1. Introduction
The research for elaborating efficient ovarian cancer (OC) diagnostic tools has been ongoing for decades. At present, no screening method is available and the disease has a highly unfavorable prognosis, mainly due to the fact that over 70% of the patients are diagnosed in late stages, i.e., stage III and IV according to the International Federation of Gynaecology and Obstetrics (FIGO). Early and specific diagnosis is essential to improve the treatment outcome, because five year survival rates for FIGO stage I reach 90%, compared to about 30% for advanced disease (FIGO stage III–IV) [
1].
One of the challenges in OC diagnosis is the correct differentiation of ovarian tumors noticed on routine transvaginal ultrasound examination. Proper preoperative risk-of-malignancy assessment is very important for making clinical decisions and treatment planning. Low-risk tumors can be followed up and re-assessed by imaging methods after a certain period of time or operated conservatively, ensuring fertility-sparing, and in a less invasive way (e.g., unilateral laparoscopic tumorectomy). In the case of high-risk lesions, additional examinations may be needed preoperatively to adequately plan the surgery and prepare the patient (computed tomography scan, colonoscopy, reservation of appropriate time for surgery, ensuring the availability of intra-operative tissue examination to assess the type of tumor during the surgery, ensuring adequately trained staff, etc.). To date, histological examination of the resected tissue still remains the gold standard. This approach results in unnecessary surgical procedures that, if a reliable non-invasive diagnostic method existed, could be avoided, because over 90% of ovarian masses detected in pre-menopausal women and up to 60% of those in post-menopausal women are benign [
2]. Moreover, correct pre-operative diagnosis of OC enables adequate referral of the patient to specialized gynecologic oncology centers where evaluation by an interdisciplinary tumor board and optimal debulking surgery is possible. Treating women with OC in specialized centers is crucial to ensure proper management and was proved to significantly improve the prognosis [
3].
Some clinical multivariate diagnostic models used in ovarian tumor differential diagnosis were reported to be quite efficient, for example the ADNEX model (The Assessment of Different NEoplasias in the adneXa), based on ultrasound features and clinical data, was reported to reach an area under the receiver operating characteristic (AUC of ROC) curves as high as 0.954 [
4]. However, despite the excellent performance, its clinical application is highly limited due to the need for highly-trained medical staff and modern equipment to perform high-quality ultrasound assessment of ovarian tumors and record the required features for the model. In addition, some ovarian benign tumors pose a particular diagnostic challenge, commonly presenting ultrasound features typical for malignant lesions [
5]. For this reason, biomarker research is more likely to provide accessible and ready-to-use diagnostic methods.
Metabolomic profiling has recently become a highly promising target in the search for non-invasive cancer diagnostic methods. Metabolome is defined as a complete set of small molecules within a biological sample. Therefore, it is a direct reflection of the current processes in the organism and is altered by pathological conditions such as carcinogenesis. We have previously showed that the serum free amino acid (SFAA) profiles are altered in ovarian cancer patients [
6] and investigated the role of amino acid profiling in screening for OC. The current study was designed to investigate the role of amino acid profiling in preoperative differential diagnosis of ovarian tumors. For this purpose, the SFAA profiles of OC, borderline ovarian tumors, and benign ovarian tumors (BOTs) were analyzed, differentiating amino acids between the groups were selected, and their performance in differential diagnosis of ovarian tumors assessed. To the best of our knowledge, this is the first study of ovarian cancer that analyzes such a wide spectrum of the SFAA profile in differential diagnosis of ovarian tumors.
4. Discussion
Differential diagnosis of ovarian tumors is an important diagnostic step enabling adequate qualification for surgical management of the patients. Because the ovaries are relatively inaccessible for a preoperative biopsy, which is also contraindicated due to the risk of iatrogenic rupture of the tumor capsule, resulting in the spread of the cancer (“surgical spill”) in the case of malignancy, the markers should ideally be obtainable from an accessible body fluid, such as blood, urine, or saliva. In recent years, thanks to technological advances, metabolomics has emerged as a promising method of searching for new OC biomarkers.
Several studies confirmed that the plasma/serum free amino acids (PFAA/SFAA) profile is significantly altered in cancer patients, e.g., lung, gastric, colorectal, breast, renal, prostate, and endometrial cancers [
10,
11,
12,
13,
14], and noted that some differences reflected the metabolic changes common to many cancers, whereas others were specific to each type of cancer [
10]. In a paper by Miyagi et al. that investigates the PFAA profiles in five types of cancer (lung, gastric, colorectal, breast, prostate), it is suggested that a decrease in glutamine, histidine, and tryptophan and an increase in proline and ornithine might reflect the metabolic changes common to all cancers [
10]. Although we indeed observe changes in the levels of glutamine, histidine, and tryptophan, the expression of proline and ornithine was not altered in our cohort. Because the PFAA profile can also differ between the early and late stages of cancer and between subtypes of cancer, we performed a detailed analysis of the PFAA profiles in various clinical subgroups.
Our results identified histidine as the most effective OC marker in almost all analyzed subgroups. Of particular value, its performance did not drop for detection of early stage cancer (AUC of 0.786 and 0.788 for early and late stage OC, respectively). Moreover, it obtained similar results in the comparison between type I or type II OC with BOT and did not differ significantly between the two OC types. Therefore, it could be considered as a universal OC biomarker and should be subject to further research. Histidine was closely followed by tryptophan, which obtained high AUC values, especially in advanced stage and high-grade (type II) OC. The depletion of those two amino acids in OC patients is in line with other studies—see
Table 5. Although, as mentioned above, the changes in histidine and tryptophan may be observed in other cancers, this study focused on amino acid profiling for distinguishing benign and malignant ovarian tumors, i.e., the situation in which a pathology in the ovaries is already detected by imaging methods. Nevertheless, the fact that histidine and tryptophan levels are decreased in other cancers is likely to negatively affect their specificity for detecting OC by increasing the number of false positive results in patients whose lesion in the ovary is in fact benign but there is a concomitant malignancy of a different organ.
The results of our study also indicate that citrulline could be considered as a type II OC marker (particularly as it did not reach statistical significance in type I OC vs. BOT analysis). Additionally, cystine could be distinguished as a potential type I OC marker because it was one of the few amino acids that reached statistical significance in the type II vs. BOT analysis. However, the performance of these two amino acids may not be sufficient for clinical use.
Our findings correspond with the results of a paper on high-grade serous OC (equivalent to type II OC in our study) using targeted metabolomics, which reports a decreased serum concentrations of five amino acids (histidine, lysine, threonine, tryptophan, citrulline) compared with healthy controls that also correlated with shorter overall survival of cancer patients [
18]. The levels of four of these amino acids (histidine, threonine, tryptophan, citrulline) were decreased in OC patients in our analysis and three (histidine, tryptophan, citrulline) were decreased in type II OC, although it should be noted that the comparison was between OC and BOT (not healthy controls). Other important findings of the above-mentioned study are that the levels of amino acids identified as significant were similar in serum, ascites fluid, and tumor tissue, and that they were positively correlated with the tumor load (i.e., recovered to concentrations typical of healthy patients after initiation of anti-cancer treatment). The authors conclude that this suggests that the depletion of certain amino acids in serum is a direct effect of tumor metabolism [
18].
A systematic review of metabolomic studies in OC was published recently [
23]. It concluded that the most frequently reported amino acid alterations in OC were the decreased levels of histidine, citrulline, alanine, and methionine [
23]. Other studies that analyzed serum samples identified altered levels of several amino acids in OC patients. These results are summarized in
Table 5. There are a lot of discrepancies in the amino acids identified as differential and some studies even reveal an opposite trend of a specific metabolite (e.g., alanine, threonine). This might be due to the adoption of different mass spectrometry-based analytic methods to identify those metabolites and different study design, especially regarding control groups. Moreover, all cited studies were based on global rather than targeted metabolomic profiling techniques in which amino acids were only a small proportion of the investigated substances, whereas our study is unique in that it focused purely on SFAA profile. Notwithstanding different study methods, the results are coherent for histidine and tryptophan, which suggests that their levels are strongly affected by OC development.
A study by Hilvo et al. [
16] additionally compared the results obtained from serum samples with matching tumor tissue samples and confirmed a linear correlation of diagnostically relevant biomarkers between serum and tumor tissue. These findings support the hypothesis that relevant metabolites originate from the tumor rather than depend on other metabolic processes in the body.
It is not clear, however, the extent to which the studies based on plasma analysis can be compared with our research in which serum samples were collected. Serum is the liquid fraction of whole blood obtained after the blood is allowed to clot and centrifuged. Plasma is obtained when whole blood is collected in tubes treated with an anticoagulant and then centrifuged to remove blood cells. Surprisingly, perhaps, a study comparing amino acid profiles in both types of blood samples revealed remarkable differences in the PFAA and SFAA profiles [
12]. In general, the amino acid concentrations were, on average, 40% lower in plasma than in serum, although the level of variation and the direction of changes varied for each individual amino acid. Nevertheless, significant differences were observed in both profiles (SFAA and PFAA) between cancer patients (clear cell renal cancer) and healthy controls, and in serum a decreased level of histidine—the same as in our study—was identified as the most effective cancer marker [
12].
In addition to the SFAA profile, in our research two clinically used OC biomarkers, CA125 and HE4, were additionally analyzed. Their generally high performance in differential diagnosis of ovarian tumors was also confirmed by our analyses. As expected, their diagnostic accuracy was lower in detecting early stage and type I OC. Although all of the analyzed amino acids failed to reach a higher AUC than CA125 and HE4, the diagnostic performance of histidine was not subject to OC stage and type. Moreover, the addition of histidine improved the diagnostic performance of all presented multivariate models based on CA125 and HE4.
Most of the amino acids identified in our research as statistically significant were proved to be involved in metabolic pathways altered during cancer growth and progression. Tryptophan depletion triggers apoptosis of effector T cells contributing to the suppression of antitumor immune responses [
24]. Considerable evidence indicates that histamine, a derivative of amino acid histidine, may be a crucial mediator in cancer growth and progression by regulating processes such as angiogenesis, cell invasion, migration, differentiation, apoptosis, and modulation of immune responses [
25]. Histidine decarboxylase that converts histidine to histamine was found to be overexpressed in several cancers, including OC tissue [
26]. Glutamine is used by tumors for nucleotide biosynthesis whereas glutamate, its derivative, serves as a donor of nitrogen for the production of other amino acids. Glutaminase, an enzyme which converts glutamine to glutamate, was found to be frequently upregulated in cancer cells [
24].
Among the limitations of this study are the number of patients and the fact that they were all from a single institute. Nonetheless, this ensured the consistency in gathering and processing the samples. The distribution of histological types of OC consisted of serous (42%), endometrioid (11%), clear cell (8%), mucinous (3%), and undifferentiated carcinomas (26%), and the frequency of the last type was much higher than reported in other European countries. This is probably due to an individual bias of the pathology department and some of these cancers could probably have been classified as high-grade serous. In four cases (10%), the type was not identified because the patients were qualified to neoadjuvant chemotherapy and the diagnosis was obtained after a paracentesis of ascites. The study also excluded cancers other than epithelial OC (i.e., germ-cell and stromal cancers). However, taking into account their extremely low incidence (less than 2% of cancers) this factor has very limited clinical impact. Another potential weakness of the presented research is the possibility of a relationship between behavioral and/or dietary patterns of the patients and alterations in the amino acid profiles [
27]. To reduce this potential bias, malnourished patients were excluded from the analysis and the blood samples were collected after overnight fasting.
The number of patients in the subgroup analyses (early vs. late stage; type I vs. type II) was especially limited, therefore much larger cohorts are needed to verify the utility of the amino acids indicated in these subgroups as relevant. Nevertheless, since only diagnosis at an early, asymptomatic stage is likely to have a significant impact on the clinical outcomes of OC patients, the subgroup analyses provide an important input. The presented study examined the role of SFAA profiles in differential diagnosis of ovarian tumors and assessed the performance of several multimarker models for pre-surgical evaluation of ovarian masses. A possible direction of future research could be the assessment of SFAA profiles in OC screening, before the actual ovarian tumor is observed in ultrasound examination. Further analyses comparing PFAA alterations in OC and other cancers are also necessary to establish the role of possible new OC biomarkers.