Disseminated Tumor Cells in Bone Marrow of Gastric Cancer Patients: Correlation with Tumor Hypoxia and Clinical Relevance

Aim. The evaluation of the clinical relevance of disseminated tumor cells (DTCs) in bone marrow (BM) of patients with gastric cancer (GC) and their association with primary tumor hypoxia. Patients and Methods. 89 resected specimens were used. DTCs were detected using immunocytochemistry, the level of tumor hypoxia using NMR spectroscopy, CD68, CD34, VEGF, and VEGFR-1 (Flt-1) expression using immunohistochemistry, and MMP-2 and MMP-9 activity using zymography. Results. DTCs were detected in 51.4% of GC patients with M0. There was significant correlation between frequency of DTCs in BM and level of tumor hypoxia (P < 0.024). DTCs presence was accompanied with Flt-1 positivity of BM. The correlation between DTCs and tumor VEGF expression in patients with M0 was shown (P < 0.0248). Activity of MMP-2 and MMP-9 in BM was linked with DTCs in patients with M0 (P < 0.05). Overall survival (OS) of patients with M0 and DTCs was shorter than that of patients without DTCs (patients in both groups were operated only) (P = 0.0497). Conclusion. Appearance of DTCs correlates with hypoxia level in primary tumors. Detection of DTCs in GC patients may be relevant indicator for adjuvant chemotherapy using.


Introduction
Gastric cancer is one of the most common cancers in Europe ranking the fifth after lung, prostate, colorectal, and bladder cancers in men and breast, colorectal, lung, and cancer of the corpus uteri in women [1]. In Ukraine in 2011 the annual age-standardized incidence rate was 21.5/100.000 ranking the fourth after lung, skin, and prostate cancers in men and 8.8/100 000 ranking the seventh after breast, skin, corpus uteri, colon, cervix uteri, and rectal cancers in women [2]. The therapy outcome of gastric cancer is still not satisfactory, and distant metastasis is kept as a key factor in the unfavorable results of gastric cancer treatment. It is known that even curative resection can not always guaranty expected longterm results of therapy that may be explained by the early dissemination of tumor cells even still before the surgical intervention [3][4][5]. It is shown that "tumor cells can disseminate from the earliest preneoplastic lesions, sometimes even before the formation of overt primary tumors" [6]. At the same time it is unable to detect early metastasis up till now despite modern tools, for example, MRT, PET.
Tumor cells leaving primary site can mainly settle in bone marrow (BM) with potency to form the metastases. These cells named disseminated tumor cells (DTCs) cannot be detected by conventional cytological methods, but they may be found both by immunocytochemistry and molecular technologies.
It was shown that DTCs in BM may be detected in 25-60% of patients with different tumors categorized as M 0 [4][5][6][7]. Conception of the detection of micrometastases in BM 2 Journal of Oncology has been introduced in clinical practice more than 30 years ago [8]. BM is considered now as a common homing-organ for DTCs that escape from epithelial tumor, first of all from breast, lung, prostate, and colorectal cancers. It was shown that persistent DTCs in BM are associated with unfavorable prognosis for breast cancer patients [9].
Perhaps firstly publication concerning DTCs in gastric cancer patients appeared in 1991 when Schlimok et al. [10] have shown the presence of DTCs in BM of 30% patients with gastric cancer without obvious distant metastases. The correlation between DTCs and some clinicopathological characteristics, in particular regional and distant metastasis, and Lauren classification was also determined. Further studies confirmed these data and focused one's attention on methodology of DTCs search and pointed out the clinical relevance of DTCs in gastric cancer patients, in particular their negative impact on survival [11][12][13][14][15][16][17]. Kolodziejczyk et al. [18] studying the influence of neoadjuvant chemotherapy on DTCs in gastric cancer patients have shown that frequency of DTCs finding was significantly decreased after chemotherapy but without substantial impact on disease outcome. Nevertheless, authors proposed to continue these studies, in particular in multicenter protocol investigations. At the same time some observations have been published where the clinical relevance of DTCs in gastric cancer patients was not observed [19][20][21]. Recently published review of Bidard et al. [22] allows to conclude that problem of DTCs in gastric cancer is relevant for clinics but it is not fully solved and needs to be clarified to improve the treatment outcome.
Our study was aimed to evaluate the influence of DTCs in BM of gastric cancer patients on survival and their prognostic significance. Moreover, our attention was focused on the assessment of possible correlation between DTCs in BM and hypoxia profile in primary tumor exploiting wellknown fact that solid tumors are hypoxic that mediates tumor aggressiveness and poor disease outcome [23].

Patients.
A total of 89 patients (62 men and 27 women) with primary gastric cancer (GC) were diagnosed and treated at the City Clinical Oncological Center (Kiev), during period 2008-2011 (Table 1). No patient received chemotherapy or radiation prior to surgery. Tissue samples were taken immediately after tumor excision. Tumors were classified and staged according to the 2002 version of the UICC staging system [24]. Histological types of tumor were evaluated by WHO histological classification (2000) [25]. All patients were thoroughly informed about the study that was approved by the local ethics committee.

Detection of Tumor Cells in Bone
Marrow. Preoperatively, 2.0-3.0 mL of BM aspirates from the sternum with conventional cautions to avoid the hit of skin epithelial cells into the sample was taken into a heparinized syringe and transferred into a tube "Sarstedt" containing EDTA K. After Ficoll-Hypaque density centrifugation (density, 1.077; Sigma-Aldrich, USA) to isolate the mononuclear cell fraction (1105 g for 20 minutes), the interphase was washed twice in  19 (21.4) phosphate-buffered saline (PBS) with removing of erythrocytes (Uti-Lyse Erythrocyte Lysing Reagent, Dako Cytomation, USA), resuspended to a concentration of 570 ⋅ 10 3 cells/30 L, and cytocentrifuged on glass slides. Specimens were air-dried from 12 to 24 hours and stained immediately or stored at -20 ∘ C. Detection of tumor cells (cytokeratin-positive cells, CKpositive cells) in BM cytospin preparations fixed in acetone was provided by APAAP method (alkaline phosphataseantialkaline phosphatase) and visualization system EnVision G/2 System/AP Rabbit/Mouse (Permanent Red) (Dako Cytomaiton, Denmark). Monoclonal mouse antibodies against panCK (clone AE1/AE3, Dako Cytomation, Denmark) were used as primary antibodies. Each assay was controlled negatively by staining of one cytospin preparation with nonspecific IgG 1 (MOPC21, Sigma). Number of tumor cells (CK-positive cells) was expressed on 10 6 BM mononuclear cells. BM samples were scored "positive" if the presence of two or more CK-positive cells per 10 6 mononuclear cells was detected (from 6 to 12 slides per patient were screened).

Detection of Flt-1-Positive Cells in Bone Marrow with
Immunohistochemical Method. Cytospins were fixed by formol-acetone solution (pH 6.6) in accordance with the instruction. Slides were treated by 0.3% Triton X-100 solution, washed by PBS and blocking of endogenous peroxidase followed by incubation in 3% bovine serum albumin to switch off nonspecific reaction antigen-antibody. Cytospins were incubated with primary polyclonal rabbit antibodies against Flt-1 (sc-316, Santa Cruz Biotechnology, Inc., USA) in optimal dilution 1 : 80 within 1 h. After washing of primary antibodies slides were processed with PolyVueHRP Detection System Components (Diagnostic BioSystems, USA).
Microvessel density (MVD), detected by immunostaining for CD34, was assessed by the hot spot method [26]. CD68positive cells were counted per 1000 cells in each slide and the number of CD68-positive cells was reported as percent. VEGF expression was assessed by scoring the number of all positive cells per 200x field. When the tumor consisted of less than 25% immunoreactive cells, the case was scored as weak, cases with 26-50% immunoreactive cells were scored as moderate, and those with 51-100% immunoreactive cells were scored as strong.

Metalloproteinase-2 and Metalloproteinase-9 Activity
Assay. Tumor specimens and BM aspirates (0.5 mL) were placed into the liquid nitrogen up to processing. Activity of MMP-2 and MMP-9 was determined for each sample by zymography in 12% polyacrylamide gel with SDS and 0.1% of gelatin as substrate [27].
2.6. 31 P NMR Spectroscopy. Level of tumor hypoxia was assessed with 31 P NMR spectroscopy. 31 P NMR spectra of perchloric acid (PCA) tumor extracts were acquired by means of a high-resolution Bruker 400 MHz spectrometer (Widebore Ultrashield, AV-400 electronics, Germany) using a probe of 5 mm inner diameter. All details of method were presented in our earlier publication [27].

Tumor Cells in Bone Marrow and Their Correlation with Clinical Variables.
Individual patient data from a total of 89 histologically confirmed gastric cancer patients were included in this study ( Table 1) (Table 2). Meanwhile, the association between the presence of CK-positive cells in BM and level of hypoxia in primary tumor was detected: severe and moderate hypoxia was found in 75% of primary tumors in patients with DTCs in BM while mild and weak hypoxia in 32.2% only ( < 0.01). Level of tumor hypoxia assessed by NMR spectroscopy [27] was ranged as follows: if the PME/Pi < 1.0, tumors are characterized by severe hypoxia, 1.0 < PME/Pi < 1.4 moderate hypoxia, 1.4 < PME/Pi < 2.0 mild hypoxia, and PME/Pi > 2.0 weak hypoxia (satisfactory oxygenation).
It was also determined that the probability of appearance of tumor cells in BM of patients with category M 0 is increased by a factor of 11.4 (odds ratio 11.4, 95% CI 2.71-47.89, 2 = 12.3, < 0.001) when tumors were characterized by severe and moderate hypoxia.

Flt-1-Positive Cells in Bone Marrow and Primary Tumor.
It was found that Flt-1 positive cells were detected both in BM and tumor, in 58.5% and 79% of patients, respectively. The presence of CK-positive cells in BM was accompanied with the Flt-1-positivity of BM in 67% and the absence of CK-positive cells with the Flt-1-positivity in 45% of cases. The correlation between the number of Flt-1-positive tumor and

Activity of Gelatinases in Primary Tumor and Bone
Marrow and CK-Positive Cells in Bone Marrow. It was shown that the association between activity of MMP-2 in tumor and presence of DTCs in BM, in particular MMP-2 activity was 9.2 ± 5.1 g/g tissue in patients with DTCs in BM whereas MMP-2 activity, was 4.1 ± 2.8 g/g in patients without DTCs in BM ( < 0.05). The association of tumor MMP-9 activity with DTCs in BM was not found. At the same time activity of both gelatinases in BM was linked with DTCs in BM; in particular activities of MMP-2 and MMP-9 were 8.6 ± 4.0 and 7.5 ± 3.4 g/g in patients with DTCs in BM and 2.8 ± 1.4 and 2.6±1.85 g/g in patients without DTCs in BM ( < 0.05 and < 0.05, resp.). It has to be noted that patients with category M 0 were analyzed only.

Overall Survival of Patients with and without Tumor Cells in Bone Marrow and Treated and Not Treated with Adjuvant Chemotherapy.
Overall survival (OS) of patients with category M 0 with DTCs in BM was shorter than that of patients with category M 0 and without DTCs in BM ( = 0.0497) (Figure 1). It has to be noted that patients in both groups were operated only. It can indicate that the detection of DTCs in BM may be considered as obligatory procedure before the decision concerning further treatment, in particular of patients with category M 0 .
Median follow-up time was 19.4 (range, 2.6-60.85) months from diagnosis for all patients (mean 21.9 ± 1.9). OS was significantly shorter in patients with M 0 category and with DTCs, compared with patients with no DTCs (log-rank test: < 0.05). If overall survival was analyzed for all patients (M 0 and M 1 categories) the difference was not statistically significant ( > 0.05).
Overall, 30 patients (33.7%) died during follow-up. In 26 patients (86.7%) death was related to gastric cancer. Of these, 20 patients (66.7%) had DTCs in BM. Of 20 patients with DTCs in BM, 8 patients (40%) had M 0 category. Of 26 patients, who died, 13 (50%) have received any kind of adjuvant chemotherapy, and 10 (38.5%) had DTCs in BM (4 patients (40%) had M 0 category). Survival in these patients was significantly shorter in patients with M 0 category with DTCs compared with those patients with no DTC: 22.5 versus 39.5 months (Student's test, < 0.05). Survival in all patients (with M 0 and M 1 categories) was almost the same (22.2 and 24.9 months, resp.) independently from DTCs in BM.
It was also found that OS of patients with DTCs in BM and Flt-1-positive cells in BM was significantly shorter than that of patients with no DTCs in BM, but with Flt-1-positive cells in BM ( = 0.0437) (Figure 2). Patients in both groups were categorized as M 0 and M 1 and treated with adjuvant chemotherapy. OS of patients with category M 0 was not influenced by Flt-1 positivity of BM ( > 0.05).
Moreover, it was evaluated that OS of patients with DTCs in BM and VEGF-positive tumor was significantly shorter than that of patients without DTCs in BM and VEGF-positive

Discussion
Our study found that tumor cells present in BM of 51.4% patients with gastric cancer with category M 0 . DTCs in BM were detected in 35-60% of gastric cancer patients with category M 0 by other authors [7]. It has to be noted that there are not enough publications concerning the significance of DTCs in BM in patients with M 0 . At the same time, namely, this information is more relevant than that in regard to patients with M 1 that are diagnosed by conventional methods. It should be noted that association between DTCs in BM and the level of hypoxia in the primary tumor was found: tumor cells in BM were found in 80% of cases where severe hypoxia was found in the primary tumor, whereas under moderate and mild hypoxia DTCs they were found only in 20% of cases ( < 0.05). It was also determined that a probability of DTCs appearance in BM is increased by a factor of 11(odds ratio 11.8, 95% CI 3.0587-45.60, 2 = 14.79, < 0.001) when primary tumors are characterized by severe hypoxia. These data indicate a possible positive impact of hypoxia-associated signaling pathways on the escape of tumor cells from the primary tumor and their dissemination into the BM to form a premetastatic niche as suggested by Kaplan et al. [28]. Close correlation was observed between DTCs and VEGF expression in primary tumor confirmed by early observation [15], but correlation between DTCs and tumor MVD was not found in contrast to other authors [14,29]. In regard to Flt-1 expression in BM it has to be noted that our study confirmed data obtained by Mimori et al. [30] that simultaneous presence of tumor cells and Flt-1-positive cells in BM is clinically relevant for metastasis.
The role of MMPs in tumor dissemination is well known, but MMPs activity in BM of gastric cancer patients was not determined till now. Our study has shown that activity of MMP-2 and MMP-9 in BM of patients with M 0 but with DTCs was increased by a factor of 3.1 and 2.9, respectively. It allows suggesting that gelatinases may play a significant role in the formation of premetastatic niche, in particular, in the reorganization of cellular microenvironment in BM.
In conclusion, it may be summarized that 51.4% of gastric cancer patients categorized as M 0 have tumor cells in BM. Presence of DTCs is correlated with level of tumor hypoxia and accompanied with Flt-1 positivity of BM. The probability of Flt-1 positivity of BM was increased by a factor of 2.7 in patients with severe and moderate hypoxia in tumor. The significant correlation between CK-positivity of BM and VEGF expression as well as MMP-2 activity in tumor was shown. There is relevant observation that activity of both gelatinases in BM correlated with presence of DTCs in BM of patients with M 0 only. OS of patients with M 0 with DTCs in BM was significantly shorter than that in patients without DTCs. It has to be noted that patients in both groups were operated only. It was also evaluated that OS of patients with DTCs and Flt-1 positivity of BM was significantly shorter than that of patients without DTCs but with Flt-1-positive BM (these patients were treated with adjuvant chemotherapy). Detection of DTCs in gastric cancer patients may be as relevant indicator for personalised cancer therapy, in particular in the choice of treatment tactic for GC patients, especially with category M 0 . Perhaps, DTCs in BM may be as an argument to use adjuvant chemotherapy, though some other factors have to be taking into account, for instance, the influence of hostile microenvironment on DTCs dormancy in BM.