and a higher risk of death during follow-up (*p* < 0.001). To further examine whether the pretreatment NLR was associated with the outcomes of HNSCC patients after curative treatment, Kaplan–Meier survival analysis was used to compare the low and high NLR subgroups. Patients with a high pretreatment NLR had a shorter overall survival (OS) time (*p* < 0.001; Figure 1b). As shown in Figure 1c,d, a high NLR was significantly associated with a reduced OS rate in both oro/hypopharyngeal cancer (*p* < 0.001) and oral cancer patients (*p* = 0.047). The results of multivariate analyses (Tables 4 and 5) revealed that the pretreatment NLR and surgical resection were significantly correlated with the risk of developing disease failure after treatment and with the OS rate. We further analyzed the predictive value of the NLR according to treatment modality. The data revealed that high NLR was associated with shorter disease-specific survival (DSS) time in patients treated with CCRT and those treated with surgery (Figure 1e,f). Moreover, in the subgroup of surgery, a high NLR was the significant predictor independent of clinical T-stage. In the subgroup of CCRT, a high NLR was associated with shorter disease-specific survival time in patients with advanced tumor stage (*p =* 0.021), but not in those with early tumor stage (*p =* 0.053).

**Figure 1.** Correlations between the baseline neutrophil-to-lymphocyte ratio (NLR) and the prognosis for patients with head and neck squamous cell carcinoma (HNSCC). (**a**) The pretreatment NLR in HNSCC patients. Box-plot showing NLR at baseline was elevated in patients with locoregional recurrence and having higher risk of death during follow-up. The data showed the third quartile (Q3) and first quartile (Q1) range of the data and data outliers. Lines indicate the median values. The survival differences are according to the pre-treatment NLR (NLR ≥ 3 vs. NLR < 3) in (**b**) all, (**c**) subgroup of oral cancer patients, and (**d**) the subgroup of patients with oro-hypopharyngeal cancer. In addition, the differences of disease-specific survival (DSS) are according to the pre-treatment NLR in (**e**) the subgroup of definite CCRT, and (**f**) the subgroup of surgery.


**Table 4.** Adjusted hazard ratio (HR) of determining factors associated with overall survival (OS) of patients with head and neck squamous cell carcinoma (HNSCC).

CI: Confidence Interval; Ref: Reference Group; N: Lymph node staging; \*: Statistically significant covariate.


**Table 5.** Adjusted hazard ratio (HR) of determining factors associated with disease failure of patients with head and neck squamous cell carcinoma (HNSCC).

> Surgery +/− neoadjuvant/adjuvant Tx 0.37 0.22–0.63 <0.001 \* \*: Statistically significant covariate.

### *3.2. Relationships of ALDH1 Expression with the Pretreatment NLR and Clinical Outcome*

We previously reported that positive ALDH1 staining was significantly related to a poor treatment response and higher disease failure rate in oral squamous cell carcinoma (SCC) [21]. Accordingly, we analyzed the predictive role of ALDH1 levels in the clinical outcome and its correlation with the pretreatment NLR in the 227 HNSCC patients. Figure 2a shows representative slides of positive and

negative ALDH1 staining in HNSCC specimens at diagnosis. IHC revealed ALDH1 overexpression in 109 (48%) in these patients. As shown in Table 2, positive ALDH1 staining was significantly associated with the risk of lymph node involvement (*p* = 0.016), a higher rate of locoregional failure (*p* < 0.001), and distant metastasis (*p* = 0.019). As shown in Figure 2b,c, positive staining of ALDH1 was significantly associated with a higher locoregional failure rate and lower OS rate. In the multivariate analysis, positive staining of ALDH1 was significantly associated with a higher risk of developing disease failure and a shorter OS time in HNSCC (Tables 6 and 7). Furthermore, the distribution of the pretreatment NLR was significantly associated with ALDH1 staining in tumor specimens (Table 2 and Figure 2d). Based on the results, we suggest that positive staining of ALDH1 is an independent predictor of shorter survival and a higher rate of disease failure, and a high pretreatment NLR plays a role in predicting ALDH1 expression levels and subsequently a poor prognosis in HNSCC.

**Figure 2.** Relationships between NLR, aldehyde dehydrogenase 1 (ALDH1) expression level, and clinical outcome. (**a**) Representative images of immunohistochemical (IHC) staining with anti-ALDH1 antibodies of oral cancer and hypopharyngeal cancer specimens. Survival differences demonstrated according to the staining of ALDH1 in (**b**) overall survival rate and (**c**) disease failure-free rate. (**d**) NLR levels in the groups of HNSCC patients with and without ALDH1 positive staining in tumor specimens. The data show the third quartile (Q3) and first quartile (Q1) range of the data and data outlier. Lines indicate the median values.


**Table 6.** Adjusted hazard ratio (HR) of determining factors associated with overall survival (OS) of patients with head and neck squamous cell carcinoma (HNSCC).

\*: Statistically significant covariate.

**Table 7.** Adjusted hazard ratio (HR) of determining factors associated with disease failure of patients with head and neck squamous cell carcinoma (HNSCC).


\*: Statistically significant covariate.

### *3.3. Predictive Role of Pretreatment NLR on Levels of CD11b+CD14*<sup>−</sup>*HLA-DR*<sup>−</sup> *Cells in Peripheral Circulation*

Accumulating evidence indicates that MDSCs, a population of cells with suppressive activity, contribute to the negative regulation of immune responses and subsequently to tumor promotion [11]. We previously reported that circulating MDSC levels were significantly increased in patients with HNSCC,

and this was associated with the clinical tumor burden [13]. In the present study, the percentage of CD11b+CD14−HLA-DR<sup>−</sup> cells, a subset of MDSCs, in the peripheral circulation of 118 patients with HNSCC was evaluated by flow cytometry. Representative flow cytometry data from two HNSCC patients are shown in Figure 3a. The mean percentage of CD11b+CD14−HLA-DR<sup>−</sup> cells in the peripheral blood mononuclear cells of the 118 HSCC patients was 11.6 ± 7.4%. As shown Figure 3b,c, the percentage of CD11b+CD14−HLA-DR<sup>−</sup> cells was significantly correlated with the risk of developing disease failure and death after treatment (*p* < 0.001). An increased MDSC level was reported to be associated with attenuating immune surveillance noted in CSC tumors [16,17,21]. Figure 3d shows that the level of MDSCs was significantly higher in the ALDH1-positive group than in the ALDH1-negative group (*p* < 0.001). The 118 patients were further divided into two groups according to the mean CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cell percentage at diagnosis (11.6%): high (≥11.6%) and low (<11.6%) groups. As shown in Table 3, a high percentage of CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cells was associated with a more advanced clinical tumor stage (T3/T4, *p* = 0.005), lymph node involvement (*p* = 0.018), a higher pretreatment NLR, and shorter survival compared with a low CD11b+CD14−HLA-DR<sup>−</sup> cell percentage. In the multivariate analysis, a higher percentage of circulating MDSCs was significantly associated with a higher risk of developing disease failure and a shorter survival in patients with HNSCC (Tables 8 and 9). We further assessed the usefulness of the NLR in predicting the CD11b+CD14−HLA-DR<sup>−</sup> cell percentage. A strong correlation was found between the pretreatment NLR and the percentage of CD11b+CD14−HLA-DR<sup>−</sup> cells in peripheral circulation of HNSCC patients (Figure 3e).

**Figure 3.** Correlation between pre-treatment NLR in the levels of circulating CD11b+CD14- HLA-DR− cells and ALDH1. (**a**) Flow cytometric analysis of circulating CD11b+CD14- HLA-DR− cells in isolated peripheral blood mononuclear cells (PBMCs). HLA-DR−CD11b+ cells were gated, and the CD14 negative population was then selected. Representative data from two cancer patients are shown (upper row, the patient with pretreatment NLR < 3; lower row, the patient with pretreatment NLR ≥ 3). Elevated circulating levels of CD11b+CD14- HLA-DR− cells associated with the higher risk of death (**b**), disease recurrence after treatment (**c**), and ALDH1 positive staining (**d**). (**e**) Positive correlation between the levels of CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cells and pre-treatment NLR in the peripheral circulation.


**Table 8.** Adjusted hazard ratio (HR) of determining factors associated with overall survival (OS) of patients with neck squamous cell carcinoma (HNSCC).

\*: Statistically significant covariate.

**Table 9.** Adjusted hazard ratio (HR) of determining factors associated with disease failure of patients with neck squamous cell carcinoma (HNSCC).


\*: Statistically significant covariate.

### **4. Discussion**

The tumor microenvironment plays an important role in cancer development and progression and may be associated with systemic inflammation [23]. Neutrophils form the first line of host immune defense against bacterial and fungal infections [24]. Compared with their role in host defenses, which is relatively well established, we are just beginning to learn about the precise role of neutrophils in cancer [25,26]. Many recent studies suggested that an elevated NLR is associated with poor survival of subjects with cancer [8,27], including head and neck cancer [10,28]. In the present study, an advantage of our analyses was that the results were based on a relatively large population of HNSCC patients from a single institute, with available information regarding staging and primary treatment details. Based on the analyses of 227 HNSCC patients who received curative treatment, an elevated pretreatment NLR was significantly associated with higher loco-regional recurrence rate and reduced OS rate. According to univariate and multivariate analyses, a pretreatment NLR of three or higher was associated with a shorter OS compared with a NLR below three. Treatment policy included surgery with or without adjuvant treatment for oral cancers and definitive radiation and chemotherapy for oropharyngeal and hypopharyngeal cancers. We further analyzed the predictive value of the NLR according to treatment modality. The data revealed that an increased NLR was a significant predictor for poor prognosis in patients treated with CCRT and those treated with surgery. Based on these results, the NLR is a useful baseline variable for assessing prognosis in HNSCC patients considered for curative treatment.

Circulating blood contains several types of immune cells that participate in the immune response. The interactions among the various populations of immune cells have been recognized as critical in forming the immune microenvironment, which provides the milieu for the anti-cancer immune response to occur [29–32]. MDSCs constitute an immature population of myeloid cells thought to be an important subset of cells that contribute to an immunosuppressive tumor microenvironment, and MDSC numbers are significantly increased in cancer patients [33,34]. Increasing evidence has demonstrated an association between suppressive neutrophils and granulocytic MDSCs related to immune suppression and their relevance to disease [35]. Many of the pro-tumor features of suppressive neutrophils are shared with granulocytic MDSCs, and the distinction between these two cell populations is a matter of intensive debate. We previously found that MDSC recruitment provided a microenvironment conducive to tumor growth and the development of treatment resistance in HNSCC [13]. To date, granulocytic MDSCs have been defined mainly as CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cell lineages in human cancers [7,22,35]. Accordingly, we characterized the proportions of CD11b+CD14−HLA-DR<sup>−</sup> myeloid cells in a cohort of HNSCC patients. FACS analyses revealed that the percentage of these MDSCs was correlated with the clinical tumor burden, disease status, and survival. We further demonstrated a positive correlation between the NLR and circulating CD11b+CD14<sup>−</sup>HLA-DR<sup>−</sup> cell level in HNSCC patients. In the present study, we showed that the pretreatment NLR was related to the circulating CD11b+CD14−HLA-DR<sup>−</sup> cell level and disease progression.

Much of the relationship between immune cells, either circulating or within tumors, and disease outcome in cancer can probably be explained by the inflammatory response that is secondary to the cancer. CSCs are becoming recognized as being responsible for metastasis and treatment resistance [36,37]. ALDH1, a detoxifying enzyme, has been identified as a novel CSC-like cell marker and is relevant to the prognosis of cancers [19,20,38,39]. Immune evasion was reported to play a role in the contribution of CSCs in tumor promotion [16,40]. CSCs can recruit cells that suppress the immune system, such as the activation of myeloid-derived suppressor cells (MDSCs), to attenuate immune surveillance [16–18,41–43]. Our previous data revealed correlations between ALDH1 expression levels and treatment resistance, CSC-like properties, higher circulating MDSC levels, and poor prognosis in oral SCC. Accordingly, we further examined the predictive value of ALDH1 for HNSCC prognosis and the correlations of the ALDH1 level with the MDSC level and NLR. By analyzing the clinical outcomes of 227 patients with HNSCC, the elevated expression of ALDH1 was correlated with a higher incidence of lymph node involvement, higher disease failure rate, and lower survival rate. Moreover, there were significant correlations among ALDH1 IHC staining, the levels of circulating MDSCs, and NLR. Patients with a higher NLR had a higher ALDH1 level in their tumors and more MDSCs in the peripheral circulation, which are associated with poor prognosis of HNSCC. The current study is limited by the inherent nature of investigating a hospital-based registry and the nonrandomized approach to treatment selection. Furthermore, we could not account for potential unmeasured selection biases regarding performance status, comorbidity, access to health care, or other patient-related factors.

### **5. Conclusions**

With the increasing use of personalized therapy, patient selection has become an important issue in assessing efficacy. The targeted therapy of CSCs may enhance the treatment response, thereby resulting in improved clinical outcomes in patients with HNSCC. In addition, MDSCs have been suggested to be a novel target for multiple cancers, and numerous clinically available agents have been developed [34]. Thus, it is imperative to identify clinically feasible parameters highly relevant to the characteristics of CSC and the level of MDSCs. The NLR is a cheaper and faster laboratory measure compared with other biomarkers, and it does not involve any additional cost. In the present study, we showed that the NLR was relevant to ALDH1 and MDSC levels and a strong prognostic indicator for HNSCC patients. Discussions based on pretreatment NLR results may help the patient decide whether the side effects of curative treatments are worth the risk. We suggest the NLR to be an important biomarker for patients that can assist the clinician and patient to make informed decisions regarding treatment options.

**Author Contributions:** Conceptualization, M.-F.C., M.-S.T. and W.-C.C.; Data curation, M.-S.T.; Formal analysis, P.-T.C.; Funding acquisition, M.-F.C.; Investigation, M.-F.C.; Methodology, W.-C.C.; Writing—original draft, M.-S.T.; Writing—review & editing, M.-F.C.

**Funding:** The grant was support by Chang Gung Memorial Hospital, Taiwan, grant CMRPG6E0372-3.

**Acknowledgments:** The authors would like to thank the Health Information and Epidemiology Laboratory (CLRPG6G0041) for their comments and assistance in data analysis.

**Conflicts of Interest:** The authors declare that they have no competing interests. All authors read and approved the final manuscript.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Role of Diffusion-Weighted Magnetic Resonance Imaging in the Differentiation of Head and Neck Masses**

### **Lutfi Kanmaz <sup>1</sup> and Erdal Karavas 2,\***


Received: 22 April 2018; Accepted: 21 May 2018; Published: 29 May 2018

**Abstract:** The purpose of this study was to evaluate the value of diffusion-weighted MRI (DW-MRI) in differentiating benign and malignant head and neck masses by comparing their apparent diffusion coefficient (ADC) values. The study included 32 patients with a neck mass >1 cm in diameter who were examined with echo planar DW-MRI. Two different diffusion gradients (b values of b = 0 and b = 1000 s/mm2) were applied. DWI and ADC maps of 32 neck masses in 32 patients were obtained. Mean ADC values of benign and malignant neck lesions were measured and compared statistically. A total of 15 (46.9%) malignant masses and 17 (53.1%) benign masses were determined. Of all the neck masses, the ADC value of cystic masses was the highest and that of lymphomas was the lowest. The mean ADC values of benign and malignant neck masses were 1.57 × <sup>10</sup>−<sup>3</sup> mm2/s and 0.90 × <sup>10</sup>−<sup>3</sup> mm2/s, respectively. The difference between mean ADC values of benign and malignant neck masses was significant (*p* < 0.01). Diffusion-weighted MRI with ADC measurements can be useful in the differential diagnosis of neck masses.

**Keywords:** neck mass; diffusion-weighted MRI; apparent diffusion coefficient

### **1. Introduction**

Quick and accurate diagnosis directly affects treatment success for patients with a neck mass, which is a common finding in ENT clinics. Inadequate or late diagnosis of a malignant mass increases the morbidity and mortality of a disease.

The rapid development of diagnostic imaging technology has provided clinical practice with new facilities for the evaluation of neck masses. These new methods are gaining importance with the advantageous factors of cost and ease of use. At present, ultrasonography (USG) and/or computed tomography (CT) are used as conventional methods for the evaluation of neck lesions.

If necessary, magnetic resonance imaging (MRI) is used for the characterization of neck masses. MRI evaluates the morphology, signal intensity and enhancement pattern of lesions. However, none of these methods can accurately differentiate benign from malignant lesions. This has led to the necessity of researching new diagnostic methods. Diffusion-weighted magnetic resonance imaging (DW-MRI) is a non-contrast enhanced technique that can be obtained during a single breath-hold. In the literature, DW-MRI was first used in the early diagnosis of stroke in neuroradiology [1,2]. In the early period, the use of this technique was limited in the central nervous system due to its sensitivity to cardiac, respiratory, and peristaltic movements. However, following improvements in the echo planar imaging technique as a fast MRI sequence, it became possible to successfully apply diffusion-weighted echo planar MRI even in other areas with high-susceptibility artifacts. DW-MRI was first applied to head and neck lesions in 2001 and promising results have been achieved [3]. Subsequent studies showed that

DW-MRI appeared to be helpful in differentiating epidermoid carcinoma and malignant lymphoma, staging neck nodal disease, and distinguishing radiotherapy-induced tissue changes from persistent or recurrent cancer. In these studies, apparent diffusion coefficient (ADC) values of tissues and lesions are calculated using diffusion-weighted images and different values in the differential diagnosis. Moreover, with the use of this imaging technique, the creation of an ADC map is an excellent method for differentiation between the viable and necrotic parts of head and neck tumors. Thus, the ADC map can be used to select the best biopsy site and to detect tumor viability in the post-treatment follow-up of patients after radiation therapy. The technique may also be useful in characterizing thyroid nodules and salivary gland neoplasms.

The purpose of this study was to evaluate the value of DW-MRI in differentiating benign and malignant head and neck masses by comparing their ADC values.

### **2. Materials and Methods**

This prospective study was performed on 43 consecutive patients who underwent MRI for a diagnosis of head and neck lesions in our center. All patients were examined with contrast-enhanced MRI and DW-MRI. The study was conducted in the Department of Otorhinolaryngology, Bakırköy Dr. Sadi Konuk Training and Research Hospital, in the period of June 2009 to June 2010.

Institutional Ethics Committee approval was obtained for the study.

### *2.1. Subjects*

A total of 11 patients were excluded from the study; four patients with neck masses <1 cm in the greatest minimal transverse diameter, two who had undergone biopsy, three due to distortion artifacts, and two with a final diagnosis of neck metastasis of a thyroid carcinoma.

Thus the final study population of 32 consecutive patients with neck masses >1 cm in diameter consisted of 12 females (37.5%) and 20 males (62.5%) with a mean age of 45.13 ± 17.08 years (range, 9–78 years). All patients were questioned in detail about age, location, and duration of the mass, associated symptoms, and then routine blood tests such as serological tests were applied. Within the head and neck examination, diagnostic pan-endoscopy of the nasal cavity, nasopharynx, oropharynx, hypopharynx, and larynx was also performed. All clinical evaluations were documented.

When patients had multiple neck masses with the same histological diagnosis, only the largest one was used for calculation of ADC values. Thus, the diffusion-weighted images and ADC maps of 32 neck masses in 32 patients were studied.

Localization of the lesions was classified according to the lymph node regions and neck facial spaces. The final diagnosis of the patients was made by histopathological examination of surgical specimens. A diagnosis of tuberculosis lymphadenitis was made by histology and culture, two undifferentiated nasopharyngeal carcinoma metastases by primary tumor biopsy and FNAB, neck metastasis in five patients with NHL diagnosed by excisional lymph node biopsy, and one adenocarcinoma metastasis by FNAB. A diagnosis of SCC metastasis was confirmed with neck dissection. Diagnosis of carotid body paraganglioma in one patient was established with MR angiography and DSA (digital subtraction angiography) before excision.

### *2.2. MR Imaging Techniques*

All the MR examinations were performed with a 1.5 Tesla MR unit (Siemens Avanto, Erlangen, Germany). Routine examination consisted of T2-weighted fast spin-echo images (with a section thickness of 4 mm, an interslice gap of 1–2 mm, a field of view (FOV) of 25–30 cm and an acquisition matrix of 256 × 224) and DW-MRIs. Before DW-MRI, T2-weighted images were obtained in the axial plane. A total of 14 transverse images covering the lesions were obtained. DW-MRIs were obtained at the section level where the largest transverse section of the lesion was detected on the MRIs which were obtained before administration of contrast material. DW-MRI was obtained using multi-slice spin-echo single-shot echo planar imaging in the axial plane. For each patient, diffusion-weighted

images and ADC maps were obtained by applying diffusion-sensitive gradients in three orthogonal directions (x, y, and z) and two different b-values (0 and 1000 s/mm2). ADC maps of the images were automatically reconstructed on the main console. Then, the region of interest (ROI) was defined by a radiologist measuring the signal intensity of the lesion. ROIs were determined on the solid appearing parts for the solid masses and on the cystic areas for the cystic lesions. The ADC values of the lesions were calculated with an ROI >1 cm2.

### *2.3. Statistical Analysis*

Statistical analyses of the study data were performed using NCSS (Number Cruncher Statistical System) 2007 and PASS (Power Analysis and Sample Size) 2008 Statistical Software (HyLown Consulting LLC., Atlanta, GA, USA). The Student's *t*-test was used to compare data between two groups. Results were stated as the mean and standard deviation. Qualitative data were compared using the chi-square test. The receiver operating characteristic (ROC) curve was applied to determine the cut-off point with the highest accuracy and sensitivity. The value of *p* < 0.05 was considered statistically significant at a 95% confidence level.

### **3. Results**

This study was performed on 32 consecutive patients with head and neck masses who underwent echo planar DW-MRI from June 2009 to June 2010. The patients consisted of 12 females (37.5%) and 20 males (62.5%) with a mean age of 45.13 ± 17.08 years (range, 9–78 years).

Malignant masses were determined in 15 (46.9%) cases and benign masses in 17 (53.1%). The benign masses consisted of five pleomorphic adenoma originating from major salivary glands, three reactive lymphadenopathies, two branchial cleft cysts, two cervical sympathetic chain schwannomas, two Whartin's tumors, one glomus tumor, one tbc lymphadenitis, and one thyroglossal duct cyst. The malignant masses were five Non-Hodgkin's lymphoma, three larynx SCC met., two undifferentiated carcinoma met., two oropharynx SCC met., one GIS adeno ca met., one primary unknown carcinoma met., and one tonsil SCC met. The diagnoses of the patients are listed in Table 1.

Of the total neck masses, the ADC value of cystic masses was the highest (1.98 × <sup>10</sup>−<sup>3</sup> mm2/s) and that of lymphomas (0.80 × <sup>10</sup>−<sup>3</sup> mm2/s) was the lowest. The mean ADC values of benign and malignant neck masses were 1.57 × <sup>10</sup>−<sup>3</sup> mm2/s and 0.90 × <sup>10</sup>−<sup>3</sup> mm2/s, respectively. The difference between the mean ADC value of benign and malignant neck masses was statistically significant (*p* < 0.01). The localizations of the masses are listed in Table 1. The numbers of malignant and benign masses were 15 (46.9%) and 17 (53.1%), respectively. The mean ADC value of benign masses with high signal intensity was statistically significantly higher than that of malignant masses with low signal intensity (*p* < 0.01) (Table 2). Malignant masses were classified in two categories as malignant lymphomas (33.3%) and carcinomas (66.7%, squamous cell carcinoma or adenocarcinoma). There was no statistically significant difference between the two categories in the malignant group (*p* > 0.05) (Table 2). When an ADC value of 1.13 × <sup>10</sup>−<sup>3</sup> mm2/s or less was used to predict malignancy, the best results were achieved with high accuracy, with sensitivity of 93.33%, specificity of 82.35%, positive predictive value of 82.35%, and a negative predictive value of 93.33% (Table 3).

The ROC curve was used to evaluate the diagnostic capability of the ADC value to differentiate benign from malignant masses. When 1.13 × <sup>10</sup>−<sup>3</sup> mm2/s was used as a threshold value in differentiating benign from malignant masses, the area under the curve was 0.918 (Table 4, Figure 1).


**Table 1.** Diagnosis and localization of 32 head and neck masses.

**Table 2.** Pathological diagnosis and distribution of mean ADC values in the examined groups.


+ Student *t* test; \*\* *p* ≤ 0.001.

**Table 3.** Calculating the threshold value.


**Table 4.** Area under the curve (AUC).


**Figure 1.** Receiver operating characteristic (ROC) curve of the ADC value.

### **4. Discussion**

Diffusion is defined as the randomized microscopic movement of water molecules and is used as a sensitive parameter for the characterization of tissue at a microscopic level. Today, in vivo measurement of diffusion is possible with DW-MRI and ADC measurements. As a result of new technological developments, MRI has become sensitive to the diffusion of water protons in biological tissues and diffusion-weighted imaging can be obtained. Intracellular and extracellular water balance is also shown in a way that is important for diagnosis and follow-up of stroke. Initially, the use of this technique was limited to brain studies because of technical problems regarding motion artifact due to cardiac, respiratory, and peristaltic movements. However, following improvements in echo planar imaging techniques, an echo planar DW-MRI can now be successfully performed even in areas with high susceptibility artifacts [4]. This technique was first used in neuroradiological imaging for diagnosis of early cerebral ischemia and has become a diagnostic tool in this area [1,5].

In 1994, Muller et al. measured the ADC of water in liver, spleen, kidney, and muscle and showed that in vivo diffusion measurements of abdominal organs obtained with MRI could prove helpful in the identification and classification of abdominal disease [6]. Subsequently, in several studies, DW-MRI has been shown to be able to be used in the differential diagnosis of lesions in the liver, kidney, and other abdominal organs with the measurement of ADCs [7,8].

In the literature, DW-MRI has also been seen to have an application area in the different regions of the head and neck. The characterization of head and neck lesions with echo planar DW-MRI by Wang et al. [3] was the first study in this area. They found that the mean ADC value of the benign lesions was statistically significantly different than that of malignant lesions. In 2003, Sumi et al. [9] studied the differential diagnosis of metastatic lymph nodes with diffusion-weighted MR imaging. Over several years, further studies have been made in this area. In another study on neck lymph nodes, ADC maps, as a new technique, have been used to determine the necrotic and non-necrotic solid areas of malignant lesions [10]. Eida et al. [11] reported that preoperative tissue characterization of salivary gland tumors could be made with ADC map construction. In another study, Abdel Razek et al. reported that the mean ADC value of malignant thyroid nodules was statistically significantly lower than that of benign ones [12]. The diffusion technique involves the diffusion motion of water protons in

the tissues. According to the diffusion of tissue, the diffusion of water molecules varies in the different regions of tissue. Therefore, the diffusion coefficient of the tissue varies depending on any change in the proportion of extracellular to intracellular water molecules. Thus, diffusion-weighted MR imaging produces different contrast and ADC values according to the microstructure of the tissues [3].

In the current study, the mean ADC value of benign masses with high signal intensity was significantly (Figure 2) higher than that of malignant (Figure 3) masses with low signal intensity. These differences in ADC values may be explained by the differences in the histopathological characteristics of benign and malignant tumors. Generally, malignant tumors show hypercellularity and have enlarged nuclei, and hyperchromatism. These histopathological characteristics reduce the diffusion space of water protons in the extracellular and intracellular regions [13,14].

**Figure 2.** Pleomorphic adenoma of the parotid gland. (**a**) Axial T2-weighted image (T2WI) shows a mass in the left parotid gland; (**b**) the lesion shows low signal intensity on DWI; and (**c**) the mass is hyperintense on the ADC map (ADC value of 1.55 <sup>×</sup> <sup>10</sup>−<sup>3</sup> mm2/s).

**Figure 3.** Undifferentiated nasopharyngeal carcinoma. (**a**) Axial T2WI shows bilateral metastatic cervical lymph nodes; (**b**) lymph node, on the left side of the neck, shows high signal intensity on DWI; and (**c**) the mass is hypointense on the ADC map (ADC value of 0.91 <sup>×</sup> <sup>10</sup>−<sup>3</sup> mm2/s).

Apparently higher ADC values for benign cystic masses may be expected because of the relatively freer mobility of water protons in the fluid. In the current study, cystic masses were not grouped separately due to low numbers. However, consistent with previous studies, the mean ADC value of three cystic masses (Figure 4) (1.98 × <sup>10</sup>−<sup>3</sup> mm2/s) was higher than that of other benign solid masses (ADC = 1.48 × <sup>10</sup>−<sup>3</sup> mm2/s). In addition, the differences in ADC values among cystic masses could be explained by the different protein concentrations. A high protein level restricts the movement of water molecules by increasing the viscosity [15].

**Figure 4.** Second branchial cleft cyst. (**a**) Axial T2WI shows a unilocular cystic mass in the left carotid space; (**b**) the lesion shows low signal intensity on DWI; and (**c**) the lesion is hyperintense on the ADC map (ADC value of 2.02 <sup>×</sup> <sup>10</sup>−<sup>3</sup> mm2/s).

Sumi et al. [16] reported that the mean ADC value of lymphomas is lower than that of metastatic lymph nodes of carcinomas due to the difference in cellular density. Maeda et al. [17] also reported that carcinomas could contain small foci of necrosis on histopathological examination that was not identifiable on conventional MRI. This investigation has also been used to explain the higher ADC values of SCCs than those of lymphomas. In the current study, although there was no statistically significant difference, the mean ADC value of lymphomas (Figure 5) (0.80 × <sup>10</sup>−<sup>3</sup> mm2/s) was lower than that of the other malignant tumors (0.96 × <sup>10</sup>−<sup>3</sup> mm2/s). In the malignant group, larynx SCC metastasis with the value of 1.15 × <sup>10</sup>−<sup>3</sup> mm2/s showed the highest values.

**Figure 5.** Non-Hodgkin lymphoma. (**a**) Axial T2WI shows metastatic cervical lymph node nearby SCM; (**b**) the lymph node, on the right side of the neck at level of lateral cervical region, shows high signal intensity on DWI; and (**c**) the mass is hypointense on the ADC map (ADC value of 0.62 <sup>×</sup> <sup>10</sup>−<sup>3</sup> mm2/s).

One of the considerations that must be taken into account when using diffusion-weighted MRI for the differential diagnosis of masses is the recognition of the fact that some malignant masses behave like a benign tumor independently of the 'cut-off' ADC value obtained at the end of the study. These masses are some primary or metastatic SCC and thyroid carcinoma metastases [3]. In the current study, a case with diagnosis of larynx SCC metastasis in the ADC value of 1.20 × <sup>10</sup>−<sup>3</sup> mm2/s showed the feature of a benign tumor. The reason that the high ADC values are non-standardized might be the presence of small foci of micronecrosis and the hypervascular tumor portions that escalate the perfusion effect in some malignant tumors and of dense extracellular fluid in follicular components in

thyroid carcinomas [3]. Considering this situation, the macroscopic solid portions were determined using MRI observations and the ADC measurements were made, and the cases with neck masses of thyroid origin were excluded from the study. Although DW-MRI is more reliable than that of other MR imaging techniques to identify micronecrosis of primary or metastatic tumors, [18–20] the differentiation between the viable and necrotic parts of head and neck tumors with DW-MRI by reconstructing ADC maps is possible for accurate biopsy results [21]. In addition, DW-MRI with ADC measurement may be used for differentiating residual or recurrent head and neck tumors from postoperative or postradiation changes [22].

With concern for inaccurate ADC measurements in small lesions according to the susceptibility artifacts and image distortions as a limitation of this imaging technique, neck masses <1 cm in the greatest minimal transverse diameter on MRI were excluded from the study. In a study by Chen et al., performed to compare periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) DW-MRI and echo planar DW-MRI techniques, it was suggested that it was possible to reduce the distortion of head and neck masses through PROPELLER diffusion-weighted MRI to a large extent [23].

One of the limitations of the current study was the area of necrosis that showed falsely higher ADC values, especially in the center of metastatic neck masses [10]. Therefore, ADC values were measured by selecting the solid portions of tumors.

In the study, three masses in the benign group indicated lower ADC values than the cut-off value in accordance with the malignant mass. In one of these masses, which was tuberculous lymphadenopathy (ADC = 0.92 × <sup>10</sup>−<sup>3</sup> mm2/s); restriction of diffusion could be explained by the presence of inflammatory cells in the pus that reduced the diffusion space of water protons [24]. In another study, this situation was explained by the thickness of the caseous material of granulomatous lesions [25]. Whartin tumor (ADC = 0.88 × <sup>10</sup>−<sup>3</sup> mm2/s) was the other case that was falsely diagnosed as a malignant tumor (Figure 6). Intense lymphoid accumulation in the stroma and proliferation of the epithelial component could be the reason for the limited motion of the water protons in the extracellular space [3]. Reactive lymphadenitis was the third case. A varying amount of fibrosis in the stroma of inflammatory cells could have resulted in the low ADC value by restriction of the diffusion of water molecules [3].

**Figure 6.** Warthin tumor. (**a**) Axial T2WI showing the mass localised in the superficial lobe and spreading into the deep lobe of the left parotid gland; (**b**) the lesion shows high signal intensity on DWI; and (**c**) the lesion is hypointense on the ADC map (ADC value of 0.88 <sup>×</sup> <sup>10</sup>−<sup>3</sup> mm2/s) and the tumor was falsely diagnosed as a malignant lesion.

In the current study, in cases with malignant diseases, the mass appeared hyperintense on diffusion images (obtained at b = 1000 s/mm2) and with low signal intensity on ADC maps and, conversely, benign masses appeared hypointense and hyperintense, respectively. There was a

statistically significant difference in ADC values between the malignant masses and benign lesions (*<sup>p</sup>* < 0.01). When an ADC value of 1.13 × <sup>10</sup>−<sup>3</sup> mm2/s or less was used to predict malignancy, the best results were achieved with high accuracy, with 93.33% sensitivity, 82.35% specificity, 82.35% positive predictive value, and 93.33% negative predictive value.

In this study, the patient groups were heterogeneous with different histopathological entities and there was a limited number of cases in each benign or malignant group. Further investigations of a larger series with a specific group of the same pathological diagnosis are necessary.

### **5. Conclusions**

In summary, DW-MRI seems to be a promising non-invasive imaging technique for characterization of head and neck masses or for any other subjects as discussed above. It can be concluded that further studies on larger series and advances of diffusion MR techniques to improve the image quality would help DW-MRI to become a routine imaging technique.

**Author Contributions:** L.K. conceived and designed the experiments; L.K. and E.K. performed the experiments; L.K. analyzed the data; L.K. and E.K. contributed reagents/materials/analysis tools; and L.K. wrote the paper.

**Acknowledgments:** The authors would like to thank Filiz Saçan for help and technical support. The authors declare that this study has not received any financial support.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Canaloplasty in Corticosteroid-Induced Glaucoma. Preliminary Results**

### **Paolo Brusini 1,\*, Claudia Tosoni <sup>1</sup> and Marco Zeppieri <sup>2</sup>**


Received: 11 December 2017; Accepted: 7 February 2018; Published: 11 February 2018

**Abstract:** Purpose: to present the mid-term results of canaloplasty in a small cohort of corticosteroid glaucoma patients. Material and Methods: Nine eyes from seven patients with various types of corticosteroid glaucoma in maximum medical therapy underwent canaloplasty. Patients underwent complete ophthalmic examination every six months. Success was defined as: post-operative intraocular pressure (IOP) ≤ 21 mmHg and ≤ 16 mmHg without ("complete success"), and with/without medical treatment ("qualified success"). The IOP reduction had to be ≥ 20. The number of medications before and after surgery was considered. The follow-up mean period was 32.7 ± 20.8 months (range 14–72 months). Results: The pre-operative mean IOP was 30.7 ± 7.2 mmHg (range: 24–45). The mean IOP at 6 and 12-month follow-up was 13.1 ± 2.6 mmHg, and 13.7 ± 1.9 mmHg, respectively. Qualified and complete success at 6 and 12 months was 100% for both of the two definitions. The number of medications used preoperatively and at the 12-month follow-up was 4.3 ± 0.7, and 0.2 ± 1.0, respectively. No serious complication was observed. Conclusions: The mid-term results of canaloplasty in patients with corticosteroid-induced glaucoma appear to be very promising. Canaloplasty should be considered as a possible alternative to filtering surgery in this form of glaucoma, when medical therapy is not sufficient to maintain the IOP within reasonable limits.

**Keywords:** canaloplasty; non-perforating surgical procedures; corticosteroid-induced glaucoma; Schlemm's canal

### **1. Introduction**

Corticosteroid-induced glaucoma is a quite common form of secondary glaucoma due to either systemic or, more frequently, topical, peri- or intraocular administration of glucocorticoids in predisposed subjects [1–5]. It is known that corticosteroids raise intraocular pressure (IOP) by lowering the facility of aqueous outflow. Quite a high percentage of normal subjects (ranging from 5% to over 40% depending on the definition of corticosteroid-responders [6,7] may undergo a significant increase of IOP after using topical corticosteroids for several days. The increasing use of intravitreal injections of triamcinolone acetonide and intravitreal implants of dexamethasone for exudative maculopathies will probably exacerbate this problem. A secondary glaucoma can develop in some cases, even though for most patients the IOP returns to baseline after ceasing steroid use. If traditional medical therapy is not able to lower IOP within the safe range, structural and functional damage can quickly develop. In these cases, a laser trabeculoplasty can be attended [8–10], but more often a surgical treatment must be performed before serious visual impairment occurs. Trabeculectomy with intra-operative antimetabolites is still considered to be the gold standard surgical procedure for different types of glaucoma, including corticosteroid-induced glaucoma [11,12]. This technique is

simple to perform and effective, however, several early and late potentially serious complications can occur. In particular, problems related to the subconjunctival bleb, and the frequent development of a cataract can be particularly disturbing in young patients, who are often the subjects that develop corticosteroid-induced glaucoma.

Canaloplasty is a non-perforating bleb-less technique, introduced some years ago, in which a 10-0 prolene suture is positioned and tensioned within Schlemm's canal, previously dilated with a viscoelastic agent, thus facilitating aqueous outflow through natural pathways [13,14].

The purpose of this study is to present the mid-term results of canaloplasty in a small cohort of patients with corticosteroid-induced glaucoma resistant to medical management.

### **2. Experimental Section**

In this non-randomized prospective study, 9 eyes from 7 patients with uncontrolled corticosteroid-induced glaucoma under maximum tolerated medical therapy underwent canaloplasty under peribulbar anesthesia. Surgery was performed either at the Department of Ophthalmology at the Azienda Ospedaliero-Universitaria "Santa Maria della Misericordia" Hospital in Udine (Italy), or at the "Città di Udine" Health Center in Udine (Italy) by one surgeon (PB), with a 10-year experience with this type of surgery, from February 2008 to July 2016.

All patients provided written informed consent. The protocol was approved by the institutional review board or ethics committee.

Inclusion criteria included: (1) patients with ocular hypertension due to corticosteroid use; (2) IOP ≥ 24 mmHg with maximum tolerated medical therapy after stopping corticosteroid use (in the case of topical administration); (3) open irido-corneal angle; (4) age over 18 years. Patients with or without typical optic nerve alterations, and with or without glaucomatous visual field defects were considered.

Exclusion criteria included: (1) elapsed time after steroids stopping shorter than 3 months; (2) age under 18 years; (3) other possible causes of glaucoma (i.e., pseudoesfoliation, previous trauma, etc.); (4) previous ocular surgery, apart from cataract and intravitreal injections; (5) refusal to undergo surgery.

The patients' demographic data, causes of secondary glaucoma, preoperative IOP, number of drugs used, visual functions and length of the follow-up at the last visit and other clinical data are reported in Table 1.


**Table 1.** Patient data before surgery.

OHT = ocular hypertension; Pre-op IOP = preoperative intraocular pressure; Dex IV CRVO = Intravitreal dexametason for central retinal vein occlusion; Triam IV EM = Triamcinolone acetonide for exudative maculopathy; Dex drops AC = dexametason drops for allergic conjunctivitis; Dex drops PK = dexametason drops after perforating keratoplasty; VA = visual acuity; VF = visual field; GSS 2 = Glaucoma Staging System 2; NA = not applicable; CF = counting fingers; SD = standard deviation.

The canaloplasty surgical technique is well known and has been extensively reported in the literature [15–19]. Briefly, surgery starts with a fornix-based conjunctival flap, and a 3 × 4 mm superficial scleral flap, which is dissected forward into the clear cornea for 1.5 mm. A deep scleral flap is then created, in order to open the Schlemm's canal. The deep scleral flap is removed and the two openings of the canal are dilated with hyaluronic acid in order to cannulate the Schlemm's canal, by means of a special microcatheter (iTrack by iScience Interventional, Menlo Park, CA, USA), connected to a flickering red laser light source for easy identification of the distal tip through the sclera. The microcatheter is inserted and pushed forward within Schlemm's canal for the entire 360◦ until it comes out of the other end of the canal opening. A 10-0, prolene suture is then tied to the distal tip and the microcatheter is withdrawn back through the canal in the opposite direction. During this maneuver, a small amount of high-molecular weight viscoelastic agent (Healon GV, Abbot Medical Optics, Santa Ana, CA, USA) is delivered within the canal every two hours of circumference, using a special screw-driven syringe. The suture is then knotted under tension in order to inwardly distend the trabecular meshwork. The superficial scleral flap is sutured with seven 10-0 Polyglactin 910 stitches to ensure a watertight closure in order to prevent any bleb formation. The conjunctival flap is then sutured with two 10-0 sutures to complete the surgery.

All patients underwent a visit once a week for the first month in order to measure IOP and detect any possible postoperative complication, then went on to have a complete ophthalmic examination every six months, including slit-lamp examination, best corrected visual acuity (BCVA), IOP measurement with Goldmann applanation tonometer, fundus examination with a 78 D Volk lens, visual field testing (Humphrey Field Analyzer (Carl Zeiss Meditec Inc. Dublin, CA, USA) 30-2 SITA standard test), retinal nerve fiber layer assessment with spectral-domain OCT and gonioscopy. Moreover, the corneal astigmatism was measured after one week and one month by means of a keratometer of Javal and corneal topography. Visual field damage severity was assessed using the Glaucoma Staging System 2 (P. Brusini, Italy) [20]. The definition of success was based on two different criteria: post-operative IOP ≤ 21 mmHg and ≤ 16 mmHg. When this goal was obtained without any medical treatment, the success was defined as "complete". When the same IOP levels were obtained with or without medical treatment, the success was defined as "qualified". Moreover, the IOP reduction had to be ≥20% for defining a case as successful. The number of medications before and after canaloplasty was also taken into consideration.

Differences between test results were calculated using the paired *t*-test for variables that showed a normal distribution. The statistical analysis was performed using SPSS 11.0 (IBM Analytics, Chicago, IL, USA). Statistical significance was defined as *p* < 0.05.

### **3. Results**

The entire standard procedure could be performed as planned in all of the nine eyes. Follow-up ranged from 14 to 72 months (mean: 32.7 ± 20.8). The mean pre-operative IOP was 30.4 ± 6.8 mmHg, ranging from 24 to 45 mmHg. The mean IOP after 6 and 12 months was 13.1 ± 2.6 mmHg, and 13.7 ± 1.9 mmHg, respectively, ranging 11 to 18 mmHg (paired *t*-test, *p* = 0.0001). The mean IOP reduction from baseline after 6 and 12 months was of 56.9% and 54.9%, respectively. The IOP values at various follow-up sessions within a period of 36 months are shown in Table 2. The scatter plot in Figure 1 shows the pre- and one-year post-operative IOP values. After the 6 and 12-month follow-up, a complete and qualified success, was obtained in all 9 eyes, using both the definitions of success (IOP ≤ 21 mmHg and ≤ 16 mmHg), with an IOP within normal limits during the entire follow-up period, except for one eye that showed an increase of IOP after two years, successfully controlled with medical therapy. The number of medications used pre- and at the 12-month follow-up was 4.3 ± 0.7, and 0.2 ± 1.0, respectively (difference statistically significant, *p* < 0.001). Only one patient (11%) was under IOP-lowering drops after one year, but medical treatment was needed in both the two patients which reached a five-year follow-up. No patient required adjunctive surgical procedures. Gonioscopy showed that the prolene suture remained correctly positioned within the Schlemm's canal for the entire follow-up period in all cases. At the last visit, visual acuity worsened of two lines in two eyes (case #1 and case #2), depending on retina conditions, and improved of three lines in one eye (case #7) due to the effects of corticosteroid treatment. A reliable visual field testing

was not possible in three eyes due either to a poor visual acuity (case #9) or to artifacts related to retinal disease (case #1 and #2). Before surgery, visual field was normal in one case (#5), showed only slight defects in four cases, whereas only small islands of vision were present in another case (#9). All of these defects remained stable over time. Optic nerve appearance (normal in all eyes but one) and retinal nerve fiber layer did not show any significant change during the follow-up period.


**Table 2.** Mean post-operative intraocular pressure (IOP).

\* *p* < 0.0001 vs. preoperative values.

**Figure 1.** Pre- and one-year post-operative IOP values.

The early post-operative complications (within four weeks from surgery) included: microhyphema in two eyes (22.2%); hypotonus (IOP < 5 mm/Hg) in one eye (11.1%); and, IOP spikes > 10 mmHg in one case (11.1%). A transient decrease in visual acuity in the first weeks after canaloplasty was a rather common finding, which was due to an induced according to-the-rule astigmatism which can reach five diopters, but usually disappear within one month. No surgery-related complications were observed after two months.

### **4. Discussion**

Surgery is sometimes needed to control ocular hypertension and delay damage progression in patients with corticosteroid-induced glaucoma, especially considering that visual field defect progression can be fast and severe if IOP is very high. However, unlike from patients with primary open-angle glaucoma or pseudoesfoliation glaucoma, which often show advanced visual field loss, patients with corticosteroid-induced glaucoma usually have normal optic nerves and visual fields at the beginning. For this reason, an IOP in the mid-teens is usually adequate in order to avoid any structural and/or functional damage. In this type of patient, even with very high pre-operative IOP levels, non-filtering surgical procedures, such as goniotomy [21], trabeculotomy [22–25], trabecular stents [26], viscocanalostomy [27] or deep sclerectomy [28] may be an interesting option, even if they are less effective than trabeculectomy in lowering IOP, considering the lower risk of complications. Nowadays, canaloplasty should be considered as a step ahead of these procedures with very interesting long-term outcomes in various forms of open-angle glaucoma [15–19].

Our mid-term results in a small cohort of patients with corticosteroid-induced glaucoma unresponsive to medical therapy appear to be particularly good in comparison with other types of glaucoma, where the mean IOP usually ranges between 15 and 17 mmHg, with a percentage of success after one year ranging between 60% and 95%, depending on the definition of success used [14–19]. In particular, if a cut-off of ≤ 16 mmHg is taken in order to define successful cases, the percentage of qualified success reported in literature is about 50% in comparison with the 100% obtained in our nine cases.

The reasons for this favorable behavior are probably various and include: (1) histopatologic studies in corticosteroid-induced glaucoma demonstrated an increased density of the cribriform meshwork and thinning of the endothelial lining of Schlemm's canal [29,30]; in cases of elevated IOP, a collapse of aqueous plexus and collector channel ostia obstructed by herniation was observed in bovine eyes [31], resulting in a decrease in the effective filtration area. Canaloplasty is able to overcome this obstacle, allowing the restoration of the aqueous humor outflow; (2) patients with this type of glaucoma are usually relatively young with well-functioning aqueous humor pathways, which is a fundamental requirement to obtain satisfactory results after canaloplasty; (3) all of our patients were under medical therapy for a short period before the operation; it is well known that topical therapy for glaucoma has negative effects in all glaucoma surgeries.

It should be noted, however, that the percentage of patients which require a pharmacological therapy, even at a lower dosage, to maintain adequate IOP control seems to increase with time.

Even if the results we obtained are very promising, it should be remembered that this was a non-randomized study with a small sample of patients without a control group. Another limitation of our study is that both eyes of two patients have been considered. Even if this could be incorrect from a statistical point of view, considering the small number of patients treated, we decided to describe all cases we treated with this surgical procedure. The study is currently still underway. New patients with corticosteroid-induced glaucoma that fit the inclusion criteria are being added and follow-up data of existing patients are being constantly updated to provide long term results and a larger cohort for our future study. Regarding these patients, multicentric randomized studies with a larger population, where canaloplasty is compared to gold standard surgery (trabeculectomy), are needed to draw more definite and robust conclusions.

### **5. Conclusions**

Canaloplasty is a very promising surgical technique in eyes with high IOP, which is usually the case in patients with corticosteroid-induced glaucoma. In our small cohort of patients, postoperative IOP was able to be maintained within physiological values, even if some medical therapy is occasionally still required. Considering that ocular hypertension is the main risk factor for structural and functional damage in corticosteroid-induced glaucoma, target IOP may not need to be very low to avoid the onset or the progression of the damage. Even if the sample taken into consideration in our study was limited, the good outcomes and the low rate of complications observed with this non-perforating procedure are very encouraging and could entice glaucoma specialists to consider early surgical treatment in the management of this kind of patient.

**Author Contributions:** P.B. and C.T. conceived and designed this research and analyzed data. C.T. visited the patients during the follow-up. P.B. wrote the paper. M.Z. checked patient's data e revised the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Communication* **The Impact of the Brain-Derived Neurotrophic Factor Gene on Trauma and Spatial Processing**

**Jessica K. Miller 1,\*, Siné McDougall 2, Sarah Thomas <sup>3</sup> and Jan Wiener <sup>4</sup>**


### Academic Editor: Nuri B. Farber

Received: 15 September 2017; Accepted: 6 November 2017; Published: 27 November 2017

**Abstract:** The influence of genes and the environment on the development of Post-Traumatic Stress Disorder (PTSD) continues to motivate neuropsychological research, with one consistent focus being the Brain-Derived Neurotrophic Factor (BDNF) gene, given its impact on the integrity of the hippocampal memory system. Research into human navigation also considers the BDNF gene in relation to hippocampal dependent spatial processing. This speculative paper brings together trauma and spatial processing for the first time and presents exploratory research into their interactions with BDNF. We propose that quantifying the impact of BDNF on trauma and spatial processing is critical and may well explain individual differences in clinical trauma treatment outcomes and in navigation performance. Research has already shown that the BDNF gene influences PTSD severity and prevalence as well as navigation behaviour. However, more data are required to demonstrate the precise hippocampal dependent processing mechanisms behind these influences in different populations and environmental conditions. This paper provides insight from recent studies and calls for further research into the relationship between allocentric processing, trauma processing and BDNF. We argue that research into these neural mechanisms could transform PTSD clinical practice and professional support for individuals in trauma-exposing occupations such as emergency response, law enforcement and the military.

**Keywords:** BDNF; Brain-Derived Neurotrophic Factor; navigation; spatial processing; trauma; trauma processing; Post-Traumatic Stress Disorder; PSTD; allocentric; hippocampus

### **1. Introduction**

Post-Traumatic Stress Disorder (PTSD) is an increasingly visible mental health issue that represents a considerable public health burden [1] across many civilian and professional populations. With mounting pressure on health, military and emergency response sectors (https://www.pdtrust. org/help/research/post-traumatic-stress/) to look after the psychological wellbeing of their staff in the face of unprecedented demand from major incidents and resource deficits, understanding PTSD has perhaps never been so critical. Fortuitously, neuropsychological research over recent years has also moved at a commensurate pace and in this paper, we seize the opportunity to reflect on the progress (and pitfalls) of that research. We review recent literature, present findings from exploratory research (provided in more detail in the Appendix A) and highlight design issues which may be key to understanding how genetic and environmental conditions interact to influence PTSD vulnerability, etiology and recovery. To do this, we look at another area of cognitive

function—navigation—which may provide us with vital information about the resilience of a specific part of our brain (the hippocampus) on which we rely to process trauma exposure [2–6].

### **2. The Neural Basis of Post-Traumatic Stress Disorder (PTSD)**

Contemporary theories of PTSD which have been developed from cognitive theories and clinical research [2–9] describe PTSD in the context of information processing. A predominant theory is that of dual representation [2,4,6–8]. Dual Representation Theory describes how trauma processing operates with two types of memory representations in the limbic system: those which are associative and those which are contextual [8]. Associative representations of trauma are typically involuntary, fear-based, and originate in the amygdala. Contextual representations, in contrast, are retrieved voluntarily and mediated by the hippocampus [10,11]. According to Dual Representation Theory, effective trauma processing involves applying context to the sensory and evocative experiences of trauma to consolidate them into long term memory and file them as "past". Trauma literature often refers to egocentric, associatively conditioned responses to stimuli as being typical in cases of post-traumatic stress, and these responses can be described by a signature symptom of PTSD, the "flashback" [9–13]. Hippocampal representations, on the other hand, provide episodic and spatial context for extreme experiences, which enables individuals to make sense of when and where traumatic incidents occurred [14–18]. However, when the hippocampus is down-regulated (e.g., by trauma or stress) it is less able to contextualise or anchor traumatic experiences in space and time, allowing them to intrude in the present, thus prolonging the stress response [4,19–24].

### **3. Brain-Derived Neurotrophic Factor (BDNF)**

Stressful or traumatic incidents in the environment are not the only causes of down-regulation in the hippocampus; genetics also has a substantial impact [1,5,24–30]. Identifying genes which are relevant to the development of PTSD has been a relentless motivator for numerous genome-wide association studies, twin studies and candidate gene studies [1,5]. A recent review [31] identified 25 such studies, many of which highlighted the specific role of a gene called Brain-Derived Neurotrophic Factor (BDNF). BDNF is expressed in the limbic system, moderating fear responses and broadly regulating the stress response [5,20,22,23,30–32]. It is also expressed outside the limbic system, such as in the retina, kidneys and prostate [33], and has been considered integral to critical periods of human development [34]. The BDNF gene codes for the BDNF protein which is then expressed to promote the growth and survival of neurons, particularly those in the hippocampus [26,31,35]. BDNF-related neuroplasticity is considered an important component in maintaining the integrity of the hippocampus [5,25,34,35].

However, this operation is complicated by the fact that the BDNF gene has two variants, derived from carrying "met" and "val" alleles, which differ in their functionality [5,15,18,20,24,25,29,31,32,35]. At a genetic level, allelic variation occurs at codon 66 on chromosome 11, resulting in an amino acid switch from valine (val) to methionine (met) and producing a val66met polymorphism which is unique to humans [18,31,35,36]. In the Caucasian population, 30% carry the met allele, either as the metmet homozygotes or valmet heterozygotes [35]. Typically, met carriers show less activity-dependent release of the BDNF protein in the hippocampus than val homozygotes [5,24,25,29,35]. This means that in met carriers (rather than val homozygotes) sufficient BDNF protein may not be released into the hippocampus for it to respond appropriately to the demands that the environment may place on it, such as the demand for consolidating traumatic experiences into long term memory [20,22,26,30–32,37,38].

Given the compounding effect of the BDNF polymorphism on hippocampal function, we would anticipate that post-traumatic stress would be more prevalent and severe in met carriers, if other environmental conditions have been controlled for. This appears to be borne out by Zhang et al.'s (2014) finding that PTSD was more prevalent and severe in met allele carriers [32]. Specifically, the study revealed that the allelic frequency of BDNF met was twofold higher in those with probable PTSD. In support of this finding, it has recently been proposed that sufficient BDNF release may be involved

in helping to prevent PTSD because its operation induces fear extinction and ensures successful trauma processing [5,20,22,30–32,37,38].

It is worth noting from BDNF and PTSD studies [32] the importance of controlling for environmental conditions. Indeed, a failure to consider the demand on the hippocampus that different environmental conditions can present may account for mixed findings to date in studies relating BDNF to PTSD [1,31,32,36]. Nonetheless, in 2014, Zhang and colleagues [32] successfully controlled for these conditions and reported a direct relationship between the BNDF gene and PTSD in a population of U.S. military Special Operations personnel.

### **4. Hippocampal Function, Navigation and BDNF**

Next, we consider how navigation can be used to assay hippocampal function [10,11,39–41]. Our situational awareness and our ability to orient ourselves and navigate our way through the world rely on two forms of mental representations, those that are hippocampal independent (egocentric representations) and those that are hippocampal dependent (allocentric, see Figure 1).


**Figure 1.** (**a**) Egocentric processing and (**b**) allocentric processing of spatial relationships.

So, given that the hippocampus facilitates (allocentric) spatial processing, individuals' navigation skills depend on effective hippocampal function and can therefore act as an index of hippocampal integrity [5,6,8,10,23,41]. There have been very few studies examining the effects of BDNF on navigation. However, there is evidence in recent neuropsychological literature for a relationship between the BDNF gene and hippocampal dependent (allocentric) spatial processing. A study from 2011 by Banner et al. [29] provides supportive evidence that met-carrying BDNF genotypes rely more on hippocampal independent (egocentric) spatial processing to complete a navigation task than valval homozygotes (see also Lövdén et al., 2011, for a similar proposal based on a study with a much smaller sample) [28]. To demonstrate this, Banner et al. assessed participants' spontaneous strategy use in a virtual maze. A higher proportion of BDNF metmet homozygotes spontaneously used egocentric strategies in comparison to valval homozygotes, whereas a higher proportion of valval homozygotes spontaneously used allocentric strategies. Both studies [28,29] made an explicit connection between less BDNF release in met carriers, lack of hippocampal engagement in spatial tasks and a bias toward implicit, associative spatial processing. In short, BDNF met carriers are generally considered to be more likely to engage in egocentric processing (which does not rely on the hippocampus) in comparison to valval homozygotes (see also [5,15,28,29,40]), who have greater access to effective allocentric processing via the hippocampus.

### **5. Bringing Together Allocentric Spatial Processing, the BDNF Gene and PTSD**

This focused literature review shows that the relationship between BDNF and hippocampal dependent processing and trauma is complicated, as illustrated in Figure 2.

**Figure 2.** The relationship between Brain-Derived Neurotrophic Factor (BDNF) genotype, Post-Traumatic Stress Disorder (PTSD), hippocampal processing, and navigation skills (represented by the anchor). BDNF genotype influences activity-dependent release of the BDNF protein used in hippocampal processing of traumatic and spatial information, potentially placing some genotypes at a disadvantage for trauma resilience and navigation competence.

With regard to trauma processing, severity and prevalence of PTSD is positively related to carrying the BDNF met allele [5,20,30–32,36–38]. With regard to spatial processing, there is evidence of an egocentric bias in BDNF met carriers but no clear differences in allocentric performance between BDNF genotypes were reported in either study [28,29]. Interestingly, egocentric bias in navigation strategy use and allocentric performance deficits have also been recently demonstrated in cases of PTSD (and trauma exposure) [2,3,6,8,40]. Neuropsychology has yet to bring these findings about BDNF, hippocampal dependent processing and trauma (or PTSD) together into one human experiment. A rodent model in 2007 [41] went so far as to demonstrate impaired spatial learning in the Morris Water Maze and significantly reduced extinction of conditioned fear in BDNF "knockout" rats. Although based on deleting rodent genes rather than genotyping human populations, this cross-discipline study stresses the possibility that cognitive spatial processing deficits and impairment in managing trauma exposure may be directly related to BDNF gene expression in the hippocampus.

In 2012, we sought to investigate how spatial processing impairment and trauma exposure processing may be related to BDNF genotypes in an exploratory extension of a human study (*n* = 150) which assessed the impact of PTSD on navigation [6,40]. *Full details of and data from the exploratory study are provided in the Appendix A*. Our intention was to determine if any bias in BDNF met carriers toward hippocampal independent spatial processing:


(c) correlated with subjective measures of self-reported navigation competence [43–45].

In summary, in the diverse sample (*n* = 150) of civilian, police and military populations, PTSD severity and prevalence were similar across BDNF groups. In the sample population, 57 participants had probable levels of PTSD and of those without probable PTSD, 60 were trauma exposed and 33 were not. Participants' navigation performance was assessed using the Alternative Route paradigm [6,10,46]. When graphed (see Figure A1 in the Appendix A), the data showed a distinctly divergent pattern of egocentric performance between BDNF valval homozygotes and met carriers, resulting in significantly higher egocentric performance in met carriers at the end of the navigation task, echoing interpretations of egocentric bias in met carriers in the previous studies [28,29]. Using self-report navigation questionnaires (specific questions from which had been shown to "predict" allocentric spatial processing in earlier studies [43–45]), we were also able to show for the first time that only BDNF valval homozygotes (not met carriers) were accurate in judging their own competence at allocentric spatial processing (see Table A1 in the Appendix A). Overall, our exploratory data were indicative of BDNF-related differences in hippocampal dependent and independent navigation behaviour, irrespective of PTSD. While interesting, it is important to note that these findings were limited by several design features. These limitations provide valuable insights for further research, and it is to those insights that we now turn.

### **6. The Future of BDNF Research**

For candidate gene (BDNF) research to shape the future of clinical trauma interventions or to influence professional practices in occupations requiring situational awareness, studies need to be able to deliver accurate and ecologically relevant data [1,3,47,48]. Our exploratory research [40] into the relationship between BDNF, PTSD and spatial processing was limited by several factors. If these factors could be addressed in replication studies, significant progress in our understanding of *gene* × *environment* interactions in trauma and navigation could be imminent. Here, we briefly critique the design limitations of recent studies (including our own) and offer suggestions for improving data quality in key areas: experimental groups, performance measurement, subjective measures of navigation, and collection of further neurological data.

Sample populations for BDNF studies can be a contentious issue, with some traditional academic disciplines [30] typically favouring large cohorts (of thousands) and genome-wide association studies over the much smaller designs and sample sizes seen in candidate gene studies [1,36,42,47,48]. While some candidate gene study sample sizes have simply been too small (*n* < 20) to adequately represent the three BDNF genotypes (see [1,28,46]), other moderate samples (*n >* 100) (see [24,26,29,32]) have been able to demonstrate the influence of the gene on PTSD when other environmental conditions within and between experimental groups have been adequately controlled. This was a lesson learnt by Zhang et al. in 2014 [32], in their replication of an earlier study from 2006 [42] which did not control for trauma exposure type or severity, time since exposure or treatment status. Another important factor to control for in studies of hippocampal function is age [2,5,6,10,25,46]. A primary recommendation is for future BDNF and PTSD research to control for: age, time since exposure, treatment status and trauma exposure type or severity (at least distinguishing between civilian exposure and occupational exposure, such as the military or blue light services).

Navigation performance as a measure or index of hippocampal integrity is also key to BDNF research design. Identifying navigation tasks which produce data that can discriminate between hippocampal dependent and independent performance is challenging [6,8,10,11,40,41,46], yet vital. In our own study, the purity of the egocentric performance measure [6,40,46] was somewhat compromised by the fact that egocentric trials in the route learning task could feasibly be solved allocentrically. Some theorists may challenge this concern on the basis that egocentric processing is generally considered more parsimonious and therefore more likely a universal default means of solving simple tasks [49,50]. Nonetheless, implementing spatial processing measures which can accurately distinguish between allocentric and egocentric processing should remain a priority for any future

studies which intend to compare functionality of the two memory systems. Similarly, disparity between studies which assess performance in allocentric and egocentric strategy use as opposed to allocentric or egocentric spontaneous strategy choice, is also something to be mindful of when comparing participants' navigation behaviours [6,28,46,51]. Spontaneous navigation behaviour and navigation behaviour over the course of a learning paradigm likely measure different components of spatial processing and need to be clarified as such in research design. Asking participants directly how they think their behaviour may have changed over the course of a navigation task is a common approach and is it our recommendation that developing post-test self-report (i.e., think aloud) [28,29,43,46] measures may provide useful insights into how individuals think they navigate, even if this contradicts with their performance data.

Understanding individuals' self-awareness of their ability to apply hippocampal dependent processing when required is not only valuable to research into the declarative hippocampal dependent memory system [26,28,29] but could be highly valuable for clinical trauma processing and professional navigation training interventions that rely on that form of information processing [2,3,6,7,9,13,23,39,40]. Developing more ecologically relevant subjective measures of spatial processing would benefit future BDNF studies greatly. The navigation questionnaire literature shows that the Santa Barbara Sense of Direction questionnaire [43] and selected questions from the Questionnaire of Spatial Representation [44] and the Fragebogen Räumliche Strategien [45] can predict navigation performance, yet the validity of the questions could be enhanced by introducing terminology and frames of reference more aligned with the types of spatial processing that sample populations may be familiar with on a day-to-day basis. For example, a subjective navigation measure for a military population could refer to topographical (landscape) changes from conflict as a frame of reference for certain questions. Whereas for policing populations, references to using satellite navigation while driving a response vehicle may be a more meaningful context. Using terms participants can relate to, they may increase task engagement, and their more focused self-reflection could enhance the validity of the self-report measures.

Finally, we consider the use of supplementary neurological data to support future research into the influence of BDNF on hippocampal processing. While research to date has clearly suggested a relationship between BDNF, hippocampal dependent processing, PTSD and navigation, the precise neural mechanisms underlying this relationship are as yet undefined. BDNF research has looked at volumetric measurement in the hippocampus using magnetic resonance imaging (MRI) and fluorescent microscopy [51–53], but whether volume differences are the result of BDNF-related protein release, neurogenesis, neuronal survival or synaptic plasticity is not clear [50–54]. Investigating these neural mechanisms is further complicated by the possibility that BDNF is released in response to different environmental conditions over the life span, meaning that age (or critical periods of development) and time since trauma exposure may need to be controlled for when investigating levels of BDNF in plasma, blood or saliva (as opposed to BDNF genotypes) [5,25,33,34,36,54–58].

Implementing a solid framework for future research into the relationship between hippocampal integrity, PSTD and spatial processing is likely a daunting, but we argue, necessary task. Such a framework could comprise combining:


### **7. Conclusions**

Understanding the *gene* × *environment* interaction in relation to both trauma exposure and spatial processing has far-reaching practical, clinical and academic implications. In practical terms, if research could accurately quantify the contribution that carrying the BDNF met allele makes to an individual being able to successfully adopt hippocampal dependent information processing techniques, this could transform how trauma management and navigation training interventions are delivered in clinical and occupational settings. If 30% of the Caucasian (and up to 50% of non-Caucasian) populations [31,35,36,60] (i.e., BDNF met carriers) could access interventions that either deliberately encouraged hippocampal dependent processing or provided workable alternatives to hippocampal dependent processing, this could equate to a substantial improvement in those intervention outcomes. The contribution that research on the hippocampus has made to 21st century neuropsychology (epitomized by the awarding of the Nobel Prize for Physiology or Medicine to Professor John O'Keefe in 2014 https://www.nobelprize.org/nobel\_prizes/medicine/laureates/2014/okeefe-facts.html) is well recognised. Further research into the BDNF gene would reinforce the value of understanding how the human hippocampus shapes our emotional and professional lives. Perhaps above all, we believe that developing this research will enable science and society to take an important step toward protecting the wellbeing and mental integrity of the hundreds of thousands of men and women who put themselves in the face of trauma as part of their everyday public service, in defence, emergency response and law enforcement. To do this, we need to embrace lessons from earlier (sometimes exploratory) research across the disciplines of genetics, trauma and navigation to ensure that as we move forward, we can offer neuroscience the caliber of data it needs to meet some pressing public issues head on.

**Acknowledgments:** This work was financially supported by Army of Angels (registered charity 1143612). Recruitment to the research was supported by Combat Stress (registered charity 26002), Dorset Constabulary and Dorset University Healthcare Foundation Trust. Ethics clearance was secured with the support of Simon Wessely (King's Centre for Military Health Research) and Chris Brewin (University College London) who also kindly provided clinical supervision. Thanks goes to Olivier de Condappa for his technical support and to Kirsten Smith for her help with recruitment.

**Author Contributions:** Jessica K. Miller and Jan Wiener conceived and designed the experiments; Jessica K. Miller performed the experiments; Jessica K. Miller analysed the data; Sine McDougall and Sarah Thomas contributed analysis and advice; Jessica K. Miller wrote the paper, edited by Sine McDougall, Jan Wiener and Sarah Thomas.

**Conflicts of Interest:** The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **Appendix A. Exploratory Research Data**

The appendix contains details and data from the exploratory research into the Brain-Derived Neurotrophic Factor (BDNF) gene, undertaken as part of a study into the impact of trauma and Post-Traumatic Stress Disorder (PTSD) on navigation behaviour [5,6,40,46].

### *Appendix A.1. Participants*

The study sample population (*n* = 150) was diverse and was recruited from a university volunteer scheme (*n* = 83), two psychotherapy treatment centres (Traumatic Stress Clinic, Camden and Islington NHS) and the Intensive Psychotherapy Treatment Centre (Dorset Healthcare University NHS Foundation Trust) (*n* = 10), two police forces (*n* = 26) (Dorset and Cambridgeshire constabularies), and a military veteran treatment centre for PTSD (Combat Stress, Tyrwhitt House, Leatherhead, Surrey: registered charity #2060002) (*n* = 25 plus *n* = 6 staff). PTSD symptom severity and prevalence rates were measured using the PTSD Diagnostic Scale (PDS) [61]. Saliva samples for BDNF genotyping were collected using DNA *Genotek Orangene*™ self-test kits (DNA Genotek, Ottawa, ON, Canada). Participants were grouped according to whether they had been exposed to trauma or not, and if they had, whether or not they had clinical or probable levels of PTSD, using the PDS [46,61]. Sample sizes

for PTSD were modest for the purpose of analysis by BDNF genotype. Of the 150 participants, 57 had probable levels of PTSD and of those who did not, 60 were trauma exposed and 33 were not trauma exposed. The frequency of the genotypes did not differ between experimental groups (*Trauma Unexposed, Trauma Exposed No PTSD*, and *PTSD*) for either the valval group, χ<sup>2</sup> = 0.24, *p* = 0.88 or the "met carrying" group, χ<sup>2</sup> = 0.04, *p* = 0.98. An independent samples *t*-test revealed no differences in PDS scores between BDNF groups, *t* (54) = −1.23, *p* = 0.23. This enabled the analysis of the influence of the BDNF gene to control for PTSD. The findings supported Zhang et al.'s conclusion in 2014 [32] that in order to demonstrate BDNF-related group differences in PTSD prevalence or severity, other environmental conditions may need to be controlled for (such as time since and severity of exposure, previous trauma and treatment access). It is important to note that this study could not assess the relationship between BDNF, spatial processing and PTSD, rather the relationship between BDNF and spatial processing.

### *Appendix A.2. Methods*

DNA saliva samples were genotyped for one single nucleotide polymorphisms (SNP): rs6265 (Val66Met), using a Taqman® allele discrimination assay. The genotyping information acquired described which BDNF alleles were carried by the (anonymous) participant, i.e., whether they carried only the val allele (and were therefore valval homozygotes), both the val allele and the met allele (and were therefore valmet heterozygotes) or if they carried only the met allele (and were metmet homozygotes).

Allocentric and egocentric performance were assessed using a navigation paradigm, the Alternative Route (AR, [46]) which required participants to learn a route through a virtual environment. The task involved testing participants on their ability to re-join a route from the same direction and different directions from which they learned it. There were six blocks to the task. Mean performance on trials which involved joining the route from the same direction (same direction trials) provided a measure of egocentric performance. Mean performance on trials which involved joining the route from a different direction to the route as it was learned (different direction trials) provided a measure of allocentric performance. An ANOVA was used to assess BDNF group differences (met carriers vs. valval homozygotes) in egocentric and allocentric performance on a navigation task. Planned contrasts were then made at individual block level (1–11) using *t*-tests (see [6,40,46]).

Participants' self-reported navigation competence was assessed using validated questionnaires; the *Santa Barbara Sense of Direction* (SBSOD) [43] and two items from the *Questionnaire of Spatial Representation* (QSR) [44] that target allocentric processing. Correlation analysis was used to analyse the relationship between self-reported competence and navigation performance within the BDNF groups (met carriers and valval homozygotes).

### *Appendix A.3. Results*

Appendix A.3.1. BDNF and Hippocampal Independent Performance

A repeated measures 2 × 6 ANOVA with the dependent variable egocentric performance, the between factor BDNF group (met carriers vs. valval homozygotes) and the within factor block (1 to 6) revealed no significant main effect of block, *F* (4.49, 138) = 0.90, *p* = 0.48, η*p*<sup>2</sup> < 0.01, nor BDNF group, *<sup>F</sup>* (1, 138) = 0.99, *<sup>p</sup>* = 0.32, <sup>η</sup>*p*<sup>2</sup> < 0.01, but a significant group × block interaction, *<sup>F</sup>* (4.49, 138) = 2.48, *p* = 0.03, η*p*<sup>2</sup> = 0.02, visible in the divergent pattern in performance in Figure A1.

**Figure A1.** Mean egocentric performance on the Alternative Route (AR) paradigm between BDNF genotypes (*n* = 140, with *n* = 96 valval homozygotes and *n* = 44 met carriers) with standard error bars, showing significant performance differences in block 6, \* *p* < 0.05.

Planned contrasts using independent samples *t*-tests revealed met carriers performed better than valval homozygotes in the final block 6 (89% SD ± 17% vs. 78% SD ± 28%), *t* (138) = 5.65, *p* = 0.006 (equal variances not assumed). Applying Bonferroni's correction (0.05/6 blocks) would require a *p* value of 0.008.

Despite the lack of significance in the overall BDNF group and egocentric performance interaction, closer examination of the data was undertaken to understand these final performance differences in block 6 and to look for any plausible explanation (other than chance) for the inverted curves in egocentric performance between the BDNF genotypes. The data showed there to be a statistically significant quadratic (rather than linear) effect of BDNF group, *F* (1, 138) = 5.59, *p* = 0.02 which supports earlier hypotheses for met carrier status explaining differences in egocentric navigation strategy preference over the course of the task [28,29,62].

Allocentric performance was comparable between BDNF genotypes, as found by both Banner et al. and Lövdén et al. in 2011 [28,29] and as illustrated in Figure A2. A repeated measures 2 × 6 ANOVA with the dependent variable (DV) of allocentric performance, the between factor BDNF group and within factor block (1 to 6) revealed a significant main effect of block (with performance increasing by block), *F* (4.07, 138) = 27.2, *p* < 0.01, η*p*<sup>2</sup> = 0.17, but no significant main effect of BDNF group, *F* (1, 138) = 2.20, *p* = 0.14, η*p*<sup>2</sup> = 0.02, and no significant interaction *F* (4.07, 138) = 1.71, *p* = 0.14, η*p*<sup>2</sup> = 0.01.

**Figure A2.** Mean allocentric performance on the AR paradigm between BDNF genotypes (*n* = 140, with *n* = 96 valval homozygotes and *n* = 44 met carriers) with standard error bars.

### Appendix A.3.2. BDNF and Self-Reported Navigation Competence

The final part of the analysis was to assess the navigation questionnaire data in relation to BDNF genotype. Participants' scores for general self-reported competence in navigation (SBSOD total score) and the allocentric targeted QSR questionnaire items were positively correlated with their performance in allocentric navigation in the AR task. As illustrated in Table A1, only valval homozygotes' self-reported competence positively correlated with their performance in allocentric navigation, suggesting that met carriers were less able than valval homozygotes to accurately describe their capacity for hippocampal dependent (allocentric) spatial processing.

**Table A1.** Pearson's correlations (r) between Questionnaire of Spatial Representation (QSR) allocentric items, the Santa Barbara Sense of Direction (SBSOD) and allocentric performance in the AR paradigm in BDNF valval homozygotes (*n* = 102) and met carriers (*n* = 45), *p* < 0.01 \*\*, *p* < 0.05 \*.


### **References**


© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **A Review of HPV-Related Head and Neck Cancer**

### **Kazuhiro Kobayashi 1, Kenji Hisamatsu 1, Natsuko Suzui 1, Akira Hara 1,2, Hiroyuki Tomita 2,\* and Tatsuhiko Miyazaki 1,\***


Received: 27 July 2018; Accepted: 22 August 2018; Published: 27 August 2018

**Abstract:** Head and neck squamous cell carcinomas (HNSCCs) arise in the mucosal lining of the upper aerodigestive tract. Tobacco and alcohol use have been reported to be associated with HNSCC. Infection with high-risk human papillomaviruses (HPVs) has recently been implicated in the pathogenesis of HNSCCs. It is now widely accepted that high-risk HPV is a cause of almost all cervical cancers as well as some forms of HNSCCs. HPV-related HNSCCs are increasing. HPV-related HNSCCs and HPV-unrelated HNSCCs differ with respect to the molecular mechanisms underlying their oncogenic processes. HPV-related HNSCCs are known to have a better prognosis response to treatment as compared with HPV-unrelated HNSCCs. Therefore, in recent years, it has been required to accurately discriminate between HPV-related and HPV-unrelated HNSCCs. To diagnose the HPV-related HNSCCs, various methods including *P16* immunohistochemistry, FISH, and genetic analyses of the HPV gene from histopathological and liquid biopsy specimens have been employed. Based on the results of the differential diagnosis, various treatments employing EGFR TKI and low-dose radiation have been employed. Here, we review the involvement of the HPV virus in HNSCCs as well as the molecular mechanism of carcinogenesis, classification, prognosis, diagnostic procedures, and therapy of the disease.

**Keywords:** human papillomavirus; human cancer; head and neck; reduction therapy

### **1. Introduction**

The role of human papillomavirus (HPV) in carcinogenicity was confirmed in 1983 following the cloning of HPV 16 type from cervical carcinoma tissue by Durst and colleagues [1]. It has since become widely accepted that high-risk HPV is a cause of almost all cervical cancers. Many cases of HPV infection are asymptomatic and resolve spontaneously, but cervical cancer may arise in cases of persistent HPV infection of the cervical basal cells [1,2]. As reviewed by Kreimer et al. [3], HPV DNA has been detected by polymerase chain reaction (PCR) in head and neck squamous cell carcinoma arising from various anatomic sites. HPV16 is the predominant HPV type, accounting for 90% of HPV DNA-positive HNSCCs. Various studies involving mainly HPV 16 have shown that viral DNA is diffusely present in tumor cells of whole tumor, and exhibits clonality when detected by in situ hybridization (ISH) [4–6]. As shown in several oral and oropharyngeal carcinoma cell lines [7–9], the retention of viral DNA during the growth of tumor cells in culture provides further evidence suggestive of viral clonality.

Various studies have reported that oral and tonsillar epithelial cells can be immortalized by full-length HPV 16 or its E6/E7 oncogenes [10–14]. Additionally, transgenic mouse models have revealed that HPV 16 E6/E7 strongly increases susceptibility to oral and oropharyngeal carcinomas [15]. Although E7 was much more competent in inducing these tumors [15], a clear synergy between E6 and E7 in causing HNSCC was discovered [16]. In an analysis of paraffin-embedded biopsies of 116 cases of laryngeal squamous cell carcinomas by in situ DNA hybridization using 35S-labelled HPV (types 6, 11, 16 and 30) DNA probes, 15/116 (12.9%) tumors were shown to contain the DNA of at least 1 HPV type. HPV 11 was the most frequent DNA type, found in 9/116 (7.8%) of the lesions; HPV 16 was found in 5.2%, and HPV 6, in 4.3% [17]. HPV was recognized as a risk factor for oropharyngeal carcinogenesis by the International Agency for Research on Cancer (IARC) in 2007 [18].

A higher frequency of oral sex and a greater number of sex partners are thought to increase the risk of HPV-related cancer in the oropharynx [19,20].

The development of a vaccine for the primary prevention of HPV infection subsequently became an urgent worldwide priority. In 2017, Muranaka was awarded the John Maddox Prize for raising public awareness of the efficacy of the HPV vaccine for the prevention of cervical and other cancers (http://senseaboutscience.org/activities/2017-john-maddox-prize/). This review summarizes the involvement of HPV virus, molecular pathological implications, classification and prognosis, and prospects for future treatment in head and neck cancers.

### **2. Involvement of HPV Virus in Head and Neck Cancers**

HPV is a DNA virus that infects the skin and mucous membranes. More than 100 types of HPV have been classified to date. Previous studies have examined the role of HPV-related carcinogenesis in uterine cervical cancer. HPV infecting the uterine cervix is divided into high- and low-risk groups. HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 68, 69, and 73 are classified as high-risk HPV [21], which is estimated to account for almost 100% of cases of cervical cancer, about 90% of cases of anal cancer, and 40% of vulva, vagina, and penile cancer. Additionally, at least 12% of pharyngeal cancer, 3% of oral cancer, and 30–60% of oropharyngeal carcinoma cases are caused by HPV infection [22]. An increase in squamous cell carcinoma of the head and neck has been reported and attracted global attention recently in the world [23]. It is now recognized that there are two types of squamous cell carcinoma of the head and neck. Classically, most oropharynx cancers are hyperdifferentiated and often show keratinization. Squamous cell carcinoma of the head and neck can be of the keratinized or nonkeratinized type. The former occurs most often in elderly males and is associated with smoking and alcohol consumption, but HPV is not involved. Conversely, nonkeratinizing squamous cell carcinoma occurs most commonly at age 40–55 years in men with little exposure to tobacco and alcohol, and HPV DNA is detected as the most characteristic feature [24–27]. Since the smoking rate in the USA is declining [28], the incidence of HPV-negative tobacco-related oropharyngeal cancer has decreased; however, that of HPV-positive oropharyngeal cancer is increasing [29]. According to repository data from the Surveillance, Epidemiology, and End Results (SEER) program, the prevalence of HPV-negative cancers decreased by 50% from 1988 to 2004, while HPV-positive oropharyngeal carcinoma increased by 225% [30]. In a study by Junor and associates involving patients at the Edinburgh Cancer Center, 41% of head and neck cancers were HPV-positive between 1999 and 2001 and 63% were HPV-positive between 2003 and 2005 [31]. HPV-positive oropharyngeal cancer is considered to be a separate disease with a causal relationship to HPV infection and a good prognosis. Several studies have shown that patients with HPV-positive oropharyngeal cancer, identified through PCR, in situ hybridization or *P16* immunohistochemistry on tumour tissues, have a significantly improved overall and disease-free survival compared with patients with HPV-negative oropharyngeal cancer patients [32–40]. In a prospective study involving 253 newly diagnosed or recurrent HNSCC patients, HPV was detected in 25% of patients. A low tumor grade and the oropharyngeal site increased the likelihood of the presence of HPV, respectively [5]. Oropharyngeal tumors are more likely to be positive for HPV (57%) compared with sites other than the tumor site and the oropharynx (14%) and the oral cavity (12%). HPV-positive oropharyngeal carcinoma occurs primarily in the tonsillar of the palatine or tonsils of the tongue. In the tonsils or tongue base, 62% of the tumors were HPV-positive, whereas in other parts of the oropharynx 25% were HPV-positive.

### **3. Pathological Molecular Mechanism in Carcinogenesis**

HPV-unrelated HNSCC cigarette smoking and alcohol has p53 mutations [41]. Deletion of 9p21–22 is also observed early in the oncogenic process, and as a result, the function of the tumor suppressor gene *P16* is lost [42]. *P16*INK4a produced by the *P16* gene forms a complex with cyclin 1-cyclin-dependent kinase 4/cyclin-dependent kinase 6 (CDK4/CDK6), inhibits phosphorylation of Rb, and inhibits transcription factor E2F-related cell rotation (pRB pathway) [43]. In HPV-associated head and neck cancer, wild-type p53 is present and mutations occur at a rate of only 10% or less. However, HPVE6 inactivates p53 resulting in a decrease in function. Furthermore, there is no deletion of *P16* in these tumors. Since HPVE7 inactivates phosphorylated Rb, which controls cell cycling of host cells, control of E2F is inhibited [44] and *P16* is overexpressed. *P16* is a tumor suppressor gene encoding a CDK repressor, which inhibits the complex formation of cyclin D1 and cyclin-dependent kinase (CDK) 4/6. Cyclin D1 and CDK 4/6 complex promotes cell cycling through the release of E2F via phosphorylation of the Rb protein, whereas the Rb protein/E2F complex also suppress the transcription of *P16*, so that when HPV-E7 inactivates the Rb protein, *P16* is overexpressed (Figure 1) [45].

**Figure 1.** Signaling pathways of high-risk HPV oncogenes. High-risk HPVs encodes two known viral oncogenes. E6 protein inactivates tumor suppressor p53 mediated DNA damage and apoptosis pathway. E7 protein inactivates tumor suppressor pRb mediated cell cycle regulation pathway.

Thus, the phenotype at the molecular level is completely different between HPV-positive and HPV-negative cancer of the head and neck. Various methods have therefore been employed for the detection of HPV in head and neck cancer such as consensus primer or type-specific PCR, real-time PCR, in situ hybridization, and serum antibody assays. For cervical cancer screening, accepted international guidelines recommend using hybrid capture II (QIAGEN) and PCR (GP 5/GP 6).

*P16* immunohistochemistry is also useful as a surrogate maker for HPV infection detection, especially in head and neck cancers. *P16* immunohistochemistry has 100% sensitivity and 79% specificity; it is the gold standard for HPV detection, based on HPV 16 E6 and E7 mRNA, in head and neck cancer specimens and is useful as a surrogate maker for clinical HPV detection. *P16* is also a useful molecular marker for judging prognosis and is a component of the WHO classification scheme, described below. *P16*-positive and *P16*-negative HNSCCs could clearly be distinguished in our specimens (Figure 2).

**Figure 2.** (**A**–**D**) A case of *P16* positive squamous cell carcinoma of Oropharynx. (**A**) HE, (**B**) *P16*, (**C**) p53, (**D**) MIB-1. (**E**–**H**) A case of *P16* positive squamous cell carcinoma of Oropharynx. (**E**) HE, (**F**) *P16*, (**G**) p53, (**H**) MIB-1.

A general consensus has been achieved for the definition of HPV associated tumors that require expression of the virus oncogenic proteins E6 and E7 that are involved in neoplastic transformation of infected cells. However, the integration of HPV-DNA into the host cell genome confirms the belief that it is an essential step for viral oncogene expression in oropharyngeal cancer, as in the case of cervical cancer. Regardless of the process leading to oncogene expression, HPV E6/E7 mRNA identification based on DNA or protein expression, for patient stratification and epidemiologic purposes, is considered a gold standard for HPV-related classification [46,47]. More accessible strategies are generally accepted. In the examination of the pathological specimen, the detection of HPV-DNA using PCR and ISH examination are typically used together with immunohistochemistry of *P16*. Various new methods have been examined for HPV testing in head and neck cancers [48]. When infected cells become malignant, HPV DNA remains in the nucleus, and viral oncoproteins, in particular against E6, are detected in virtually all cases of HPV-driven OPC cases [49,50]. There are reports that HPV-DNA and HPV16 E6 antibodies in oral and in body fluids can be used for detection of HPV-infected head and neck cancers and prediction of the risk of recurrence [51].

Studies using next-generation sequencing of HNSCCs have also been reported in recent years. Matthias Lechner did 20 HPV+ and 20 HPV-laser capture microdissection pharyngeal carcinoma studies. HPV-positive and HPV-negative oropharyngeal cancers are divided into two different subgroups. A TP53 mutation was detected in 100% of HPV-negative cases and invalidation of G1/S checkpoint by CDKN2A/B deficiency and/or CCND1 amplification was shown to occur in the majority of HPV tumors [52]. In a study that examined the somatic mutations of 279 HNSCCs performed in 2015, In HPV-positive oropharyngeal carcinoma, deletion of TRAF3, activation mutation of PIK3CA, and amplification of E2F1 were observed. In HPV-negative HNSCC, subsets recognizing simultaneous mutations of CASP8 with amplicons on 11q with CCND1, FADD, BIRC2, and YAP1, or with HRAS, were observed. Either type of tumor was abnormal with respect to target cell cycle, death, NF-κB and

other oncogenic pathways [53]. In a study of 92 cases of HNSCC using the next generation sequencer, TPK 53was the most common mutation, occurring in 47 (51%) patients followed by CDKN 2A (*n* = 23, 25%), CCND 1 (*n* = 22, 24%), and PIK 3 CA (*n* = 19, 21%). Changes in TPV, CDKN2A, and CCND1 genes occurred more frequently in HPV-negative tumors, but the total amount of mutations was similar between HPV-negative and HPV-positive tumors. HPV-positive tumors were significantly associated with immune-related genes compared to HPV-negative tumors. Mutations in NOTCH1 (*p* = 0.027), CDKN2A (*p* < 0.001), and TP53 (*p* = 0.038) were significantly associated with decreased overall survival. FAT1 mutation was highly enriched in cisplatin responders and targetable alterations such as PIK3CA E545K and CDKN2A R58X were found in 14 patients (15%) [54].

As an emerging technology, liquid biopsies, involving the use of a small amount of DNA and mRNA collected from blood samples, have been used in recent cancer studies. Allen et al. [55] reported a novel in vitro diagnostic approach in which miRNA is examined from exposed cancer cells using sera from HNSCC patients. Of 377 miRNAs detected, 16 different miRNAs were found to be differentially expressed when comparing cells exposed to serum from HNSCC versus healthy individuals. Real-time PCR analysis revealed that serum from HNSCC patients downregulated the expression of 5 genes involved in carcinogenesis and 2 of these—P53 and SLC2A1—are direct targets for the detected miRNAs. This technique has potential for a new therapeutic approach using tumor-specific cell lines or single cell in vitro assays, and the possibility of more specific diagnosis could facilitate individual treatment and early detection of primary tumors or metastasis.

### **4. Classification and Prognosis of Head and Neck Cancers**

In general, HPV-positive oropharyngeal carcinoma is highly susceptible to radiation and anticancer drugs and has a better prognosis compared with HPV-negative cancer. HPV-negative oropharyngeal carcinoma is caused by the disappearance of function due to p53 gene mutation [56–59]. In a prospective phase II clinical trial of pharyngeal and laryngeal cancer patients by the Eastern Cooperative Oncology Group, 63% of 60 cases were HPV 16 positive, and all of the HPV16-positive cases were also positive for *P16*. For HPV-positive and HPV-negative cancers the respective response rates were 84% versus 57%, the 2-year progression-free survival rates were 86% vs. 53%, the overall 2-year survival rates were 95% vs. 62%, and the prognosis for HPV-positive cancer was significantly better [37]. Previous reports have examined the prognosis of HPV in relation to *P16*, and the outcomes of oropharyngeal cancer in relation to tobacco exposure. In a phase III study involving 400 oropharyngeal carcinomas, the 2-year progression-free survival rates for HPV-positive and HPV-negative cancer were 72% vs. 51% and the 2-year overall survival rates were 88% vs. 67%. Due to differences in prognosis, squamous cell carcinoma of the head and neck was classified in the new WHO scheme as HPV-negative or HPV-positive. Classification by *P16* immunostaining or HPV testing is recommended. In addition, some recently proposed classification schemes are based on the EGFR status according to 2 categories: HPV-positive/*P16* positive squamous cell carcinoma and HPV-negative/*P16* negative squamous cell carcinoma [60]. Cetuximab, a monoclonal antibody that inhibits the function of EGFR, is known to have efficacy in colorectal cancer [61,62] and head and neck cancer [63,64], and it was also shown to be more specific and cost-effective for these types of cancers.

### **5. Prospects for Treatment**

For the reasons outlined herein, HPV virus is expected to be a therapeutic target in the treatment of human cancer [65] We believe that the prognosis for cervical cancer can be greatly improved through the implementation of methods for early detection such as cytodiagnosis, HPV screening, and *P16* immunostaining, Furthermore, HPV vaccination may also be useful for preventing HPV-related cancers other than cervical cancer. For example, most cases of oropharyngeal carcinoma are caused by HPV 16 (about 90%) and HPV 18, and HPV vaccination for this condition can be expected to have a greater disease-suppressing effect than in cervical cancer. Randomized trials have provided strong evidence for high efficacy of the 2 FDA-approved VLP vaccines: the bivalent HPV16/18 vaccine (CervarixÒ, GlaxoSmithKline Biologicals, GlaxoSmithKline plc, Brentford, UK) and the quadrivalent HPV 6/11/16/18 vaccine (Gardasil™, Merck Sharp and Dohme, Merck & Co., Kenilworth, NJ, USA) against cervical, vaginal, and vulvar HPV16/18 infections and related diseases, and against anal HPV16/18 infections in women [66–70]. One of the methods proposed for prevention of HPV-related oropharyngeal cancer is vaccination. The US Centers for Disease Control Advisory Committee on Immunization Practices (ACIP) has recommended HPV vaccination for females and males between the ages of 11 and 12 years, starting as early as 9 years, with booster doses at up to 26 and 21 years for females and males, respectively. In order for the vaccination to be effective in preventing oropharyngeal cancer, the protective effect must last for at least 2 decades, and ongoing studies have shown no waning of systemic antibodies at 8 years after vaccination. However, in June 2013, the Ministry of Health, Labour and Welfare of Japan suspended proactive recommendation of the vaccine after unconfirmed reports of adverse events. To investigate any potential association between the vaccine and reported symptoms, the Nagoya City Council conducted a questionnaire-based survey. The anonymous postal questionnaire investigated the onset of 24 symptoms, associated hospital visits, frequency, and influence on school attendance. A total of 29,846 residents responded. No significant increase in occurrence of any of the 24 reported post-HPV vaccination symptoms was found. The results suggest no causal association between the HPV vaccines and reported symptoms [71].

Also, Reduction surgery or minimally invasive treatment should be considered in cases of HPV virus-related oropharyngeal carcinoma. Although limited to the T1 and T2 stages of oropharynx cancer, transoral robotic surgery approved by the FDA since 2009. In addition, many reduction surgeries and post-operative adjuvant therapies based on pathologic staging are being studied [72–74].

As mentioned above, it is known that EGFR is expressed in HNSCCs, and combination use of cetuximab and radiation is, therefore, being studied for treatment instead of standard cisplatin therapy [38,61,64,75–78]. Several researchers have hypothesized that radiation dose reduction is feasible and safe for some HPV-positive patients when induction chemotherapy (IC) is used for patient selection. Three points have been raised to support this assertion. First, HPV-positive HNSCCs are considered to be more radiosensitive than HPV-negative HNSCCs [79]. Second, doses comparable to the supplemental radiation dosage is sufficient for the treatment of patients with asymptomatic disease [80]. Finally, the response to chemotherapy can predict the future response to subsequent radiation therapy. Chen et al. [79] conducted an arm Phase II trial (NCT 01716195) in 44 patients with stage III/IV *P16*-positive OPCs treated 2 cycles of IC (paclitaxel and carboplatin was for 21 days), followed by radiation and paclitaxel. The radiation dose was also reduced in complete or partial responders (54 Gy, *n* = 24), and even in patients with partial or no response (*n* = 20, 60 Gy instead of standard 70 Gy). Except for 1patient, all patients completed the IC. Two years PFS and local region control were 92% (95% CI 77–97) and 95% (95% CI 80–99), respectively. The two-year degree of freedom for grade 3 or adverse events of worsening mucosa and esophagus was 85% (95% CI 80–90) in patients treated with 54 Gy. These results were achieved with doses reduced by 15–20% compared with these in the ECOG 2399 trial using the same protocol, except that a dose of 70 Gy, dose was used as a standard chemoradiotherapy regimen. Besides that, Removal of chemotherapy and Alternative to the "conventional" photon beam therapy are considered as new treatment methods [80].

Differences in the prognosis and etiologic mechanisms of HPV-related head and neck cancer from conventional head and neck cancers (mostly HPV-negative) suggest that the detection of HPV may significantly change the future diagnosis, treatment, and management [31]. HPV not only plays a role in the development of pharyngeal cancer but is also involved in 23.5% of oral cancer and 24% of laryngeal cancer cases, suggesting that indications for HPV vaccination could be expanded to also include oral and laryngeal cancer [37].

**Author Contributions:** K.K., K.H., N.S., A.H., H.T. and T.M. all contributed to the manuscript and have reviewed it.

**Funding:** This research was fund ed by the KAKENHI grant from the Ministry of Education, Culture, Sports, Science, and Technology of Japan: grant No. 26430111 (HT).

**Acknowledgments:** We thank all the members who work in the Division of Pathology.

**Conflicts of Interest:** The authors declare no conflicts of interest in this work.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **The Role of Procalcitonin in the Diagnosis of Meningitis: A Literature Review**

### **Dimitrios Velissaris 1, Martina Pintea 2, Nikolaos Pantzaris 3, Eirini Spatha 3, Vassilios Karamouzos 4, Charalampos Pierrakos <sup>5</sup> and Menelaos Karanikolas 6,\***


Received: 4 June 2018; Accepted: 8 June 2018; Published: 11 June 2018

**Abstract:** Objective: To review the current published literature on the use of procalcitonin as a diagnostic and prognostic marker in adult patients with meningitis. Methods: We conducted a PubMed search to identify all relevant publications regarding the diagnostic and prognostic value of serum procalcitonin in patients with a known or suspected central nervous system infection. We also reviewed the bibliographies of all identified manuscripts in an attempt to identify additional relevant references. Results: A significant body of evidence suggests that serum procalcitonin has a promising role and can be a useful biomarker in the assessment of patients with meningitis. Conclusions: Our literature review suggests that data on the role of Cerebrospinal Fluid (CSF) procalcitonin are limited, whereas serum procalcitonin (S–PCT) is probably a useful tool in the evaluation of patients with a known or suspected central nervous system infection and can help distinguish between bacterial and viral meningitis.

**Keywords:** procalcitonin; bacterial meningitis; viral meningitis; antibiotic therapy; biomarker; differential diagnosis

### **1. Introduction**

Meningitis is a serious medical condition and can be a major cause of morbidity and mortality. Early diagnosis and timely initiation of appropriate antibiotic therapy is crucial for reducing mortality from bacterial meningitis.

The term "biomarker", as used in daily clinical practice, refers to molecules and biological products used as markers for the assessment of disease progression or as indicators for the presence of an abnormal clinical state. Biomarkers can be specific cells or genes, gene products, enzymes or hormones, have characteristic defined biological properties, and can be detected and measured in biological fluids (plasma, serum, cerebrospinal fluid, bronchoalveolar lavage) or body tissues. More than 178 biomarkers have been identified in the field of sepsis, but none seem to have sufficient specificity or sensitivity for routine use in daily clinical practice [1] and some require considerable time, effort, and costs to measure. In addition, the reliability and validity of certain proposed biomarkers have not been thoroughly tested [2]. Procalcitonin (PCT) and C-reactive Protein (CRP) are the biomarkers most commonly used, but have a limited ability to distinguish sepsis from other inflammatory and non-inflammatory states or to predict outcomes.

Serum procalcitonin (S–PCT) has been used as biomarker in sepsis because S–PCT levels are elevated in bacterial, parasitic, or fungal infections, while they remain normal or only slightly elevated in viral infections. Because early recognition of viral versus bacterial meningitis is critical for the prompt initiation of treatment to improve prognosis, a reliable method distinguishing bacterial from viral meningitis could help clinicians limit inappropriate antibiotic treatment. This review was conducted to evaluate the current knowledge on the use of S–PCT as a tool for the diagnosis of meningitis and for distinguishing bacterial meningitis (BM) from viral meningitis (VM).

### **2. Methods**

To identify relevant publications of interest, we conducted a PubMed search on 24 May 2018 using the terms 'procalcitonin and meningitis' as "Title/Abstract" or as "MeSH Terms". Because "Procalcitonin" is not a "Mesh Term" in PubMed, we also used the term "Calcitonin" in "Mesh Terms" during the search. The structure of the search in the "Search details" window of the PubMed website was (procalcitonin [Title/Abstract] OR "calcitonin" [MeSH Terms]) AND (meningitis [Title/Abstract] OR "meningitis" [MeSH Terms]). For the purposes of this review, we then limited the search to "Humans" and only considered manuscripts presenting data on adults. We also reviewed the bibliographies of all identified manuscripts to identify additional relevant publications. For the purposes of this review, we included all types of publications, including case reports, case series, and review articles, regardless of publication date. Publications in languages other than English were included only if they had a meaningful detailed abstract in English.

### **3. Results**

The PubMed literature search generated 157 references, but the number of references was reduced to 125 after limiting the search to "Humans". After further evaluation, 38 publications were included in this review.

In a prospective study published in 1999, Viallon et al. evaluated 105 emergency department patients admitted with suspected meningitis. Based on clinical findings, gram staining, cultures, and Cerebrospinal Fluid (CSF) chemical analysis, 23 patients had bacterial meningitis, 57 had viral meningitis, and meningitis was ruled out in 25 patients. Although two patients with previous antibiotic therapy had S–PCT levels of <0.5, S–PCT was the best marker for differentiating between bacterial and viral meningitis: Using S–PCT of >0.2 ng/mL as the threshold, S–PCT sensitivity and specificity approaches 100% for the diagnosis of acute bacterial meningitis [3].

A prospective study published by Schwarz in 2000 included 30 patients with meningitis (16 with acute bacterial and 14 with abacterial meningitis) and assessed whether S–PCT levels were elevated in patients with bacterial meningitis. Results of the study showed that, although false negative results can occur, S–PCT is a useful variable for distinguishing bacterial from non-bacterial meningitis: Because S–PCT levels do not increase in cases of viral meningitis, even with viral sepsis, increased S–PCT levels indicated bacterial origins of an infection with high specificity [4].

In 2000, Viallon et al. published the results of a prospective study on 179 patients admitted to the emergency department on suspicion of meningitis. Of those, 32 patients had bacterial meningitis and 90 had viral meningitis, whereas 57 patients did not have meningitis. The authors assessed the role of CSF parameters (cytology, protein, glucose, lactate) and serum parameters (CRP, S–PCT) for differentiating between bacterial and viral meningitis, and demonstrated that S–PCT was the most discriminant variable, using a threshold value of 0.93 ng/mL in their population [5].

Shimetani et al. reported extremely high CSF CRP levels in patients with bacterial meningitis, but only in 10% of patients with viral meningitis. Among patients with bacterial meningitis, S–PCT levels were more elevated in those with a more serious infection. PCT levels in CSF did not differ significantly between patients with bacterial, viral, or mycotic meningitis. However, S–PCT levels were very high in all bacterial meningitis patients, especially in the most serious cases [6].

A report by Hoffmann et al., published in 2001, assessed S–PCT levels in 12 adult patients with meningitis and suggested that S–PCT has limited diagnostic value in adults with bacterial meningitis, especially in cases with unusual agents or nosocomial origin. Increased S–PCT levels in bacterial meningitis may indicate the presence of bacterial inflammation outside the Central Nervous System (CNS) [7].

A review by O'Connor published in 2001 included manuscripts published from 1990–2001 to assess the diagnostic usefulness of S–PCT in critical illness. This publication concluded that, although there is debate regarding the superiority of S–PCT as a sepsis biomarker compared to other biomarkers, a number of studies support the usefulness of S–PCT in differentiating between bacterial and viral meningitis [8].

A prospective study on 45 adult patients with CNS infection (20 with bacterial meningitis and 25 with tick-borne encephalitis), published in 2001, evaluated the role of S–PCT and CSF procalcitonin in differentiating acute bacterial vs. viral meningitis. Median S–PCT level was 6.45 ng/mL (0.25–43.76 ng/mL) in patients with bacterial meningitis vs. 0.27 ng/mL (0.05–0.44 ng/mL) in patients with viral meningitis, and the authors concluded that S–PCT and CSF PCT concentrations >0.5 ng/mL seem to be reliable indicator of bacterial CNS infection [9].

A study by Martinez et al. in 2002 attempted to evaluate the role of S–PCT monitoring in the differential diagnosis of ventriculitis in adult Intensive Care Unit (ICU) patients. The study included 15 consecutive ICU patients with ventriculitis and a ventricular catheter in place, and compared these data with 10 patients with community-acquired bacterial meningitis. The authors concluded that, in contrast to bacterial meningitis, monitoring of S–PCT alone is not helpful for the diagnosis of ventriculitis [10].

A case report published in 2004 presented the case of a 73-year-old woman with progressively worsening headaches, nausea, vomiting, and neck stiffness. As her clinical condition deteriorated, she developed diffuse brain edema and hydrocephalus, requiring external ventricular drainage (EVD) and admission to a neurologic ICU, with a subsequent diagnosis of severe post-myelographic chemical meningitis. The authors compared CSF and serum inflammatory markers of this patient versus seven patients with proven bacterial meningitis and concluded that S–PCT may be useful in differentiating between bacterial and chemical causes of CNS inflammation [11].

A study by Kepa et al. assessed the role of CSF PCT and S–PCT levels in the differential diagnosis of adults with CNS infections. The study included 17 patients with bacterial meningoencephalitis and 16 patients with lymphocytic meningitis and showed that CSF and plasma PCT levels were significantly different between these two patient groups. These results supported the usefulness of measuring plasma PCT levels in the differential diagnosis of CNS infections in adults. With regards to the role of CSF PCT, the authors concluded that CSF PCT levels are less important for differential diagnosis, but correlate with the severity of bacterial meningoencephalitis and can be taken into consideration when predicting prognosis and outcomes [12].

In 2005, a study by Viallon et al. described the change in S–PCT levels during the treatment of 48 patients with community-acquired acute bacterial meningitis. Bacterial infection was documented in 45 patients and initial antibiotic treatment was effective in all patients. Serum PCT levels were measured on admission and on day two, and showed that S–PCT levels decline rapidly with appropriate antibiotic therapy. The authors concluded that the rapid decline of S–PCT levels reduces the value of performing lumbar puncture 48 to 72 h after admission to assess the effectiveness of antibiotic therapy [13].

Ernst and colleagues evaluated serum and CSF procalcitonin levels in patients with dementia disorders and neuro-inflammation. The study included 40 patients with probable Alzheimer's disease, 12 patients with frontotemporal dementia (FTD), 8 patients with dementia with Lewy bodies (DLB), 12 patients with vascular dementia (VD), 16 patients with acute neuroinflammation, and 50 non-dementia control patients (18 surgery patients and 32 patients with other neurologic diseases)and showed that, compared to non-dementia controls, CSF procalcitonin levels were increased

in patients with dementia diseases and acute neuro-inflammation. In addition, in matched serum samples, S–PCT levels were elevated in meningitis patients, but not in dementia patients [14].

An observational cohort study by Knudsen et al. included 55 patients with suspected meningitis and compared the diagnostic value of serum sCD163 levels, CRP, and procalcitonin in bacterial infection and meningitis and showed that, although elevated serum sCD163 levels seem to be the most specific biomarker for differentiating between bacterial and non-bacterial disease (specificity 0.91; sensitivity 0.47), the overall diagnostic accuracy of CRP (Area Under the Curve (AUC) = 0.91) and PCT (AUC = 0.87) were superior compared to sCD163. The authors concluded that, although PCT and CRP had very high accuracy for distinguishing between bacterial and viral infection, none of them were useful as an independent tool for diagnosis in patients presenting with purulent meningitis [15].

A prospective multicenter study, published in 2007, included 151 patients with bacterial or nonbacterial meningitis and negative initial Gram stains from three teaching hospitals in France and reported laboratory data, including results of CSF analysis (CSF leukocyte count, percentage of CSF leukocyte, CSF/blood glucose ratio, CSF protein), serum CRP, and serum PCT, together with clinical findings and outcomes. The study evaluated the accuracy of laboratory results in differentiating between bacterial and non-bacterial meningitis in patients with meningitis and a negative gram stain, and concluded that CSF laboratory results have some role in distinguishing bacterial from non-bacterial meningitis, whereas serum CRP (AUC 0.81 (95% CI 0.58–0.92) and S–PCT levels (AUC 0.98, 95% CI, 0.83–1.00) are excellent predictors of bacterial meningitis, with S CPT being clearly superior (*p* < 0.05) [16].

A prospective study from the Saint-Etienne University Hospital in France collected data from all patients admitted to the emergency unit with suspected meningitis between 1997 and 2009. Data were collected on 97 patients with bacterial meningitis and 218 patients with viral meningitis, but, after 62 patients with Bacterial Meningitis (BM) were excluded for various reasons, the study only included data from 35 patients with BM and negative direct CSF examination. The aim of the study was to determine the ability of several parameters used for the diagnosis of acute meningitis in differentiating between bacterial and viral meningitis in adult patients with a negative CSF examination. In this study, S–PCT had a 95% sensitivity, 100% specificity, and 100% negative predictive value, as well as a 97% positive predictive value for distinguishing BM versus Viral Meningitis (VM) when using a diagnostic cut-off level of 0.28 ng/mL (AUC, 0.99; 95% CI, 0.99 to 1) [17].

A prospective study on 36 adult patients with acute meningitis was published in 2012. The aim of the study was to evaluate the role of serum procalcitonin levels over time during treatment for central nervous system infections. Serum procalcitonin levels were measured before the initiation of treatment and 24 and 72 h after treatment started. Results showed that mean PCT levels were higher in patients who did not improve and that the reduction of serum PCT levels were more significant after 72 h in patients who improved. The authors emphasized the role of serum PCT levels as a marker for follow up in treating patients with bacterial meningitis [18].

A study by Choi assessed the value of serum procalcitonin in differentiating post-operative bacterial meningitis (PBM) versus postoperative aseptic meningitis (PAM) after neurological surgery and included patients who had cerebrospinal fluid pleocytosis within 14 days of surgery. The study compared PCT in 14 patients with PBM against 64 patients with PAM and showed that serum PCT had limited value for diagnosing PBM and serum PCT levels of ≥0.15 ng/mL had 80% specificity. However, CPT combined with other biomarkers can be a useful adjunct for increasing diagnostic sensitivity [19].

An observational study by Tian et al., from Guadong, China and published in 2014, investigated the value of procalcitonin in the discrimination between sepsis and systemic inflammatory response syndrome (SIRS). The study included patients treated in a neurological intensive care unit and serum levels of C-reactive protein and S–PCT were evaluated on admission day, on the day of diagnosis of SIRS or sepsis, and on days three and seven after the diagnosis. Results of the study showed significant differences in S–PCT levels between groups at all stages of sepsis. The authors concluded that S–PCT has significant value as an index for discriminating sepsis from SIRS and in determining sepsis severity [20].

A study by Abdelkader et al. published in 2014 evaluated 40 patients with suspected acute meningitis and negative gram stains compared to 10 healthy controls. The goal of the study was to evaluate the role of S–PCT in differentiating bacterial from aseptic meningitis in patients with negative cerebrospinal fluid (CSF) examination on admission and after three days of treatment, and to assess the role of PCT and other inflammatory markers in relation to treatment efficacy. In this study, patients in the bacterial group had significantly higher serum PCT on admission compared to the aseptic group (2.49 ± 2.54 vs. 0.89 ± 0.69, *p* < 0. 001), and there was a significant difference in bacterial versus. aseptic meningitis, even after three days of treatment (1.70 ± 1.58 vs. 0.64 ± 0.51, *p* < 0.001) [21].

A prospective observational study, published by Shen et al. in 2015, assessed the diagnostic value of serum and CSF PCT levels in 120 patients with meningitis-like symptoms and showed that both S–PCT and CSF PCT levels were increased in patients with bacterial meningitis (BM). The area under the Receiver Operator Characteristic (ROC) curve was 0.96 (CI 0.93–1.00) for S–PCT versus 0.9 (CI 0.83–0.96) for CSF PCT in the diagnosis of BM. When using 0.88 ng/mL as a threshold, S–PCT had an 87% sensitivity and 100% specificity for the diagnosis of BM. The study concluded that both S–PCT and CSF PCT have value for the diagnosis of BM, but the diagnostic value of S–PCT is superior [22].

In another prospective observational study, Omar et al. collected data on CRP, S–PCT, and CSF cultures every other day in 36 adult patients with severe head trauma and ventriculostomy, and observed elevated S–PCT concentration in all five patients who developed ventriculostomy-related infections. Mean serum PCT was <2.0 ng/mL in patients with negative CSF cultures versus 4.18 ng/mL in patients with positive cultures. The study concluded that an early increase of S–PCT levels is a valid indicator of bacterial CNS infection in patients with head trauma and External Ventricular Drainage (EVD) [23].

A retrospective clinical study by Li et al. assessed the diagnostic value of CSF procalcitonin combined with CSF lactate levels in distinguishing post-neurosurgical bacterial meningitis (PNBM) from aseptic meningitis in 178 hospitalized patients with suspected PNMB (50 patients with PNBM vs. 128 patients without PNBM). Median (min, max) CSF procalcitonin levels were 0.2 (0–3.1) in patients with PNBM versus 0 (0–0.5) in patients with non-PNBM (*p* < 0.001), and ROC analysis revealed a cut-off value of 0.075 ng/mL (AUC = 0.746, sensitivity 68.0%; specificity 72.7%, *p* < 0.001) for CSF procalcitonin. Similarly, median (min, max) CSF lactate levels were 5.3 (2.2–10.6) in patients with PNBM versus 2.3 (1.2–5.4) in patients with non-PNBM (*p* < 0.001), and ROC analysis revealed a cut-off value of 3.45 mmol/L (AUC = 0.943, sensitivity 90.0%; specificity 84.4%, *p* < 0.001) for CSF lactate. The study showed that PNBM patients have significantly higher CSF procalcitonin and CSF lactate levels compared with non-PNBM patients and concluded that CSF lactate and PCT levels have significant diagnostic value for PNBM, and could be useful in differentiating PNBM from non-PNBM [24].

A publication by Konstantinidis et al. in 2015 evaluated CSF procalcitonin levels and compared CSF procalcitonin levels with CSF levels of other established markers of infection, such as CRP, high-sensitivity CRP, and White Blood Cells (WBC) in 30 ICU, Medicine, Neurology, Hematology, and Pediatric patients with bacterial (*n* = 19) or viral (*n* = 11) meningitis, and in 28 patients with non-infectious diseases. In this study, CSF PCT levels were 4.714 ± 1.59 ng/mL in bacterial meningitis versus 0.1327 ± 0.03 ng/mL in patients with viral meningitis versus <0.1 ng/mL in patients with non-infectious diseases, with the authors concluding that S–PCT can be helpful in distinguishing bacterial meningitis from viral meningitis and other noninfectious CNS diseases [25].

A meta-analysis published in 2015 by Vikse et al. included nine studies with a total of 725 patients and concluded that serum procalcitonin had a pooled sensitivity of 0.90 (95% CI 0.84–0.94), a specificity of 0.98 (0.97–0.99), a positive likelihood ratio of 27.3 (8.2–91.1), a negative likelihood ratio

of 0.13 (0.07–0.26), a diagnostic odds ratio of 287.0 (58.5–1409.0), and, thus, is far superior to CRP for rapid differentiation between bacterial and viral meningitis [26].

Kim et al. compared S–PCT levels in 26 patients with tuberculosis meningitis versus 70 patients with BM and 49 patients with VM in a retrospective study and showed that low S–PCT levels (≤1.27 ng/mL) independently distinguished tuberculosis meningitis from bacterial meningitis, with a 96.2% sensitivity and a 62.9% specificity. However, S–PCT levels were not significantly different in patients with tuberculosis versus viral meningitis. Logistic regression showed that an S–PCT level of >0.4 ng/mL was an independent predictor of a poor prognosis in patients with tuberculosis meningitis and had a negative correlation with Glasgow Coma Scale (GCS) scores at discharge (*r* = 0.437, *p* = 0.026) [27].

A prospective observational study published by Morales-Casado et al. in 2016 assessed the role of 32 clinical and epidemiological variables as predictors of bacterial meningitis in 154 patients aged over 15 years who presented in the Emergency Department with symptoms of acute meningitis. Multivariate logistic regression showed that four variables (S–PCT, CSF lactate ≥33 mg/dL, CSF glucose <60% of blood value, and CSF polymorphonuclears ≥50%) were excellent tools for the prediction of bacterial meningitis; the model using S–PCT ≥0.8 ng/mL and CSF lactate ≥33 mg/dL had an AUC of 0.992, with a 99% sensitivity and a 98% specificity for predicting bacterial meningitis (95% CI: 0.979–1; *p* < 001) [28].

Another prospective observational study by Morales-Casado et al. evaluated the usefulness of inflammatory markers for the diagnosis of bacterial meningitis in 220 patients and showed that S–PCT had the highest AUC (0.972; 95% CI, 0.946–0.998; *p* < 001) for the diagnosis of BM. Using 0.52 ng/mL as a cutoff, S–PCT had 93% sensitivity and 86% specificity for the diagnosis of BM overall, but sensitivity was 96% and specificity was 75% in patients >75 years old [29].

In 2016, Wei and colleagues published a systematic review and meta-analysis on the role of procalcitonin in the diagnosis of bacterial versus non-bacterial meningitis. The review included twenty-two studies with a total of 2058 patients; diagnostic accuracy of S–PCT and CSF PCT was assessed using the bivariate model and analysis showed that PCT is a useful biomarker for the diagnosis of bacterial meningitis: CSF PCT had 0.86 specificity and 0.8 sensitivity, whereas S–PCT had 0.97 specificity and 0.95 sensitivity [30].

A prospective, observational study published in 2016 by Morales Casado et al. evaluated serum PCT and C-reactive protein as markers for detection of bacterial meningitis in 98 patients diagnosed with acute meningitis in the emergency department (38 pts with BM, 33 with VM, 15 with probable VM, and 12 with presumptively diagnosed, partially treated acute meningitis). Data analysis showed that S–PCT levels were significantly higher in patients with BM (11.47 ± 7.76 ng/mL vs. 0.10 ± 0.15 ng/mL in viral meningitis, *p* < 0.001). Using 1.1 ng/mL as cutoff, S–PCT as diagnostic tool achieved 94.6% sensitivity, 72.4% specificity, 95.4% NPV, and 69.2% PPV, and AUC was 0.965 (95% CI, 0.921–1; *p* < 0.001). Based on these results, the authors concluded that S–PCT performs better than CRP in the detection of bacterial meningitis [31].

In a prospective observational study published in 2017, Zhang et al. measured S–PCT and CSF PCT, high-sensitivity C-reactive proteins (Hs-CRP), proteins, chloride, and glucose in three patient groups: 24 patients with suppurative meningitis, 20 with VM, and 22 with tuberculous meningitis (TBM). S–PCT values were significantly higher in the suppurative meningitis group, but declined significantly in suppurative meningitis patients after 72 h and seven days of treatment. In addition, admission CSF PCT levels were significantly lower in VM compared to TBM and suppurative meningitis patients, but CSF PCT values did not change significantly with treatment. The authors concluded that S–PCT changes over time can be useful in evaluating disease progression and response to treatment in patients with suppurative meningitis [32].

A retrospective clinical study on 80 patients with BM and 58 with VM, published by Park in 2017, showed that S–PCT >0.12 ng/mL is a significant marker for differentiating BM from VM, and also that

S–PCT levels >7.26 ng/mL are associated with higher risk of death (OR = 9.09, 95% CI: 1.74–47.12, *p* = 0.016) [33].

Last, a prospective observational study published by Li et al. in 2017 included 143 ICU patients (49 with BM, 25 with TBM, 34 with viral meningitis/encephalitis (VM/E), 15 with autoimmune encephalitis (AIE), and 20 with non-inflammatory nervous system diseases (NINSD) to assess the value of CSF PCT, S–PCT and other biomarkers in the diagnosis of BM. In this study, CSF PCT levels (median, range) were significantly (*P* < 0.01) higher in BM patients (0.22, 0.13–0.54 ng/mL) compared to TBM (0.12, 0.07–0.16 ng/mL), VM/E (0.09, 0.07–0.11 ng/mL), AIE (0.06, 0.05–0.10 ng/mL), or NINSD (0.07, 0.06–0.08 ng/mL) patients. Furthermore, CSF PCT had the highest area under the receiver operating characteristic curve (AUROC) (0.881; 95% CI 0.810–0.932; cutoff 0.15 ng/mL; sensitivity 69.39%; specificity 91.49%), whereas S–PCT was less useful (AUROC 0.759, 95% CI 0.669–0.849, cutoff 0.19 ng/mL, with sensitivity 67.35% and specificity 75.53% for the diagnosis of BM [34]. Findings of clinical studies evaluating the role of PCT in patients with known or suspected meningitis are summarized in Table 1. Quantitative measures, including AUC, sensitivity, specificity, and cut-off points, reported in clinical studies included in this review are summarized in Table 2.


**Table 1.** Clinical studies evaluating the role of procalcitonin (PCT) in patients with known or suspected meningitis.


#### **Table 1.** *Cont.*


**Table 1.** *Cont.*

ABM = Acute Bacterial Meningitis, AD = Alzheimer's disease, AIE = Autoimmune Encephalitis, ASM = Aseptic Meningitis, AUC = Area Under the Curve, AUROC = Area Under The Receiver Operating Characteristic Curve, BM = Bacterial Meningitis, CI = Confidence Interval, CRP = C-reactive protein, CSF = cerebrospinal fluid, DLB = Dementia with Lewy bodies, ED = Emergency Department, EVD = External Ventricular Drainage, FTD = Frontotemporal dementia, LP = Lumbar Puncture, NBM = Nonbacterial Meningitis, NICU = Neurological Intensive Care Unit, NINSD = Non-Inflammatory Nervous System Disease, NPV = Negative Predictive Value, PMN = Polymorphonuclear, PNBM = Post-Neurosurgical Bacterial Meningitis, PPV = Positive Predictive Value, Pts = patients, S–PCT = Serum Procalcitonin, SIRS = Systemic Inflammatory Response Syndrome, Tb = Tuberculosis, TBE = Tick-Borne Encephalitis, TBM = Tuberculous Meningitis, VD = Vascular Dementia, VM/E = Viral Meningitis/Encephalitis, VM = Viral Meningitis.

**Table 2.** Area under the Curve (AUC), sensitivity, specificity, and cut-off points reported in clinical studies evaluating Serum Procalcitonin (S–PCT) or Cerebrospinal Fluid Procalcitonin (CSF PCT) in patients with known or suspected meningitis.



**Table 2.** *Cont.*

### **4. Discussion**

Procalcitonin, a precursor of calcitonin, is a 116 amino acid peptide and a member of the calcitonin superfamily of peptides, with a molecular weight of 14.5 kDa. PCT is synthesized by the parafollicular C cells of the thyroid gland and is involved in calcium homeostasis. In addition, PCT is also produced by the neuroendocrine cells of the lung and the intestine. In the CNS, cells likely to be sources of CSF PCT include the neurons, astrocytes, and microglia in the parenchyma and meningeal cells. Normal S–PCT concentration is <0.05 ng/mL, with a reported half-life of 25–30 h [35]. Procalcitonin is considered a sensitive and specific marker of certain bacterial infections, such as pneumonia, meningitis, and pyelonephritis, and has been used as tool for the assessment of disease severity. In addition to bacterial infections, increased PCT levels have been identified in other clinical conditions, including severe fungal infections, trauma, burns, major surgery, and medical therapy that stimulates cytokine production. In these cases, S–PCT levels are less elevated and rarely exceed 0.5 ng/mL [36]. There are several hypotheses regarding the pathophysiology of PCT; most suggesting that procalcitonin may be involved in calcium metabolism, cytokine network, and the modulation of nitric oxide (NO)

synthesis [37]. In sepsis, PCT hypersecretion emanates from multiple tissues throughout the body, which are not traditionally viewed as endocrine tissues. It is likely that PCT in sepsis potentiates the inflammation cascade by increasing leukocyte-derived cytokines and augmenting reactive oxygen species [38]. PCT secretion is stimulated in bacterial infections by various cytokines, such as IL–1, IL–6, and tumor necrosis factor-alpha. In contrast, PCT production is down-regulated in viral infections, probably due to increased interferon gamma production. Consequently, PCT is considered a useful tool for diagnosing sepsis and repeat S–PCT measurements over time can be used to monitor response to therapy. Most currently used inflammatory markers do not reliably differentiate between a systemic inflammatory response and sepsis. However, because PCT is, generally, not induced by severe viral infections or non-infectious inflammatory reactions, PCT can help distinguish bacterial from viral infections and differentiate between infectious and non-infectious origins of systemic inflammatory response syndrome (SIRS), acute respiratory distress syndrome (ARDS), pancreatitis, cardiogenic shock, and acute rejection of transplanted organs [39].

In CNS infections, disruption of the blood brain barrier (BBB) has been documented in patients with bacterial infections and in experimental models [40,41]. Elevated CSF PCT levels in bacterial meningitis patients seem to be the result of this mechanism and some studies have shown higher CSF PCT levels in patients with Gram-negative bacteria compared to patients with Gram-positive bacteria [42]. In CNS infection cases, microglia and meningeal cells express the responding receptors (Toll-like receptors [TLRs]) to the invading bacteria [43]. Several questions regarding PCT synthesis and secretion by brain cells during bacterial meningitis have not been resolved and need further investigation.

Prompt diagnosis and appropriate antimicrobial treatment are of paramount importance and can contribute to a reduced morbidity and mortality in sepsis. Procalcitonin is an acute-phase protein with faster kinetics than C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR), and is detectable in serum within 4–6 h after the onset of a bacterial infection. PCT serum levels peak within 24 h and start to decline by approximately 50% daily with effective treatment [44–46]. Although there is no "gold standard" for the diagnosis of most infections, several biomarkers have been used as tools to monitor disease progression. An ideal marker should help with early diagnosis and therapeutic decision-making in bacterial infections and should also help clinicians assess the course and prognosis and, in that regard, PCT seems to be superior compared to other commonly used biomarkers.

Despite advances in diagnosis and treatment, bacterial meningitis is a neurological emergency, requires treatment in a high acuity care unit, and remains an important cause of mortality. The diagnosis and management of bacterial meningitis requires various biological tests and a multidisciplinary approach. Empiric antimicrobial and adjunctive therapy should start as soon as there is clinical suspicion of meningitis. Regarding laboratory findings, a left shift in peripheral white blood cell count, elevated serum PCT and C-reactive protein, CSF pleocytosis with predominance of polymorphonuclear leukocytes, and decreased glucose concentration are predictive of bacterial meningitis. CSF analysis is a gold standard for the diagnosis of meningitis: CSF gram staining reveals bacteria in 50% to 80% of cases and cultures are positive in 80% of cases, at best. However, the sensitivity of both tests is <50% in patients already receiving antibiotics. CSF leukocyte count, protein, glucose and lactate concentration, and a latex agglutination test adapted for the rapid direct detection of soluble bacterial antigens in CSF lack the specificity and sensitivity for the diagnosis of meningitis and can only define a clinical probability. In fact, the relatively imprecise nature of the cutoff values for these markers can make their interpretation difficult. Furthermore, in the early phases of acute bacterial and viral meningitis, differential diagnosis is difficult because signs and symptoms are often non-specific and, therefore, differentiation between bacterial and viral meningitis remains a difficult problem for clinicians. Biomarkers, like CRP, procalcitonin, or sTREM–1, may be useful for diagnosis and can help differentiate between viral and bacterial meningitis [47,48]. Serum and CSF PCT levels can be more useful in the diagnosis of bacterial meningitis and in distinguishing bacterial from viral meningitis. A systematic review published by Markanday in 2015 compared serum PCT versus CRP as markers for

bacterial infection and showed that, compared to CRP, S–PCT had a higher sensitivity and specificity for differentiating bacterial from noninfectious causes of inflammation [49].

Diagnosis of meningitis requires a detailed history and physical examination combined with high clinical suspicion and appropriate cultures. Prognosis of meningitis depends on rapid diagnosis, identification of the cause, and prompt implementation of appropriate antibiotic treatment. Because clinical and laboratory data available within a few hours after hospital admission are not reliable (except for when bacteria are found in CSF under the microscope), use of biological markers has been proposed as a tool to improve the accuracy of initial diagnosis and, in this setting, serum and CSF procalcitonin measurements seem to be of great value. Because of its high specificity and positive predictive value, elevated S–PCT concentrations (>0.5 ng/mL) indicates ongoing and, potentially, severe systemic infection. Because C-reactive protein is the inflammatory marker most widely used in emergency departments to discriminate bacterial from viral infections, Gerdes et al. published a meta-analysis in 1998 from 35 studies aiming to assess the usefulness of CRP in discriminating bacterial from viral meningitis. The meta-analysis showed that, although the majority of authors propose using CRP as an additional tool for discriminating bacterial from viral meningitis, only negative CRP tests are highly informative in most clinical settings [50]. However, procalcitonin seems to be a valuable tool for discriminating the causative factor of meningitis as, since 1997 and 1998, two French studies showed that, using a cut-off range of 0.5–2 ng/mL, S–PCT had 100% sensitivity and specificity in discriminating bacterial from viral meningitis [51]. Similarly, a more recent meta-analysis published by Vikse et al. in 2015 showed that procalcitonin has a 90% sensitivity and a 98% specificity in the discrimination between bacterial and viral meningitis [26].

This review aims to provide clinicians with an overview of the role of S–PCT and CSF-PCT as diagnostic markers in CNS infections. Several publications assessing the role of PCT as a guide for antibiotic therapy in adult meningitis patients suggest that S–PCT is a sensitive, specific, and prognostic marker of bacterial infections, therefore, S–PCT and CSF PCT measurement can help differentiate bacterial from viral meningitis. Interpretation of PCT levels must take into consideration the clinical presentation of the CNS infection. In addition, knowledge of assay characteristics is important for setting specific cut-off values and functional assay sensitivities. Most clinical studies presented in this review have limitations, including small sample sizes and inconsistent reporting of laboratory findings: In some, AUC values and cut-off values were not reported, while a few were published in languages other than English and, therefore, we only included in the review data reported in the Abstract. Therefore, even though S–PCT seems to be a useful biomarker for the diagnosis and possible prognosis in patients with BM, additional data from larger, well-designed studies are needed to better evaluate the role of procalcitonin in the differentiation between viral and bacterial meningitis and as tool to improve the overall management of patients with meningitis.

### **5. Conclusions**

Serum PCT is a biomarker with high sensitivity and specificity in identifying patients with sepsis and can be useful for the diagnosis of bacterial infections. This literature review identified several studies evaluating the role of S–PCT in the assessment of patients with central nervous system infections. Published data suggests that, compared to other acute phase biomarkers, S–PCT is superior as a sepsis biomarker in acute meningitis and can help differentiate bacterial from viral meningitis. Combined with good clinical judgment and appropriate use of antimicrobial agents, S–PCT could be a valuable adjunct in the timely diagnosis and management of sepsis in patients with CNS infection.

**Author Contributions:** D.V. did literature search, wrote and edited manuscript; M.P. did literature search and wrote manuscript; N.P. did literature search; E.S. did literature search; V.K. did literature search; C.P. did literature search, M.K. did literature search, edited, revised and submitted the manuscript.

**Acknowledgments:** This work, including the costs to publish in open access was supported in its entirety by Department funds, without financial support by industry or by any external grants.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Imaging Characteristics of Malignant Sinonasal Tumors**

### **Masaya Kawaguchi 1,2, Hiroki Kato 1,\*, Hiroyuki Tomita 2,\*, Keisuke Mizuta 3, Mitsuhiro Aoki 3, Akira Hara <sup>2</sup> and Masayuki Matsuo <sup>1</sup>**


Received: 30 October 2017; Accepted: 4 December 2017; Published: 6 December 2017

**Abstract:** Malignancies of the nasal cavity and paranasal sinuses account for 1% of all malignancies and 3% of malignancies of the upper aerodigestive tract. In the sinonasal tract, nearly half of all malignancies arise in the nasal cavity, whereas most of the remaining malignancies arise in the maxillary or ethmoid sinus. Squamous cell carcinoma is the most common histological subtype of malignant tumors occurring in this area, followed by other epithelial carcinomas, lymphomas, and malignant soft tissue tumors. Although many of these tumors present with nonspecific symptoms, each tumor exhibits characteristic imaging features. Although complex anatomy and various normal variants of the sinonasal tract cause difficulty in identifying the origin and extension of large sinonasal tumors, the invasion of vital structures such as the brain, optic nerves, and internal carotid artery affects patients' prognosis. Thus, diagnostic imaging plays a key role in predicting the histological subtype and in evaluating a tumor extension into adjacent structures. This article describes the computed tomography and magnetic resonance imaging findings for malignant sinonasal tumors.

**Keywords:** sinonasal tract; malignant tumor; CT; MRI

### **1. Introduction**

Sinonasal neoplasms are relatively rare and malignant sinonasal neoplasms are more common than their benign counterparts. Sinonasal malignancies comprise only 3% of all head and neck cancers and 1% of all malignancies [1–3]. The complex anatomy of the region and the rare occurrence of these tumors pose diagnostic and therapeutic challenges. Of the various histological subtypes of malignant sinonasal tumors, squamous cell carcinoma (SCC) is the most common subtype, whereas the other subtypes, such as adenocarcinoma, minor salivary gland carcinoma, undifferentiated carcinoma, neuroendocrine carcinoma, and nonepithelial malignancies (such as sarcoma, lymphoma, plasmacytoma, olfactory neuroblastoma, and melanoma) are considerably less common. The treatment modalities vary depending on the tumor histological subtype, location, and extent of the disease and include surgery, radiation, chemotherapy, or a combination of two or more of these modalities. The prognosis of the patients largely depends on tumor histology, location, and stage.

### **2. Anatomy**

The sinonasal tract comprises the nasal cavity, maxillary sinus, ethmoid sinus, frontal sinus, and sphenoid sinus. It includes various tissue types such as epithelium, mucosal epithelium, vessel, nerve, cartilage, bone, and lymphatic tissue. The maxillary, ethmoid, nasal, frontal, palatine, sphenoid, and lacrimal bones are also included. This area comprises bone and cartilage lined with ciliated respiratory epithelium and is located between the orbit and the oral cavity. This area is also close to the frontal cortex through the cribriform plate of the ethmoid bone and is connected with the cerebrum by vessels, lymph channels, and nerves.

The nasal cavity is divided by the nasal septum in the midline. Bilateral nasal cavities include the superior, middle, and inferior turbinates; in addition, the nasal meatus is located under each of them. The common nasal meatus lies between the nasal turbinate and the nasal septum, and the olfactory cleft is located superior to the lower border of the middle turbinate. The paranasal sinuses are connected to the nasal cavity and categorized according to the location of the ostium. The middle meatus drains the maxillary, frontal, and anterior ethmoid sinuses, whereas the superior meatus drains the posterior ethmoid and sphenoidal sinuses. These are located close to important structures, such as the cavernous sinus, the internal carotid artery, the pituitary gland, and the optic nerve.

The upper third of the nasal cavity, the frontal sinus, and parts of the ethmoid and sphenoid sinuses are supplied by the ophthalmic artery, whereas most of the remaining sinonasal tracts are supplied by facial and maxillary arteries. The olfactory mucosa in the upper part of the nasal cavity is innervated by the olfactory nerve. The ophthalmic nerve (V1) provides sensory innervation to the ethmoid sinus, sphenoid sinus, lateral wall of the nasal cavity, and the anterior part of the nasal septum. The maxillary nerve (V2) provides sensory innervation to the maxillary sinus. Submandibular lymph nodes drain the anterior components of the sinonasal tract (anterior drainage pathway), whereas the retropharyngeal lymph nodes drain the posterior components of the sinonasal tract (posterior drainage pathway). These routes of lymphatic drainage finally reach the superior deep cervical lymph nodes.

### **3. Clinical Presentation**

The clinical presentations of sinonasal malignancies are nonspecific and identical to those of inflammatory sinus disease, such as nasal obstruction, rhinorrhea, epistaxis, headache, and facial pain. In addition, these malignancies are often asymptomatic until they erode and invade the adjacent structures. Therefore, sinonasal malignancies are usually diagnosed in advanced stages, and the survival rate and prognosis of these patients remain poor. Sinonasal malignancies should be considered in patients with unilateral nasal obstruction or recurrent epistaxis. Progressive sinonasal malignancies that often invade the adjacent structures produce characteristic symptoms, such as cranial neuropathies due to intracranial extension, facial subcutaneous soft tissue swelling, exophthalmos, diplopia, visual disturbances, eye movement disorders, olfactory dysfunction, and respiratory symptoms. Early detection and adequate treatment are required to improve the survival and mortality rates.

### **4. Imaging**

Computed tomography (CT) and magnetic resonance imaging (MRI) are very useful tools for the assessment of tumor size, nature, extent, and invasion. The evaluation of the potential extension into adjacent regions impacts the surgical or therapeutic planning, particularly in cases with the involvement of the anterior and middle cranial fossa, orbit, pterygopalatine fossa, palate, or infratemporal fossa (masticator and parapharyngeal space). CT is the most commonly used imaging modality because of its wider availability, easy access, lower cost, and potential to offer greater anatomic detail. In comparison to MRI, CT is particularly effective in delineating calcification and evaluating the pattern of bone invasion. Intralesional calcifications are observed in some sinonasal disorders, such as adenocarcinoma, olfactory neuroblastoma, inverted papilloma, fibrous dysplasia, osteoma, osteosarcoma, cartilaginous tumor, fungal sinusitis, and dentigerous tumor. Characteristic patterns of bone invasion help predict the tumor histology. High grade malignancies show extensive bony destruction, whereas small round cell tumors show permeative invasion and lack of bony destruction. Benign lesions and low grade malignancies may cause bony expansion due to their slow and expansile growth. Contrast-enhanced CT is invaluable for the identification of the feeding artery (because of its

high spatial resolution) and for the diagnosis of hypervascular tumors. In contrast, MRI provides higher contrast resolution and affords an excellent characterization of the soft tissue components of the tumor. The signal intensity within a tumor varies according to the tissue components. Malignant tumors usually exhibit nonspecific hyperintensity on T2-weighted images (T2WI) and hypo- to isointensity on T1-weighted images (T1WI). On T2WI, mucinous or cartilaginous tumors show marked hyperintensity, hypercellular tumors show slight hyperintensity, and tumors with fibrosis, calcification, or flow void show hypointensity. On T1WI, hyperintensity within a tumor is indicative of the presence of methemoglobin, melanin, lipid, protein, and mineral elements. Diffusion-weighted image (DWI) with measurement of apparent diffusion coefficient (ADC) captures the degree of Brownian movement of the water molecules in tissues, which serves as a useful imaging biomarker. Low-ADC lesions with strong diffusion restriction indicate hypercellularity, abscess, or hemorrhage, whereas high-ADC lesions indicate hypocellularity, mucus, cartilage, or fluid. Therefore, DWI with ADC measurement is usually useful to differentiate between benign and malignant tumors. The ADC values of malignant sinonasal tumors (0.87–1.10 × <sup>10</sup>−<sup>3</sup> mm2/s) have been shown to be significantly lower than those of benign sinonasal lesions (1.35–1.78 × <sup>10</sup>−<sup>3</sup> mm2/s) [4–6]. Contrast-enhanced MRI is a useful method in detecting perineural spread and dural invasion. Perineural spread can be diagnosed as nerve thickening, widening of the neural foramen, loss of perineural fat, and enhancement of the nerve. Linear enhancement alone is not a conclusive sign of dural invasion; however, dural thickening > 5 mm, pial enhancement, or the presence of focal dural nodules are indicative of dural invasion.

### **5. Sinonasal Malignancies**

### *5.1. Squamous Cell Carcinoma*

Squamous cell carcinoma (SCC) is the most common histological subtype and accounts for more than half of all sinonasal malignant tumors. It most commonly affects patients in the sixth and seventh decade of life with male predominance. The maxillary sinus is the most frequently affected site, followed by the nasal cavity and the ethmoid sinus. Primary SCCs of the sphenoid sinus and frontal sinus are rare. In the past, sinonasal SCC has been linked to cigarette smoking and occupational exposures [7]. However, in recent years, the generational changes in sexual behavior may have led to an increased positivity rate for human papillomavirus (HPV) among patients with sinonasal SCC; consequently, HPV has been identified in 32–62% of all sinonasal SCCs [8,9]. HPV positivity is more common in SCCs of the nasal cavity than in paranasal SCCs [9]. As with oropharyngeal SCCs, patients with a HPV-positive sinonasal SCC have a better prognosis than patients with a HPV-negative sinonasal SCC [8,9].

Sinonasal SCCs are characterized by aggressive bony destruction of the adjacent sinus walls (Figure 1). Because sinonasal SCCs are often detected at an advanced stage, the invasion of the contralateral sinonasal area, orbital wall, infratemporal fossa, and skull base is sometimes observed. Hypoxia is a common feature in most cases of SCC, and prolonged oxygen deprivation often leads to chronic hypoxic stress and consequent tumor necrosis [10]. Thus, intratumoral necrosis is also one of the characteristic findings in SCCs. On MRI, isointensity on T1WI, slight hyperintensity on T2WI, and moderate enhancement on contrast-enhanced T1WI are typical and nonspecific imaging findings for SCCs. Smaller lesions are typically homogeneous, whereas larger tumors are usually more heterogeneous and exhibit areas of necrosis and hemorrhage [11]. In the maxillary sinus, the ADC values of SCC (0.95 × <sup>10</sup>−<sup>3</sup> mm2/s) were higher than those of non-Hodgkin's lymphoma (NHL) (0.61 × <sup>10</sup>−<sup>3</sup> mm2/s) [12].

**Figure 1.** Squamous cell carcinoma of the left maxillary sinus. Contrast-enhanced CT image showing an ill-demarcated, heterogeneously enhanced bulky mass with extensive bony destruction (arrows).

### *5.2. Adenocarcinoma*

Adenocarcinomas account for 10–20% of all sinonasal malignancies. Sinonasal adenocarcinomas are categorized into salivary type adenocarcinomas and non-salivary type adenocarcinomas [13]. The latter are further classified into intestinal-type adenocarcinomas (ITAC) and nonintestinal-type adenocarcinomas (non-ITAC) [13]. Sinonasal ITACs, which histopathologically resemble colorectal adenocarcinoma, can occur sporadically or are associated with occupational exposure to hardwood and leather dust. Sinonasal non-ITACs do not exhibit the histopathological features of sinonasal ITACs or of salivary type adenocarcinomas; they are categorized into high-grade type and low-grade type. Most sinonasal non-ITACs are of the low-grade type, whereas high-grade non-ITACs are rare. Although the age of the patients with sinonasal ITAC and non-ITAC may vary widely, patients in the sixth decade of life are most commonly affected. The nasal cavity is the most common site for ITAC and non-ITAC, whereas the paranasal sinuses are less commonly affected [14].

On CT, sinonasal adenocarcinomas appear as a soft-tissue mass and occasionally exhibit areas of calcification, which reflect the mucin content. In unilateral olfactory cleft adenocarcinomas, the bulging of the nasal septum across the midline and widening of the olfactory cleft are observed [15]. High-grade adenocarcinomas often show bone destruction. Adenocarcinomas arising from the ethmoid sinus may potentially extend to the skull base and intracranially to the frontal lobes [16]. On MRI, the signal intensity of the adenocarcinomas varies according to their mucin content, cellularity and the presence of hemorrhage. Mucin-producing adenocarcinomas usually show hyperintensity on T2WI and exhibit gradual enhancement on contrast-enhanced T1WI, whereas adenocarcinomas without mucin production show iso- to hypointensity on T2WI. The imaging characteristics of adenocarcinomas are often indistinguishable from those of SCCs (Figure 2).

**Figure 2.** Non-intestinal type adenocarcinomas of the right maxillary sinus. T2-weighted image showing a heterogeneously hyperintense lesion (arrow).

### *5.3. Adenoid Cystic Carcinoma*

Adenoid cystic carcinomaa (ACC) are slow-growing and relentless salivary gland tumors comprising epithelial and myoepithelial neoplastic cells. ACCs are the most common malignant salivary gland tumors of the sinonasal tract; sinonasal ACCs account for 10–25% of all head and neck ACCs. The average age at presentation is the fifth to sixth decade of life. The maxillary sinus is the most commonly affected primary site, followed by the nasal cavity, ethmoid sinus, and sphenoid sinus [17]. There are three distinct histopathological subtypes of ACC: tubular, cribriform, and solid subtype. ACCs are characterized by wide local infiltration, perineural spread, a propensity for local recurrence, and late distant metastasis. Bone invasion (41%), perineural invasion (40%), and angioinvasion (3.8%) are observed in the surgical specimens of sinonasal ACCs [17]. Lymph node and distant metastases are uncommon at presentation, but the reported overall recurrence rate is 56.2% [17]. The most common sites of distant metastases are the lungs, followed by the liver and bone.

Low-grade sinonasal ACCs may present as polypoid lesions that remodel the bone and mimic a simple polyp, whereas high-grade sinonasal ACCs may present as large irregular masses with bone destruction and heterogeneous density or signal intensity [18]. The growth pattern of maxillary sinus ACCs can be classified into expansile type with minimal bony defects and destructive type with extensive bony defects, and these tumors usually extend to the nasal cavity and, occasionally, to the retroantral fat pad, pterygopalatine fossa, or orbit [19]. ACCs show isointensity on T1WI and iso- to hyperintensity on T2WI, depending on the amount of cellularity (Figure 3). ACCs exhibit the greatest propensity for perineural spread, and the maxillary division of the trigeminal nerve is most commonly affected by sinonasal ACCs. These tumors sometimes easily extend into intracranial components including the cavernous sinus and the Gasserian ganglion, which are far away from the original site [20,21]. Furthermore, for the surgeons, it is often important to first evaluate on images whether the tumor is resectable or not and far away from vital structures. In cases with an advanced tumor, fluid collection and thickened mucosa caused by the isolated sinuses sometimes make it difficult to diagnose and stage the disease.

**Figure 3.** Adenoid cystic carcinoma of the left maxillary sinus and nasal cavity. T2-weighted image showing a well-demarcated, lobulated, heterogeneously and strongly hyperintense lesion (arrow).

### *5.4. Sinonasal Undifferentiated Carcinoma*

Sinonasal undifferentiated carcinoma (SNUC) is a rare and highly aggressive malignancy, which accounts for approximately 3–5% of all sinonasal cancers. SNUC is a clinicopathologically distinct carcinoma of uncertain histogenesis with no glandular or squamous features. The median age at presentation is the sixth decade of life; the reported male-to-female ratio is 2–3:1. SNUC most commonly arises from the superior nasal cavity and the ethmoid sinus. SNUC usually presents as a rapidly enlarging tumor, and the majority of these patients presents with Stage IV disease. Orbital and intracranial invasion, nodal involvement, and distant metastasis are frequent findings. The recurrence rate is 42.3%. The time to recurrence ranges from 3 to 33 months; 32.1% of patients die of local disease, whereas 14.3% of patients die of metastatic disease [22].

Most SNUCs are larger than 4 cm in maximal diameter at presentation and have ill-defined margins [23]. The aggressive nature of the tumor is reflected in the bone destruction and invasion of adjacent structures, including the paranasal sinuses, anterior fossa, orbit, pterygopalatine fossa, parapharyngeal space, and cavernous sinus [23]. On CT, SNUCs usually appear as a noncalcified mass and show variable contrast enhancement and areas of central necrosis. On MRI, SNUCs show isointensity on T1WI, iso- to hyperintensity on T2WI, and exhibit heterogeneous enhancement on contrast-enhanced T1WI. Owing to the nonspecific imaging findings, it is typically difficult to distinguish between SNUCs and SCCs (Figure 4).

**Figure 4.** Sinonasal undifferentiated carcinoma of the right nasal cavity. Contrast-enhanced CT image showing an ill-demarcated, heterogeneously enhanced lesion (arrow).

### *5.5. Malignant Lymphoma*

The head and neck region is the second most common site for extranodal lymphomas after the gastrointestinal tract. NHL is the second most common malignancy in the sinonasal tract after SCC. Patients classically present in the sixth to eighth decades of life; the reported male-to-female ratio is 2:1. Diffuse large B-cell lymphoma (DLBCL) most commonly arises from the paranasal sinuses; the maxillary sinus is the most common site of involvement, although DLBCL may arise from the nasal cavity. NK/T-cell lymphoma most commonly involves the nasal cavity and shows a predilection for occurrence in Asian and South American populations. B-cell lymphoma has a better prognosis than T-cell lymphoma.

On CT, sinonasal lymphomas frequently show both infiltrative or permeative bony invasion and exhibit varying degrees of regional bony destruction [12]. NHLs with permeative-type tumor invasion typically cross the sinus wall and exhibit remnants of sinus wall as a linear structure within the tumor (Figure 5) [24]. In contrast, bony resorption or remodeling caused by the lymphoma may also be accompanied by bone sclerosis [25]. NHLs usually show isointensity on T1WI and slightly hyperintensity on T2WI [11]. Although the signal intensity of NHLs is nonspecific, the ADC measurement helps differentiate these tumors from other malignancies. In the maxillary sinus, the ADC values of NHL (0.61 × <sup>10</sup>−<sup>3</sup> mm2/s) were shown to be lower than those of SCCs (0.95 × <sup>10</sup>−<sup>3</sup> mm2/s), which reflects the greater cellularity of NHLs [12]. Although NHLs usually appear as a homogeneously enhanced mass, necrotic areas within the tumor are occasionally observed in NK/T-cell lymphoma [26,27].

**Figure 5.** Diffuse large B-cell lymphoma of the left maxillary sinus. Contrast-enhanced CT image showing a homogeneously enhanced lesion accompanied by remaining sinus walls as a linear structure within the tumor (arrows).

### *5.6. Extramedullary Plasmacytoma*

Extramedullary plasmacytomas (EPMs) (also referred to as extraosseous plasmacytomas) are characterized by soft-tissue monochronal plasma cell proliferation with no evidence of underlying multiple myeloma; these tumors account for 4% of all plasma cell tumors. EPMs usually occur in the sixth decade of life; the reported male-to-female ratio is 3–4:1. Approximately 80% of EPMs involve the head and neck region; the nasal cavity and paranasal sinuses are most commonly affected, followed by the nasopharynx, oropharynx, and larynx [28]. Regional recurrence or spread to other osseous sites may occur. Approximately 15% of the patients develop multiple myeloma [28]. The prognosis depends on the tumor size (>5 cm) and nodal involvement [29].

On CT, EMPs typically appear as well-defined, polypoid soft-tissue masses, which exhibit homogenous enhancement. Large tumors may show areas of necrosis, destruction of the adjacent bone, infiltration of the adjacent structures, and vascular encasement [30,31]. On MRI, EMPs show isointensity on T1WI, iso- to slight hyperintensity on T2WI, and exhibit variable enhancement on contrast-enhanced T1WI [30,31] (Figure 6). Because they are highly vascularized tumors, vascular flow void may be observed within the tumor.

**Figure 6.** Extramedullary plasmacytoma of the bilateral nasal cavity. T2-weighted image showing a homogeneously isointense lesion (arrows).

### *5.7. Olfactory Neuroblastoma*

Olfactory neuroblastomas (ONB) arise from the specialized sensory neuroepithelial (neuroectodermal) olfactory cells that are normally found in the cribriform plate, the superior turbinate, and the upper third of the nasal septum. ONBs account for 3% of all sinonasal tumors [32]. ONB may occur at any age, with a peak incidence in the fifth and sixth decades of life; the reported male-to-female ratio is 1.2:1. Direct tumor extension into the adjacent paranasal sinuses, cribriform plate, skull base, orbit, and intracranial cavity is frequently observed. Cervical lymph node metastasis (20–25%) and distant metastases (10–40%) develop over the course of the disease. The most frequent sites of distant metastasis are the lungs, liver, and bone. The most commonly used staging systems are the modified Kadish and Dulguerov classifications.

On CT, ONBs appear as a homogeneous, well-defined soft-tissue mass. Scattered speckled calcifications may be observed within the tumor. The tumor commonly extends into the ethmoid and maxillary sinuses, but rarely involves the sphenoid sinus. CT is essential for the evaluation of osseous involvement of the cribriform plate, fovea ethmoidalis, and lamina papyracea. On MRI, ONBs usually show hypointensity relative to the gray matter on T1WI and hyperintensity relative to the gray matter on T2WI [33]. These tumors demonstrate an avid and homogeneous enhancement except for occasional areas of necrosis or hemorrhage (Figure 7). When an intracranial extension is present, the peripheral or marginal cysts are a characteristic and specific feature of ONBs [34].

**Figure 7.** Olfactory neuroblastoma of the right nasal cavity. Contrast-enhanced CT image showing a homogeneously enhanced lesion in the right olfactory cleft (arrow).

### *5.8. Malignant Melanoma*

Malignant melanoma (MM) originates from the pigment-producing cells (melanocytes) predominantly located in the skin. Sinonasal mucosal MMs account for 0.5–2% of all MMs and approximately 4% of all head and neck MMs. The age at occurrence may vary widely; the peak incidence is in the seventh decade of life. The tumor exhibits no gender predilection. The nasal cavity (including nasal septum, inferior and middle turbinates, and lateral nasal wall) is the second most common site for mucosal melanoma followed by the oral cavity. The paranasal sinuses are rarely affected; of these, the maxillary sinus is the most commonly affected. The occurrence of regional and distant metastases is relatively common (≥25% in most series) [35].

CT findings are nonspecific; however, CT is useful for the evaluation of the remodeling of the the surrounding bone or bony erosion [36]. On T1WI, sinonasal MMs that contain melanin or hemorrhage usually show iso- to hyperintensity relative to the gray matter; however, amelanotic melanoma may also show hypointensity (Figure 8). T1 shortening more often appears as a reflection of the paramagnetic effects associated with the products of hemorrhage rather than the presence of melanin [37]. On T2WI, sinonasal MMs typically show hyperintensity relative to the gray matter; however, melanotic melanomas may show iso- to hypointensity. MMs typically show a heterogeneous strong contrast enhancement owing to the rich vascular network.

**Figure 8.** Malignant melanoma of the right nasal cavity. T1-weighted image showing heterogeneously hyperintense areas within tumor (arrow).

### *5.9. Rhabdomyosarcoma*

Rhabdomyosarcoma (RMS) is a malignant mesenchymal tumor with skeletal muscle differentiation and is one of the most common pediatric soft tissue sarcomas, accounting for 3–5% of all malignancies in childhood. RMS is classified into embryonal, alveolar, pleomorphic, and spindle-cell subtypes. The mean age at diagnosis is 5–6 years, and 72–81% of patients are younger than 10 years; the reported male-to-female ratio is 1.3:1 [38]. Approximately 40% of all RMSs occur in the head and neck region; the most common sites are the orbit, nasopharynx, middle ear and mastoid, and sinonasal tracts. Embryonal RMS is the most common histopathological subtype occurring in the head and neck region and in the genitourinary system and is typically associated with a favorable prognosis.

The average diameter of head and neck RMSs (HNRMS) is 4.5 cm. Most of the HNRMSs have ill-defined margins with adjacent bony destruction and extension into the surrounding spaces [39,40]. On CT, HNRMSs appear as an isodense or slightly hypodense mass and show homogeneous enhancement on contrast-enhanced CT (Figure 9). Intratumoral calcification and hemorrhage rarely occur in HNRMS. On MRI, HNRMSs show isointensity relative to muscle on T1WI, and moderate hyperintensity relative to muscle on T2WI [39,40]. On contrast-enhanced T1WI, HNRMSs show various enhancement patterns; however, the majority of HNRMS shows moderate homogenous enhancement.

**Figure 9.** Rhabdomyosarcoma of the right nasal cavity. Contrast-enhanced CT image showing a homogeneously, slightly enhanced lesion (arrow).

### *5.10. Malignant Peripheral Nerve Sheath Tumor*

Malignant peripheral nerve sheath tumors (MPNST) are highly aggressive malignant mesenchymal tumors that usually arise from peripheral nerves or cells of the peripheral nerve sheath and show variable differentiation toward one of the cellular components of the nerve sheath. MPNSTs account for approximately 5–10% of all soft tissue sarcomas, of which only approximately 8–16% occur in the head and neck region [41]. They occur mainly in adults; the age at presentation may vary widely, although the peak incidence is in the fifth decade of life. They are commonly associated with neurofibromatosis type 1 (NF1) but can also arise through sporadic mutation [41]. The cases associated with NF1 occur in younger patients (mean patient age in the third to fourth decades). MPNSTs may occur in the sinonasal tract, nasopharynx, oral cavity, and orbit. Approximately two thirds of the cases metastasize, usually hematogenously, to the lung and bone.

On CT, MPNSTs appear as a large, hypodense soft-tissue mass with infiltration of the adjacent structures. On MRI, 51% of MPNST on T1WI and 78% of MPNST on T2WI exhibit heterogeneous signal intensity [42] (Figure 10). Compared to neurofibromas, the peripheral enhancement pattern, perilesional edema, and intratumoral cystic change are characteristic MRI findings of MPNSTs [42]. Compared to non-neurogenic malignant soft tissue tumors, the intermuscular distribution, nodular morphology with a fusiform shape, location on the course of a large nerve, and homogeneity of the signal intensity and enhancement are characteristic MRI findings of MPNSTs [43].

**Figure 10.** Malignant peripheral nerve sheath tumor of the right maxillary sinus. T2-weighted image showing a heterogeneously hypo- to hyperintense lesion (arrow).

### **6. Conclusions**

Although the radiological differentiation of sinonasal malignancies is very difficult because of the similarity of imaging findings, the tumor location, growth pattern into adjacent bone, tumor homogeneity, internal signal intensity, contrast enhancement pattern, and DWI with ADC measurement may facilitate an adequate diagnosis. CT and MRI are useful tools for pretreatment evaluation of the characterization, localization, and distribution of malignant sinonasal tumors.

**Author Contributions:** M.K. reviewed the literature and wrote the first draft. H.K. reviewed the draft and revised the manuscript. H.T., M.K., A.M., H.A., and M.M. revised parts of the manuscript. All authors read and approved the final version of the paper.

**Conflicts of Interest:** The authors declare no conflicts of interest.

### **References**


© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Case Report*
