INTRODUCTION

Ovarian tumors belong to the most insidious oncologicconditions in whole human pathology. Ovarian surface epithelialtumours are a heterogenic group of neoplasms in which a widespectrum of the clinical behaviour can be seen. Histopathologicalscope of these tumours ranges from the benign cystic tumours upto the overt malignant high-grade carcinomas. Ovarian cancer isone of the most sinister forms of female genital tract malignancy,with the lowest survival rates compared with other gynecologictumors (1). This is in part because of lack of clinical symptoms atearly stage or their unspecificity, as for example is the case withvisceral pain, in which the patient is unable to point the locationof the pain source (2, 3). The fact, that statistically benign formsof them prevail, typically presenting as a cyst or multicystic massdoes not help, since there is a high risk of either misinterpretationor simply an omission of sometimes disproportionally minutefocus of malignant lesion within grossly benign tumor. Moreover,existence of the well-known category of borderline tumors posesalmost inherent source of misinterpretation by diagnosingpathologist in spite of the formally well established guidelinesthat had been worked out to tell borderline tumor fromrespectively, bona fide malignant and entirely benign one.Therefore the need to find some objective adjuncts for thehistopathological diagnosis and prediction of the biologicalbehaviour of these tumors is enormous. This stimulated us toundertake investigations of the elemental content and distributionin ovarian tumors. Carcinogenesis of ovarian tumors is amultistep process in which varied genetic pathways can betriggered. Because of this, there is a certain amount of

controversy whether this group represents a biological continuumof increasing malignancy or rather it reflects dissimilar neoplasticpathways, that notwithstanding result in somewhat convergentprocesses. Current view is that low grade carcinomas stem froma different precursor than the high-grade ones do - benign andborderline neoplasms in case of the former and fimbrial Fallopiantube precursor in the latter (4). This raises the question of possiblemolecular or elemental differences between various ovariantumours and because of that, necessitates appropriateinvestigations. A number of studies have been published onevaluation of trace elements distributions and concentrations inanimal models of human cancers. Feldstein et al. (5) performedchemical elemental analysis on mice inoculated with the varioustypes of cancer. The experiment showed the changes in theconcentration of trace elements as a function of time frominoculation. They discovered a considerable growth of theconcentration of rubidium and a decline of the concentration ofiron as a function of time elapsed from inoculation. Changes inthe concentration of some trace elements in breast cancer tissuewere studied by Silva et al. (6). They found that theconcentrations of Ca, Fe, Cu and Zn are higher in canceroustissues compared to the control, although, the level of Fe wasfound to be greater in benign tumours than in malignant ones.Geraki et al. (7) also studied the breast cancers and noticed agreater concentration of K, Cu, Zn and Fe in the cancerous tissuescompared to the control ones. An attempt to use the informationon the trace elements distribution in a human brain tissues intumour type classification was undertaken in the study bySzczerbowska-Boruchowska et al. where multivariatediscriminant analysis (MDA) was used (8). Wandzilak et al. (9)

JOURNALOF PHYSIOLOGYAND PHARMACOLOGY 2017, 68, 5, 699-707

www.jpp.krakow.pl

L. CHMURA1, M. GRZELAK2, M. CZYZYCKI2, P. WROBEL2, M. BRZYSZCZYK, R. JACH3, D. ADAMEK 1, M. LANKOSZ2

QUANTIFICATION OF SELECTED ELEMENTS IN OVARIAN TUMOURS AND THEIR POTENTIALS AS A TISSUE CLASSIFIER

1Chair of Pathomorphology, Faculty of Medicine, Jagiellonian University, Cracow, Poland; 2Faculty of Physics and Applied Computer Science, University of Science and Technology, Cracow, Poland;

3Department of Gynecology and Oncology, Faculty of Madicine, Jagiellonian University Medical College, Cracow, Poland

Neoplastic and healthy ovarian tissues were analysed to identify the changes in the spatial distribution and concentrationof elements using synchrotron induced micro X-ray fluorescence spectroscopy. High-resolution distribution maps ofminor and trace elements were drawn. Significant amounts of elements such as P, S, Cl, K, Ca, Fe, Cu, Zn, Br and Rbwere present in all neoplastic tissues analysed. The study showed significant diversifications in elemental distributionsdepending on the structure of tissue. The efficacy of micro X-ray fluorescence spectroscopy to distinguish betweenvarious types of ovarian tumours based on the concentrations of studied elements was confirmed by multivariatediscriminant analysis. Our analysis showed that the most important elements for tissue classification are S, Cl, K, Fe,Zn, Br and Rb.

K e y w o r d s :ovarian tumours, trace elements, elemental distribution, human samples, micro X-ray fluorescence spectroscopy

utilized the X-ray fluorescence microspectroscopy to study therelationship between the amount of the trace metals in the brainand the progress of the carcinogenesis. This study showed that theX-ray fluorescence spectroscopy is a well suited methodology forthe investigation of the levels and spatial distributions ofelements in the human brain tissues. By using the metals as aguide, it was possible to distinguish, in a non-destructive way, therelatively homogeneous cancerous tissues, the structures likeblood vessels and the calcifications. The researchers not onlyshowed that it is possible to locate the tumour tissue, but they alsofound that the concentration of the trace metals could be used tosupport the assessment of the malignancy grade of brainneoplasms. It was discovered that in the high grades ofmalignancy certain trace elements such as iron and calcium wereless abundant than in normal tissue, whilst zinc behavedconversely. This means that the trace metals could be used as abiomarker to evaluate the aggressiveness of brain tumours inpatients. Al-Ebraheem et al. investigated the emerging patterns inthe distribution of the trace elements in ovarian, and breastcancer, both invasive one and in-situ (10). This study enabled toelucidate the changes in the trace elements levels in the relationto the progression from ductal carcinoma in situ to the invasiveductal carcinoma of the breast. In this work, cellular distributionof Ca, Cu, Fe and Zn has been investigated. The study shows thatthe trends in differences in the levels of Ca, Fe, Cu and Znelements for invasive ductal carcinoma and ovarian cancers aresimilar. The statistical analysis showed a substantial higher levelsof Ca, Cu and Zn concentration in the cancer tissue whencompared to the normal surrounding tissue for both the ovarianand the invasive ductal carcinoma cases. However, in thisinvestigation formalin fixed paraffin embedded tissue was used,which is not entirely reliable approach. Our previousmeasurements performed with the use of brain tissue confirmed,that formalin fixation and paraffin embedding can causedegradation of tissue and affect their elemental composition (11).Therefore in our research on ovary tumours we used freeze driedsamples. Selection of the material to the experiment relied onintraoperative fresh, unfixed material that was subjected toroutine quick histopathological assessment and on the condition,that the histopathological examination provided conclusiveresults as for the character of the tumour in the provided tissuesample. The proposed study is an attempt to answer the questionwhether the concentrations of minor- and trace elements in themalignant tissues can be used for differentiation and/orclassification (diagnosis) of ovarian tumours.

MATERIALS AND METHODS

The applied procedures were in accordance with the ethicalstandards and approved by the local ethical committee (numberof approval: KBET/179/B/2013). The samples channelled to theelemental micro-imaging were taken intraoperatively from theovarian tumours of different types. Primarily the samples wereused for provisional intraoperational diagnostic evaluation,after which a subsequent microtome sections were taken to mapthe distribution of chemical elements. Samples for the firstpurpose were cryo-sectioned to a thickness of 5 µm and stainedwith hematoxylin-eosin and then subjected to histologicalexamination to identify the type and possible malignancy of thetumour. An adjacent sample of tissue was cut to a thickness of20 µm and then placed on X-ray-transparent Ultralene foil,stretched on a polymer disc. Next the slide was dried at a –80°Ctemperature. The size of each slide mounted on the foil wasapproximately 1 cm2. Altogether 12 samples of canceroustissues and three controls we analysed during themeasurements.

Among the analysed samples, there were epithelial tumoursof different biological behaviour: benign - one sample ofmucinous cystadenoma, and two samples of endometrial cysts;borderline - one sample of atypical proliferative mucinous tumorand two samples of atypical proliferative serous tumor;malignant - one sample of endometrioid adenocarcinoma, onesample of clear cell carcinoma and two samples of high gradeserous carcinoma, as well as stromal tumors (two samples offibroma) and control tissue samples from a stroma of anadjacent, uninvolved ovary. The measurements of experimentalsections were performed at the beamline FLUO of ANKAsynchrotron light source, Karlsruhe, Germany. The beam wasmonochromatised with the use of Si (111) monochromator andthen focused with glass capillary. The final beam spot size was13 × 19 µm2 FWHM. For 2D mapping the primary photonenergy was set to 17.2 keV. The fluorescence radiation wasdetected with the use a SDD detector. The sample surface wasoriented at 45° to the beam and the detector. The surface areas ofthe sections measured from 100 × 100 µm to 500 × 500 µmdepending on the size of the structure of interest. Themeasurement time was 5 seconds per point and the pixel sizewas 15 × 20 µm2. NIST SRM 1832 and SRM 1833 referencematerials were used in order to recalculate the elementalfluorescence intensities into the elemental mass deposits per unitarea. Both the samples were analysed in 25 points with countingtime of 10 seconds and additional filtration (2 mm Al) of theprimary beam. The data was analysed with use of the AXILprogram. The measured intensities were converted into the massdeposits per unit area. The maps of elemental distribution wereexpressed in µg/cm2. To verify the statistical significance of thedifferences in the elemental composition of various tumour typesthe Kruskal-Wallis test H was applied (12). The Kruskal-Wallistest is a rank-based, non-parametric test. The null hypothesis ofeach of this tests is that all populations have identicaldistribution functions. The Kruskal-Wallis test can be used todetermine if there are statistically significant differencesbetween three or more groups of an independent variables. Thedetailed theoretical basis of the Kruskal-Wallis test is presentedelsewhere (12).

The Multiple Discriminant Analysis was applied todistinguish the ovarian tissue samples by their elemental content.For this purpose, from each mapped homogenous area the 5% ofpixels representing the highest and the lowest concentrations ofeach element were removed. From those remaining, 100 pixelswere randomly selected. In statistical analysis all the quantifiedelements, i.e. Br, Ca, Cl, Cu, Fe, K, Mn, P, Rb, S, Se, Sr, Zn wereconsidered. The mass deposits per unit area of those elementswere converted to a logarithmic scale (base 10) to make thedistribution more Gaussian-like, which is a standard procedurein such analyses. Elements that make it possible to distinguishbetween various types of cancers were identified using a forwardstepwise procedure in which elements with the highestdiscrimination power are added to the model in a stepwisemanner. All the variables were included in the final model. As aresult of this analysis, discrimination functions were obtainedwhich are linear combinations of mass deposits per unit area ofthe elements considered in the model. From all the obtainedfunctions, we selected those which offered the bestdifferentiation between samples representing different types oftumour.

RESULTS

The spectrum of characteristic X-rays excited in neoplastictissue is presented in Fig. 1. Also lower limit of detection values(LLD values) of measured elements were calculated and are

700

701

Fig. 2. SR-XRF maps of elemental distribution in benign tumour tissue and optical microscope image of tissue. Data presented in µg/cm2.

Fig. 1. A typical sum spectrum excited inovarian cancer tissue with the use ofsynchrotron radiation (beam energy 17.2keV; total acquisition time: 6000 s).

element Br Ca Cl Cu Fe K Mn P Rb S Se Sr

LLD 0.16 1.81 6.75 2.90 0.36 3.28 0.66 24.6 0.21 11.1 0.12 0.42

Table 1. Limits of detection. Limits of detection expressed in ng/cm2 of measured elements in ovarian tissue sample obtained foracquisition time; 5 s.

presented in Table 1. Significant amounts of elements such as P, S,Cl, K, Ca, Fe, Cu, Zn, Br and Rb were present in all neoplastictissues analysed. There were also small amounts of Ti, Cr, Mn, Ni,and Pb. Figs. 2, 3 and 4 show distribution maps of P, S, Cl, K, Ca,Fe, Cu, Zn in the benign tumour, borderline tumour and malignantone, respectively. The presented maps show significantdiversifications in elemental distributions depending on the

structure of tissue. It was found, that for the borderline tumour Cuis mainly located in the neoplastic cells. As mentioned before, foreach scanned area 100 pixels were selected randomly. These pixelswere used for calculations of the average surface densities ofmeasured elements in investigated tissues. In Table 2 the calculatedmean mass deposits per unit area of elements and theiruncertainties (single st. dev.) were presented. The results presented

702

Fig. 3. SR-XRF maps of elemental distribution in borderline tumour tissue and optical microscope image of tissue. Data presentedin µg/cm2.

Benign tumor Borderline

tumor

Malignant

tumor

Stromal

tumor

Control

tissue

Br 4.4 ± 1.9 5.21 ± 0.78 4.19 ± 0.61 4.59 ± 0.44 4.26 ± 0.45

Ca 540 ± 24 265 ± 27 178 ± 23 258±24 197 ± 23

Cl 7800 ± 290 5090 ± 390 4020 ± 100 6090 ± 990 5020 ± 630

Cu 2.9 ± 0.8 2.6 ± 0.3 2.5 ± 0.2 2.3 ± 0.3 2.6 ± 0.2

Fe 20.2 ± 7.7 21.5 ± 7.2 23.01 ± 2.06 26.30 ± 5.4 97 ± 69

K 5390 ± 1020 4390 ± 470 5310 ± 320 4820 ± 980 5570 ± 760

Mn 1.32 ± 0.14 1.019 ± 0.092 0.945 ± 0.066 3.9 ± 1.9 12.9 ± 10.8

P 6160 ± 470 4380 ± 290 4960 ± 390 4540 ± 680 5567 ± 807

Rb 4.36 ± 0.49 3.98 ± 0.53 5.49 ± 0.50 4.01 ± 0.79 5.12 ± 0.71

S 4650 ± 560 4510 ± 580 3270 ± 180 4270 ± 270 4140 ± 290

Se 0.658 ± 0.027 0.529 ± 0.027 0.478 ± 0.024 0.436 ± 0.049 0.498 ± 0.024

Sr 0.49 ± 0.17 0.56 ± 0.16 0.401 ± 0.099 0.319 ± 0.056 0.329 ± 0.067

Zn 44 ± 12 22.4 ± 1.8 40.8 ± 5.9 22.2 ± 5.3 34.7 ± 3.9

Table 2. The average mass deposits per unit area of elements in ng/cm2.

703

Fig. 4. SR-XRF maps of elemental distribution in malignant tissue and optical microscope image of tissue (top left photo). Datapresented in µg/cm2.

Fig. 5. Mean concentrations of Zn(ng/cm2) in various types of ovariantumours and control tissue. Error barsrepresent ± one standard deviation ofuncertainty.

Fig. 6. Mean concentrations of S(ng/cm2) in various types of ovariantumours and control tissue. Error barsrepresent ± one standard deviation ofuncertainty.

in Table 2 show significant lower concentrations of Cu and Zn inmalignant tissue. The relationships shown in Table 2, reveal thedifferences in the concentrations of elements in tissues representingvarious type of ovarian cancers and control tissue. Based on thevalues of the Kruskal-Wallis test presented in Table 3, it was found

that the elements of the greatest importance in distinguishingbetween types of ovarian tumours are Ca, Fe, Mn, S, P, Se and Zn.Graphical interpretations of the statistical evaluation are presentedin Figs. 5, 6, 7 and 8 for Zn, S, Pand Se in relation to the type ofovarian tumour. Statistically significant differences were found in

704

Fig. 7. Mean concentrations of P(ng/cm2) in various types of ovariantumours and control tissue. Error barsrepresent ± one standard deviation ofuncertainty.

Fig. 8. Mean concentrations of Se(ng/cm2) in various types of ovariantumours and control tissue. Error barsrepresent ± one standard deviation ofuncertainty.

Fig. 9. The scatter plot of observationsin the space of discriminant variablesfor different types of ovarian tumours.

Zn mass fractions between the control tissue sample and theneoplastic tissue, distinguishing separately benign, borderline,malignant and also separately stromal tumour.

For the multiple discriminant analysis concentrations ofelements are represented in the formulae by their symbols:

D1 = –2, 498K + 1, 501S + 0, 824Cl - 0, 596Fe (1)D2 = 2, 673S - 2, 952Cl + 1, 191Br + 0, 904Rb - 0, 690Zn (2)

Noteworthy is that for the first discrimination function Fe,Cl, S and K are dominant, whereas for the second - Zn, Rb, Br,Cl and S. In Fig. 9 the pixels in the maps of the studied tissuesused in MDAare represented by points. It can be noticed that thegroups of points representing the same type of tumour are clearlyseparated. It is also of note that those representing non-malignant moiety: stromal, borderline and benign tumours aregrouped in the same area of the graph and are located close toone another, whereas 'true' cancers are located on the oppositeside. It can also be noticed that the points representing controlsamples are separated from tumours.

In Fig. 10 the examples of correlations between Zn and S fordifferent type of ovarian tumours and for the control tissue areshown. It can be observed that for the control tissue (Fig. 10A),the elemental content is spatially correlated, whereas in the tissuerepresenting the cancer (Fig. 10B) there is almost no correlation.For borderline tumour (Fig. 10C) a reasonably good correlationwas obtained. However two different groups of correlations arevisible. This phenomenon could be caused by the fact, thatmeasured pixel were positioned on different types of cells.

DISCUSSION

In the MDA, the elements that played the highest role in thegeneral discrimination were: K, S, Cl, Fe, S, Br, Rb and Zn. Theexplanation that the discrimination is based on major compoundsof cytoplasmic and extracellular fluid - K,Cl can be explained onthe basis of general ion composition of cellular fluid as theseelements represent the main elemental compounds of cytosole. Aneoplasm can be viewed as a structure with a greater amount ofcytoplasm per sample area because of the abundance of cells in aproliferating tumour and an increased ratio of tumour cells tosurrounding stroma. As stroma is made mainly of fibroblasts,vasculature and extracellular matrix rich in collagen (13) and assuch thus lower on cytoplasmic fluid load, the difference inaforementioned elements emerges. In Fig. 3 this increasedconcentration of K and Cl is visible as rims of higherconcentration (brighter colours) at the edge of the borderlinetumor, a location occupied by neoplastic cells. Analogicalexplanation can be assumed in terms of Rb, a biological proxy ofpotassium with similar biochemical characteristics (14).

Of importance however is the significance of transition traceelements (Fe, Zn) for distinguishing between tumour typeswhich is in concordance with postulated role of these metals inneoplastic processes. Increased levels of iron can directlyparticipate in carcinogenesis by means of inducing oxidativestress through Fenton reaction (15). Zinc is one of the mostabundant trace metals in human cells (16) and an elementinvolved in many metabolic processes such as cell proliferation,DNA repair and defence against production of reactive oxygenspecies (17, 18). Elevated levels of zinc were a feature ofsamples of malignant tumours in this study which is inconcordance with yet another function of zinc in biologicalsystems, as this metal is also a compound of metal matrixproteinases (MMPs) - enzymes involved in digestion ofextracellular matrix. This mechanism is crucial for cancer

705

Fig. 10. Spatial correlation between Zn and S for control sample (A), cancer sample (B) and borderline sample (C) together with theirrespective correlation coefficients.

H P

Br 2.7667 0.5976

Ca 15.6908 0.0035

Cl 1.944 0.7461

Cu 1.0641 0.8999

Fe 21.1713 0.0003

K 3.044 0.5505

Mn 11.1147 0.0253

P 6.5011 0.1647

Rb 5.0443 0.2828

S 13.5364 0.0089

Se 11.3261 0.0231

Sr 4.2199 0.3771

Zn 10.4992 0.0328

H, the Kruskal-Wallis test H statistic; P, the level of significance.

Table 3. Values of the H statistic obtained by the Kruskal-Wallistest for a comparison of element contents in various types ofovarian tumors and the control sample.

progression and infiltration of surrounding tissues (19).However, given that the confidence intervals are not separatebetween the groups (Fig. 5), this result would need to beconfirmed on a larger set of samples.

The spatial correlation between Zn and S content in benigntumor tissue (Fig. 10A), in contrary to the lack of such correlationin cancer (Fig. 10B) supposedly reflects decoupling of normallysynchronized metabolic processes, what takes place in cancerouslytransformed cells (noteworthy, the dimension of pixels 15 × 20 µmfor which the data were separately acquisitioned is such that it maybe entirely contained within the cross-section of the single averagecell). Glutathione is a key protein for sulfur storage in human cells(20). It has been shown in the work of Pitynski et al. that thedistribution of gamma-glutamyl transferase, a transmembraneenzyme protein involved in glutathione metabolism normallyabundant in human cells is deranged in high grade endometrioidadenocarcinoma of the uterus (21). One may also speculate that theloss of spatial correlation between Zn and S hypotheticallyaccounting for the aforementioned 'decoupling' (of metabolicprocesses), may be related not only to intracellular processes withina particular cancer cell but, it may as well reflect a loss of propercooperation between cancer cells and their non-cancerous directcellular neighbours present in the tumour 'tissue'. Conversely it canbe an effect of qualitative change in the malignant tumorenvironment, the composition of which differs from the non-malignant counterparts, for example with an accumulation oftumour associated macrophages in progressive malignant tumours(22). Interestingly, accordingly to the findings of Winters et al. theZn content in macrophages can be significantly changed,supposedly as a response to extrinsic signalling molecules likecytokines (in particular, granulocyte macrophage colony-stimulating factor - GM-CSF) and the level of Zn in macrophagesis bound with their antimicrobial activity (23). Though technicallythis report does not refer to tumor-associated macrophages,however it unequivocally provides an evidence of dynamicchanges of Zn content in macrophages and hence one mayrightfully speculate that also in tumor/cancer conditions, thecontent of this element in macrophages may change. As a result,the relative change between S and Zn may follow in tumor-associated macrophages in response to so far unknown factors evenchanging 'gross content' of Zn in tumor tissue as our results show.

In MDA plot, benign tumor represent a separate cluster,whereas for the rest of groups such good separation was notobtained. This is probably because of the unique characteristicsof tumor growth pattern in comparison with other types includedin the analysis, as benign surface epithelial neoplasms of ovarygrow in a manner of cystic spaces lined with a thin layer of cells.This can result in that the actual points selected at random fromthe mapped areas will, with higher probability, come from anadjacent area of ovarian stroma or even from a fluid filled cysticcavity than from 'true' cancer cells. This means, that this groupwould rather deviate from others as the latter represent morecomplex or solid growth pattern.

The obtained results show presence of the differences in theconcentration of trace elements, however direct role of each ofthem remains elusive. The fact of variability between differentneoplastic groups has a potential to be applied as means todistinguish various histopathological types of cancer. For thisapplication however, employing synchrotron radiation facilitiesis highly problematic if not impossible at current state of the artas turnover times would not be acceptable. For practicalapplication purposes, a more simplistic model of analysis basedon unique characteristics of tumor would be appreciated. On thisbasis it would be possible to use a more simple setup whichwould utilize micro X-ray lamp.

As sample load included in the study was low, the presentedanalysis can be viewed as merely the first step into providing a

deeper insight into the intricacies of trace and other elementsdistribution in cancerous ovarian tissue.

This work assessed the possibility to distinguish basicovarian tumour types with regard of their elementalcomposition. To take into account all the quantified elementssimultaneously, multiple discriminant analysis was used.Statistical analysis performed on the maps obtained using theXRF technique showed that it was possible to distinguishbetween essential types of ovarian tumours on the basis ofconcentrations of elements measured in homogenous areas ofthe samples. Of greatest significance for this distinction were S,Cl, K, Fe, Br, Rb and Zn. The results presented here show thatthe XRF method has much greater potential for theclassification of samples due to its higher sensitivity to high-Zelements. The results could also be another step towardsdeveloping methods for identification of tumours and especiallyfor separation of benign from malignant and borderline forms ofthe ovarian tumors that could be complementary tohistopathological examination.

Acknowledgements: We acknowledge the Synchrotron LightSource ANKA for access to the FLUO beamline. The researchleading to these results has received funding from the EuropeanCommunity's Seventh Framework Program CALIPSO, thePolish Ministry of Science and Higher Education grants:W116/IAEA/2014, W57/IAEA/2015, W22/IAEA/2017 andIAEA Research Contract No. 18199.

Conflict of interests: None declared.

REFERENCES

1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CACancer J Clin 2017; 67: 7-30.

2. Smith LH. Early clinical detection of ovarian cancer: areview of the evidence. Expert Rev Anticancer Ther 2006; 6:1045-152.

3. Leppert W, Zajaczkowska R, Wordliczek J, Dobrogowski J,Woron J, Krzakowski M. Pathophysiology and clinicalcharacteristics of pain in most common locations in cancerpatients. J Physiol Pharmacol 2016; 67: 787-799.

4. Kurman RJ, Shih IM. The origin and pathogenesis ofepithelial ovarian cancer: a proposed unifying theory. Am JSurg Pathol 2010; 34: 433-443.

5. Feldstein H, Cohen Y, Shenberg C, et al. Comparisonbetween levels of trace elements in normal and cancerinoculated mice by XRF and PIXE. Biol Trace Elem Res1998; 61: 169-180.

6. Silva MP, Silva DM, Conceicao AL, Ribeiro-Silva A, PolettiME. Role of Ca, Fe, Cu and Zn in breast cancer: study by X-ray fluorescence techniques and immunohistochemicalanalysis, XRay Spectrom 2013; 42: 303-311.

7. Geraki K, Farquharson MJ, Bradley DA, Hugtenburg RP. Asynchrotron XRF study on trace elements and potassium inbreast tissue. Nucl Instrum Methods Phys Res B 2004; 213:564-568.

8. Szczerbowska-Boruchowska M, Lankosz M, Adamek D. Firststep toward the 'fingerprinting'of brain tumours based onsynchrotron radiation X-ray fluorescence and multiplediscriminant analysis. J Biol Inorg Chem 2011; 16: 1217-1226.

9. Wandzilak A, Czyzycki M, Radwanska E, Adamek D, GerakiK, Lankosz M. X-ray fluorescence study of the concentrationof selected trace and minor elements in human brain tumours.Spectrochim Acta Part B At Spectrosc 2015; 114: 52-57.

10. Al-Ebraheem A, Dao E, Geraki K, Farquharson MJ.Emerging patterns in the distribution of trace elements in

706

ovarian, invasive and in-situ breast cancer. J Phys Conf Ser2014; 499: 012014. doi:10.1088/1742-6596/499/1/012014

11. Chwiej J, Szczerbowska-Boruchowska M, Lankosz M.,Setkowicz Z. Preparation of tissue samples for X-rayfluorescence microscopy. Spectrochim Acta Part B AtSpectrosc 2005; 60: 1531-1537.

12. Corder GW, Foreman DI. Nonparametric Statistics: A Step-by-Step Approach. Hoboken, New Jersey, John Wiley &Sons 2014, pp. 117-138.

13. Bremnes RM, Donnem T, Al-Saad S, et al. The role of tumorstroma in cancer progression and prognosis: emphasis oncarcinoma-associated fibroblasts and non-small cell lungcancer. J Thorac Oncol 2011; 6: 209-217.

14. Relman AS. The physiological behavior of rubidium andcesium in relation to that of potassium. Yale J Biol Med1956; 29: 248-262.

15. Toyokuni S. Role of iron in carcinogenesis: cancer as aferrotoxic disease. Cancer Sci 2009; 100: 9-16.

16. Emsley J. Nature's Building Blocks: An A-Z Guide to theElements. Oxford University Press 2011, p. 621.

17. Bray TM, Bettger WJ. The physiological role of zinc as anantioxidant. Free Radic Biol Med 1990; 8: 281-291.

18. Powell SR. The antioxidant properties of zinc. J Nutr 2000;130 (Suppl. 5S): 1447S-1454S.

19. Turunen SP, Tatti-Bugaeva O, Lehti K. Membrane-typematrix metalloproteases as diverse effectors of cancerprogression. Biochim Biophys Acta 2017; 1864: 1974-1988.

20. Nimni ME, Han B, Cordoba F. Are we getting enough sulfurin our diet? Nutr Metab (Lond) 2007; 4: 24.doi.org/10.1186/1743-7075-4-24

21. Pitynski K, Ozimek T, Galuszka N, et al. Association of theimmunohistochemical detection of gamma-glutamyltransferase expression with clinicopathological findings inpostmenopausal women with endometrioid adenocarcinomaof the uterus. J Physiol Pharmacol 2016; 67: 395-402.

22. Rybicka A, Eyileten C, Taciak B, et al. Tumour-associatedmacrophages influence canine mammary cancer stem-likecells enhancing their pro-angiogenic properties. J PhysiolPharmacol 2016; 67: 491-500.

23. Winters MS, Chan Q, Caruso JA, Deepe GS Jr. Metallomicanalysis of macrophages infected with Histoplasmacapsulatum reveals a fundamental role for zinc in hostdefenses. J Infect Dis 2010; 202: 1136-1145.

R e c e i v e d :May 30, 2017A c c e p t e d :September 25, 2017

Author's address: Dr. Lukasz P. Chmura, Chair ofPathomorphology, Faculty of Medicine, Jagiellonian University,16, Grzegorzecka Street, 31-531 Cracow, Poland.E-mail: l.chmura@uj.edu.pl

707