ZYXIN EXPRESSION LEVELS IN NON-SMALL CELL LUNG CANCER PATIENTS EKSPRESIJA ZIKSINA KOD OBOLELIH OD NESITNOĆELIJSKOG KARCINOMA PLUĆA Authors

Background/Aim. Non-small cell lung cancer (NSCLC) is the most common cause of cancerrelated mortality worldwide. Early detection represents one of the most promising approaches to reduce lung cancer mortality. Zyxin is a member of the focal adhesion protein family, recently identified as a potential early diagnostic marker for NSCLC. The aim of this study was to evaluate zyxin expression levels in NSCLC patients and compare serum zyxin expression profiles between early and advanced clinical stages, different histological subtypes and histological grades. Methods. Blood samples were obtained from 90 patients diagnosed with NSCLC in all clinical stages and 30 patients without the clinical and radiological findings and previous history of malignancy. For the quantitative determination of human zyxin (ZYX) concentrations in serum we used enzyme-linked immunoadsorbent assay (ELISA). Results. Zyxin protein exhibited higher plasma levels in NSCLC patients as compared to the control samples with exceptionally significant difference (p=0.00). The ROC curve demonstrated a high specificity with AUC=0.912. There were no statistically significant differences in the zyxin values between two most common NSCLC types, adenocarcinoma and squamous cell carcinoma (p=0.758). There were no statistically significant differences in the zyxin values between different clinical stages (p=0.518). Sample of only 3 patients had well-differentiated tumor, no useful data may be extracted from it. There were no statistically significant differences in the zyxin values between patients with moderately differentiated tumor and poorly differentiated tumor (p=0.48). Conclusion. We found that zyxin was overexpressed in NSCLC, but its expression level was not closely correlated with tumor size and advanced TNM stage. Our results suggest that zyxin exhibited potential as an early diagnostic plasma-based tumor marker for non-small cell lung cancer, with the same importance for adenocarcinoma and squamous cell carcinoma.


Introduction
Lung cancer is the leading cause of cancer deaths worldwide.The incidence is approximately 14% in both genders (second after prostate cancer in men and breast cancer in women) 1 .Every year, lung cancer causes more than 1.7 million deaths, more than breast, colon and prostate cancers combined.Lung cancer is classified into two major types: small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC).NSCLC is divided into adenocarcinoma (approximately 63%), squamous cell carcinoma (approximately 30%), and large-cell carcinoma (approximately 7%) subtypes, accounting for approximately 85% of all new lung cancer cases. 2 Five-year survival rate of NSCLC still remains < 20% 1 .Most patients are diagnosed in older age (≅65 years) and in late-stage (IIIB-IV) where surgical resection is not a standard procedure anymore, according to the guidelines of the American Joint Committee on Cancer (AJCC) showing low overall survival rates at 5 years (5% for IIIB and 1% for IV stages) 3 .However, detection at an earlier stage and treatment by immediate resection are the cornerstones of reducing NSCLC death rates.
Despite the tremendous efforts made to discover blood-based tests for the early diagnosis during the past decades, no tumor markers are available with selectivity to effectively diagnose lung cancer.The most widely used blood-based tumor marker screening includes carcinoembrionic antigen (CEA), cytokeratin 19 fragment (CYFRA 21-1), squamous cell carcinoma antigen (SCCA), and neuron-specific enolase (NSE), without the evidence for their significance in the early diagnosis of lung cancer 4,5 .
Over the past decade, proteomic analysis has become the main tool for investigation of tumor biology.The goal of proteomics is to characterize proteins by evaluation of their expression, functions and interactions, and also may provide information about posttranslational modifications of proteins and evaluate their value as specific disease biomarkers 6 .Any biomarker is defined as a specific that is objectively measured and evaluated as an indicator of normal physiological processes, pathogenic processes and diseases or pharmacological responses to a specified therapeutic intervention 7 .Many studies reported elevation of serum haptoglobin (HP) in NSCLC patients 8,9 , elevation of serum amyloid alpha (SAA) 10,11 and tissue metalloproteinase inhibitor 2 (TIMP2) 12 , and reduction of pigment epithelium-derived factor (PEDF) in the pleural effusion and the serum samples 13 , in comparison to controls.Higher levels of leucine-rich alpha-2-glycoprotein (LRG1) were found in urine samples of cancer patients in comparison to healthy subjects 14 , high level of gelsolin expression was significantly associated with death risk of NSCLC patients 15 .Zyxin (ZYX; Uniprot ID, Q15942) showed potential to be used for early diagnosis of NSCLC.Analysis of zyxin values at the different clinical stages demonstrated that the levels of this peptide were alredy elevated at early stages of NSCLC 16 .
Zyxin is a zinc-binding phosphoprotein known as a member of the focal adhesion protein family.In normal cells zyxin is involved in cell adhesion, cytoskeleton remodeling 17 , stress fibers self-monitoring and repair in response to mechanical stress 18 .But, during mitosis, zyxin also acts as a participant in mitotic control by forming a complex with h-warts/LATS1 on the mitotic apparatus 19 .Zyxin has been already reported as being associated with tumorigenesis.The role of zyxin as a key player in the epithelial-mesenchymal transition (EMT) mechanism 20 and its association to lung cancer as a down regulator of TGF-β inducing cell motility 21 has been recently discussed.Zyxin expression correlates with cancer cell lines with higher malignancy, its activation may play a critical role in regulating Yap activation during tumourigenesis 22 .Upregulation of zyxin in hepatocellular carcinoma had been previously reported 23 , a peptide fragment apparently derived from truncated zyxin has been identified in serum samples from colorectal cancer patients 24 .The expression level of zyxin corresponding to tumorigenesis or tumor mass in the human body has been quite controversial 20 .
The purpose of this study is to evaluate zyxin expression levels in NSCLC patients and compare serum zyxin expression profiles between different histological subtypes and histological grades.

Methods
A total of 120 patients were recruited from the Clinic for Lung Diseases, Clinical Center Niš, Serbia, between October 2015 and August 2017.Blood samples were obtained from 90 patients diagnosed with NSCLC in all stages and without prior history of other cancers, including adenocarcinoma and squamous cell carcinoma, and 30 patients without the clinical and radiological findings and previous history of malignancy.None of the patients received chemotherapy, radiotherapy, hormone therapy, or other related antitumor therapies prior recruiting.
The NSCLC patients were classified into clinical disease stages I (n=9), II (n=12), III (n=30) and IV (n=39) according to the 7th edition of the American Joint Committee on Cancer TNM staging system 3 .Blood samples were taken prior to surgery for stages I-IIIA and prior to treatment for advanced stage NSCLC patients using serum separator tube.Sera were allowed to clot for two hours at room temperature and then centrifuged at 1000 × g for 15 minutes.Immediately following centrifugation, all specimens were stored at −80°C until being analyzed.
For the quantitative determination of human zyxin (ZYX) concentrations in serum we used Cusabio Human Zyxin (ZYX) ELISA Kit, Catalog Number CSB-EL027165HU (Cusabio Technology LLC, 8400 Baltimore Avenue，Room 332 College Park, MD 20740, USA).This assay employs the quantitative sandwich enzyme immunoassay technique.Antibody specific for zyxin has been pre-coated onto a microplate.Standards and samples are pipetted into the wells and any zyxin present is bound by the immobilized antibody.After removing any unbound substances, a biotin-conjugated antibody specific for zyxin is added to the wells.After washing, avidin conjugated Horseradish Peroxidase (HRP) is added to the wells.Following a wash to remove any unbound avidin-enzyme reagent, a substrate solution is added to the wells and color develops in proportion to the amount of zyxin bound in the initial step.The color development is stopped and the intensity of the color is measured.We created a standard curve by reducing the data using computer software "Curve Expert" capable of generating a four parameter logistic (4-PL) curve-fit.The minimum detectable dose of human zyxin is less than 5.8 pg/ml.Detection range between standards is 23.5 pg/ml -1500 pg/ml.
This study was approved by the Research Ethics Committee of the Clinical Center Niš.The informed written consents were collected from all eligible patients and the entire study was performed based on the Declaration of Helsinki.

Introduction to the data and statistical analysis
The data tracked numerous factors, but the ones relevant for this analysis are merely a dichotomous categorical variable and a continuous variable, specifically whether the patient is diagnosed with lung carcinoma, and what the patient's serum level of zyxin is in pg/ml.These two will suffice to establish a link between carcinoma diagnosis and zyxin level.A sample of the data is presented in Table 1.
Once the descriptive statistics have been computed the data will be analyzed using simple group-mean comparison.In point of fact, the analysis will be conducted using between-groups statistical methods, with the groups being determined using the dichotomous categorical variable Dg_Ca representing carcinoma diagnosis.The analysis has been conducted using the R programming language, version 3.3.1 built for the x64 processor architecture, using the following packages: WRS2 25 , effsize 26 , and pastecs 27 .

Results
The 90 blood samples were collected from patients who have been diagnosed as NSCLC, the average age was 63.56 ± 6.344 years, the majority of patients were males and smokers (Table 1).The blood samples from 30 control subjects with other pulmonary diseases were collected as controls by matching their age (62.43 ± 9.380 years), sex, smoking history and their duration and intensity.Patient demographics and clinical profiles are presented in Table 2.

Descriptive statistics and assumptions of general linear models
Taken as a whole, the data have an n of 120, and the mean of the variable measuring zyxin is 426.6800277, with a median of 426.2166666 and a of 249.0966559.The large relative value of the standard deviation compared to the mean casts serious doubts on the assumption of globally normal data, and the histogram in Figure 1, shows that the distribution is likely heavytailed.
However, more crucial information for our purposes is the analysis of the data as split into groups based on carcinoma diagnosis.To simplify terminology, henceforth the group without the carcinoma diagnosis will be referred to as 'control' and the group with the carcinoma diagnosis will be referred to as 'effect.'With that said, the descriptive statistics of the control group show that the n is 30 and that the mean is 168.0501111, while the median is 127.3333334 and the is 155.5513055.The distribution of the value with the control group can be seen on the histogram in Figure 2.
Despite promising mean/median results, this distribution is not normal which the QQ plot in Figure 3 shows.The s-shape to the curve indicates heavy skew which is also evident in the histogram.Similar results are found by applying standard normality tests with the Shapiro-Wilk test allowing us to reject the of normality with a p-value of 0.0015193.
The descriptive statistics of the effect group show that the n is 90 and that the mean is 512.8899999 while the median is 471.7666667 and the is 212.1738935.The distribution of the value with the effect group can be seen on the histogram in Figure 4.
Despite promising mean/median results, this distribution is not normal which the QQ plot in Figure 5 shows.The deviation from the angle to the curve indicates an issue with the kurtosis.This is possible to see by examinging the shape of the histogram suggesting a platykurtic distribution.Similar results are found by applying standard normality tests with the Shapiro-Wilk test allowing us to reject the of normality with a p-value of 2.3073684*10 -5 .
The groups have unequal sizes, unequal variances, and heavily violate the assumption of normality.With the effect size being what it is, we can confirm that the requirements of the General Linear Model (of which the t-test, normally used in situations like this, is an example) are not met, and, indeed, the requirements of parametric models in general are not met.However, the difference between the means of these two groups is large: more than double the size, in fact.This means that the effect we are looking for is very large, and thus, we should be able to identify it with great specificity if we apply a statistical technique with sufficient power and sufficiently lax parametric requirements.

Hypothesis testing on difference in Zyxine level between NSCLC patients and control subjects
We are dealing with two groups of which one is platykurtic and the other is heavily skewed.Assuming their distributions are equal may influence our results.The alternative is to apply a hybrid approach in which we deal with each of our issues with GLM assumptions in turn by employing a different solution.
Unequal variances are most simply resolved using the Welch modification of the t-test 28 .This test is a modification of the familiar Student's T-test (and is commonly employed in its stead in various libraries of statistical software) and pools variances in both populations thus producing an altered test statistic and a modified measure of degrees of freedom The remainder of the test is the same.To successfully combine a test that deals with heterogeneity of variances and with a violation of the assumption of normality, a combination of the Welch test and the Yuen modification 29 is necessary.Yuen's approach re-creates the Welch test using, instead of means, trimmed means, i.e., means with outlier values on the edges of the distribution clipped to some pre-determined value, generally expressed as a proportion in percent.This level is expressed as .However, Wilcox 30 suggests that if the group sample sizes are unequal, which is true in our case, a more robust bootstrapped version of the Yuen-Welch is employed.This version of the test uses resampling techniques to estimate the confidence interval for the critical value of the test statistic.This helps reduce the probability of Type I error to the nominal level.The procedure, roughly, proceeds by first computing the trimmed means of the sample and Yuen's estimate of the Squared Standard Errors, like so: Then, for the j-th group analyzed randomly re-sample with replacement from the available data set observations.Using the samples generated by this Monte Carlo approach compute the same values as the initial Yuen approach and label them and .Then calculate the following: This value represents an estimation of the distribution of Repeat the preceding steps generating a sequence of values (for our test we selected that this number be ).Then sort these values in ascending order.Let represent the T-value occupying the i-th place in the sorted array with . Compute and round it to the nearest integer and let .
The confidence interval of (being the true mean of the difference in groups) is We use the reference implementation in the WRS2 package with 2000 resampling steps and a trimming percentage of .The result of this analysis shows a p-value of 0, indicating a level to small for the computer to measure, with a test statistic of 7.3819, and a 95% confidence interval for the difference in means of (247.1077,437.9487).This is an exceptionally significant result, showing that we can confidently reject the of the means of the two groups being equal.There is certainly a difference with zyxin having a significantly larger value in the group diagnosed with carcinoma as compared to the control group of patients without the carcinoma diagnosis.Higher plasma levels in NSCLC patients as compared to the control samples are illustrated in Figure 6.
The remaining question is how large the effect is.The traditional approach is to employ Cohen's d effect measure for this purpose, and doing so yields a value of 1.7263458 which corresponds according to Sawilowsky 31 to a very large effect size.However, Cohen's d depends on the same parametric assumptions our data violate.Therefore in accordance to Wilcox and Tian 32 , we employ the explanatory measure of effect size, which equals 0.889821 and with a 95% confidence interval of 0.8118965 to 0.9635671.The authors of the approach categorize a value of 0.5 as a large effect, therefore, we are led to assume that our value represents a very large effect in line with Cohen's d results.
The ROC curve generated using zyxin values demonstrated a high specificity toward NSCLC with AUC = 0.912 as shown in Figure 7.

Hypothesis testing on difference in Zyxin level between different histological types of NSCLC
Using the exact method outlined above, it was possible to test for a difference in zyxin levels between groups with one of two most common NSCLC types, adenocarcinoma and squamous cell carcinoma.The analysis shows a test statistic of 0.3114692 with an associated pvalue of 0.758, showing a difference that is not statistically significant.In practical terms, the actual measured means are 522.1514and 502.3056, showing the sort of effect only a very large sample size might be able to prove statistically significant.

Hypothesis testing on difference in Zyxin level between clinical disease stages
Clinical disease stages of NSCLC delineate four separate groups, demanding a slight change in approach, specifically, instead of using a modified T-test to determine a difference in mean between groups, a robust ANOVA-equivalent is used instead, one which tests the hypothesis of equal trimmed means with a and a B=2000 without demanding assumptions of normality or heteroscedasticity.Performing this test produces a test statistic 0.8133 and a corresponding p-value of 0.518 with an effect size estimate of 0.29.Actual measured means are 443.10 for stage I, 562.38 for stage II, 439.93 for stage III, and 470.10 for stage IV.The difference is not large in absolute terms, first, indicating that a larger sample size is required, and second it is not statistically significant, as indicated by the p-value.Figure 8 shows a scatter plot of zyxin values for the different clinical stages of NSCLC.

Hypothesis testing on difference in Zyxin level between histological grades
Given the nature of histological grades, the decision was made to test only grades 2 and 3, since with applying to the grade of 1, no useful data may be extracted from it (n=3).Using the two-group comparison solution described above, it was possible to test for a difference in zyxin levels between groups with grades of two and three.The analysis shows a test statistic of -0.646 with an associated p-value of 0.4845, showing a difference that is not statistically significant.In practical terms, the actual measured means are 522.3604and 546.1500, showing the sort of effect only a very large sample size might be able to prove statistically significant.

Discussion
Low-dose computed tomography screening reduces lung cancer-related mortality, at least for subjects fulfilling National Lung Screening Trial (NLST) inclusion criteria 33 or under US Preventive Services Task Force (USPSTF) recommendations 34 .The use of lung cancer predictor models could help defining the subjects with higher risks 35 .In the near future, ongoing research on lung cancer biomarkers could increase accuracy of lung cancer low-dose CT screening.Currently, there are no validated biomarkers for early lung cancer detection.
Kim YJ et al previously demonstrated that the zyxin levels were already elevated at early stages of NSCLC.This study applied highly multiplexed liquid chromatography-selected reaction monitoring (LC-SRM) assay to verify the biomarker candidates in plasma samples for lung cancer.Zyxin was identified as a potential early diagnostic marker for NSCLC 16 .
The aim of our study is to evaluate zyxin expression levels in NSCLC patients and compare serum zyxin expression profiles between early and advanced clinical stages, different histological subtypes and histological grades.
In this manuscript, we have demonstrated that zyxin was overexpressed in NSCLC, but its expression level was not closely correlated with tumor size and advanced TNM stage.Zyxin protein exhibited higher plasma levels in NSCLC patients as compared to the control samples with exceptionally significant difference.
The NSCLC group was comprised of plasma samples of patients at two different histological types of the disease.Analysis of the zyxin values at the different histological types of NSCLC demonstrated equal levels for adenocarcinoma and squamous cell carcinoma.There were no statistically significant differences in the zyxin values between two most common NSCLC types.To the best of our knowledge, no studies have investigated the possibility of such differences.
Analysis of the ELISA zyxin values at the different clinical stages of the disaese showed significant elevation in the plasma of NSCLC patients already at early stages.This result was consistent with the results from the previous report 16 .There were no statistically significant differences in the zyxin values between different clinical stages.
A recent study by Ma B et al showed that zyxin expression correlates with cancer cell lines with higher malignancy.Zyxin is upregulated in human breast cancer and positively correlates with histological stages and metastasis 22 .We have not established correlation between degree of differentiation of a tumor and zyxin level due to lack of data for histological grade for 38 (42.2%) patients.Sample of only 3 patients had well-differentiated tumor, no useful data may be extracted from it.There were no statistically significant differences in the zyxin values between patients with moderately differentiated tumor and poorly differentiated tumor.
In conclusion, our results suggest that zyxin fulfilled the criteria for a potential early diagnostic plasma-based tumor marker for non-small cell lung cancer, with the same importance for adenocarcinoma and squamous cell carcinoma.Early detection represents a very promising approach to reduce lung cancer mortality.The results of these analyses give us reason to be hopeful, considering high ELISA zyxin values at clinical stages I and II.As one can anticipate, over the coming decade, effective biomarkers in combination with low-dose computed tomography may provide effective tools for the non-small cell lung cancer early detection and improve survival rates in these patients.x -mean; SDstandard deviation.

Table 1 .
A sample of the data under observation

Table 2 .
Patient demographics and clinical profiles Histological type of NSCLC (World Health Organization -WHO histological classification of tumors of the lung) 2 **Disease stage (7th Edition of the tumor, node, metastasis (TNM) classification of malignant tumors) 3 *** Histological grade (The 2015 World Health Organization Classification of Lung Tumors) 2 *