Connective tissue of cervical carcinoma xenografts: Associations with tumor hypoxia and interstitial fluid pressure and its assessment by DCE-MRI and DW-MRI

Abstract

Background. A high fraction of stroma in malignant tissues is associated with tumor progression, metastasis, and poor prognosis. Possible correlations between the stromal and physiologic microenvironments of tumors and the potential of dynamic contrast-enhanced (DCE) and diffusion-weighted (DW) magnetic resonance imaging (MRI) in quantification of the stromal microenvironment were investigated in this study. Material and methods. CK-160 cervical carcinoma xenografts were used as preclinical tumor model. A total of 43 tumors were included in the study, and of these tumors, 17 were used to search for correlations between the stromal and physiologic microenvironments, 11 were subjected to DCE-MRI, and 15 were subjected to DW-MRI. DCE-MRI and DW-MRI were carried out at 1.5 T with a clinical MR scanner and a slotted tube resonator transceiver coil constructed for mice. Fraction of connective tissue (CTF_Col) and fraction of hypoxic tissue (HF_Pim) were determined by immunohistochemistry. A Millar SPC 320 catheter was used to measure tumor interstitial fluid pressure (IFP). Results. CTF_Col showed a positive correlation to IFP and an inverse correlation to HF_Pim. The apparent diffusion coefficient assessed by DW-MRI was inversely correlated to CTF_Col, whereas no correlation was found between DCE-MRI-derived parameters and CTF_Col. Conclusion. DW-MRI is a potentially useful method for characterizing the stromal microenvironment of tumors.

Tumor tissues consist of a compartment of malignant parenchymal cells and a compartment of supporting connective tissue, or stroma. The stroma exhibits blood vessels, immune cells, and activated connective tissue cells, primarily cancer-associated fibroblasts [1]. Together with the parenchymal cells, the stromal cells produce a fibrotic extracellular matrix that provides a scaffold for tumor growth. In addition to its supportive structural role, the extracellular matrix acts as a reservoir for growth factors, influences cell signaling, directs angiogenesis, and guides cell migration [2]. A high level of collagen and other extracellular matrix proteins has been documented to be a barrier to the transport of chemotherapeutic agents in tumor tissues [3,4] and has been shown to be associated with progession, metastasis, and poor prognosis in several histologic types of cancer [5–7].

Interactions between the parenchymal tumor cells and their stromal microenvironment lead to uncontrolled neovascularization and the development of an abnormal microvasculature and a metabolic and physiologic microenvironment characterized by nutrient deprivation, hypoxia, acidosis, and interstitial hypertension [8]. These microenvironmental abnormalities may cause genomic instability and promote malignant progression in tumor tissues [9–11]. Preclinical studies have revealed that metastatic dissemination and resistance to radiation treatment may be a consequence of tumor hypoxia [12,13] and interstitial hypertension [14–17]. Clinical investigations have shown that low oxygen tension in the primary tumor is associated with locoregional treatment failure, development of metastatic disease, and poor disease-free and overall survival rates in several malignant diseases [18,19]. Studies of patients with advanced cervical cancer have revealed that high interstitial fluid pressure (IFP) in the primary tumor is linked to high incidence of distant metastases and pelvic recurrence after radiation treatment with curative intent [20,21].

Cancer patients having developed tumors with significant microenvironmental abnormalities may benefit from particularly aggressive treatment. Imaging methods for non-invasive characterization of the microenvironment of tumors are therefore urgently needed. Preclinical studies have revealed that dynamic contrast-enhanced (DCE) and diffusion-weighted (DW) magnetic resonance imaging (MRI) may provide valuable information on the physiologic microenvironment of tumors [22–27]. Significant correlations have been found between DCE-MRI-derived parametric images and extent of hypoxia and interstitial hypertension [22–24] and between DW-MRI-derived parametric images and cell density and extent of necrosis [25–27]. Clinical investigations of patients with cervical carcinoma and other malignant diseases have given results consistent with these preclinical observations [28–33].

However, studies of parameters of the stromal microenvironment of tumors and their relationship to parameters of the physiologic microenvironment are sparse [34,35]. The aim of the present study was to investigate whether the fraction of connective tissue in tumors may be associated with the fraction of hypoxic tissue and/or the IFP and, furthermore, whether the fraction of connective tissue can be assessed by gadolinium diethylene-triamine penta-acetic acid (Gd-DTPA)-based DCE-MRI and/or DW-MRI.

Material and methods

Tumor model

CK-160 human cervical carcinoma xenografts growing in adult female BALB/c nu/nu mice were used as tumor model. The CK-160 line was established from a pelvic lymph node metastasis of 65-year-old woman with FIGO stage IIB squamous cell carcinoma of the uterine cervix [36]. Tumors were initiated from cells cultured in RPMI-1640 (25 mmol/l HEPES and L-glutamine) medium supplemented with 13% bovine calf serum, 250 mg/l penicillin, and 50 mg/l streptomycin. Approximately 5.0 × 10⁵ cells in 10 μl of Hanks’ balanced salt solution were inoculated in the gastrocnemius muscle. Tumors with volumes of 100 800 mm³ were included in the study. MRI and IFP measurements were carried out with mice anesthetized with fentanyl citrate (0.63 mg/kg), fluanisone (20 mg/kg), and midazolam (10 mg/kg). Animal care and experimental procedures were in accordance with the Interdisciplinary Principles and Guidelines for the Use of Animals in Research, Marketing, and Education (New York Academy of Sciences, New York, NY, USA).

MRI

MRI was performed by using a 1.5-T whole-body clinical scanner (Signa; General Electric, Milwaukee, WI, USA) and a slotted tube resonator transceiver coil constructed for mice. The coil was insulated with styrofoam to prevent excessive heat loss from the animals. The body core temperature of the mice was kept at 37–38°C during imaging by using a thermostatically regulated heating pad. The tumors were positioned in the isocenter of the magnet and were imaged axially in a single section through the tumor center.

DCE-MRI was carried out as described earlier [37]. Gd-DTPA (Schering, Berlin, Germany) was diluted in 0.9% saline to a concentration of 0.06 M and administered in the tail vein in a bolus dose of 5.0 ml/kg (0.3 mmol/kg), corresponding to an injection volume of 0.12 0.15 ml. The administration was performed manually during a period of 5 s after the mice had been placed in the MR scanner by using a 24 G-neoflon connected to a syringe by a polyethylene tubing. T₁-weighted images (TR = 200 ms, TE = 3.2 ms, and α_T1 = 80°) were recorded at a spatial resolution of 0.23 × 0.23 × 2.0 mm³ and a time resolution of 14 s by using a spoiled gradient recalled sequence. Two calibration tubes, one with 0.5 mmol/l Gd-DTPA in 0.9% saline and the other with 0.9% saline only, were placed adjacent to the mice in the coil. The DCE-MRI was carried out by using an image matrix of 256 × 128, a field of view of 6 × 3 cm², and one excitation. Two proton density images (TR = 900 ms, TE = 3.2 ms, and α_PD = 20°) and three T₁-weighted images were acquired before Gd-DTPA was administered, and T₁-weighted images were recorded for 15 min after the administration of Gd-DTPA. Gd-DTPA concentrations were calculated from signal intensities by using the method of Hittmair et al. [38]. The DCE-MRI series were analyzed on a voxel-by-voxel basis by using the arterial input function of Benjaminsen et al. [37] and the Tofts pharmacokinetic model [39] to produce parametric images of K^trans (the volume transfer constant of the contrast agent between the blood plasma and the extravascular extracellular space) and v_e (the distribution volume of the contrast agent per unit volume of tissue). Images of IAUGC (initial area under Gd-DTPA curve) and k_ep (the rate constant of the contrast agent between the extravascular extracellular space and the blood plasma) were produced by using a procedure similar to that described by Zahra et al. [40]. It was verified that the Tofts pharmacokinetic model gave good curve fits to the experimental data, consistent with our previous studies of CK-160 tumors [41]. Attempts were also made to analyze the DCE-MRI data by using the extended Tofts pharmacokinetic model, which includes the blood volume fraction (V_b) as a parameter in addition to K^trans and v_e [39]. However, in agreement with previous observations [42], the temporal resolution of our DCE-MRI series was found to be too low to allow reliable assessment of V_b. The strengths and limitations of our DCE-MRI method have been discussed elsewhere [37,43].

DW-MRI was carried out by applying a diffusion-weighted single-shot fast spin echo sequence with ETL = 84 and TR = 5002 ms. The diffusion-weighted images were recorded at a spatial resolution of 0.39 × 0.39 × 2.0 mm³ by using an image matrix of 256 × 256, a field of view of 10 × 10 cm², and 10 excitations. Diffusion sensitization gradients were applied in six non-collinear directions with the following x, y, and z physical gradient combinations: [1 0 1], [− 1 0 1], [0 1 1], [0 1 1], [1 1 0], [− 1 1 0]. Five different diffusion-weightings with diffusion encoding constants of b = 200, 400, 600, 800, and 1000 s/mm² and corresponding echo times of TE = 85, 95.5, 103, 108.9, and 113.9 ms were used to determine ADC. An image without diffusion weighting (b = 0) was recorded for each TE value to compensate for the different TEs associated with the different b values. The b = 0 images were not included in the calculation of ADC. The total scan time of our DW-MRI method was ˜27 min.

Maps of the apparent diffusion coefficient (ADC) were produced with in-house-made software developed in Matlab. Briefly, the directional diffusion images were averaged on a voxel-by-voxel basis to non-directional diffusion images. ADC values were calculated for each voxel by fitting signal intensities (S) to the mono-exponential model equation:

by using a linear least square fit algorithm. A mono-exponential model and b values from 200 to 1000 s/mm² were chosen to avoid confounding effects of perfusion, as recommended by Padhani et al. [44]. The signal decay of a large number of voxels was investigated to verify that the mono-exponential model gave good fits to the data. The fits generally had a correlation coefficient of 0.98 0.99.

Preparation of histologic sections from imaged tumors

To search for spatial correlations between parametric MR images and tumor histology, histologic sections were prepared from the scanned cross-sections of imaged tumors. We marked off the central axial scan on the tumors with a felt-tip pen and euthanized the mice immediately after imaging. The tumors were then cut into two halves along the scan, resected, and fixed in liquid N₂. Histologic sections, separated by ˜ 400 μm, were cut from the central axial tumor plane, corresponding to the 2-mm-thick MR scan.

Histologic analyses

Histologic sections were immunostained for connective tissue or hypoxia by using a peroxidase-based indirect staining method [45]. An anti-collagen rabbit polyclonal antibody (Dako, Glostrup, Denmark) or an anti-pimonidazole rabbit polyclonal antibody (gift from Professor Raleigh, University of North Carolina) was used as primary antibody. Diaminobenzidine was used as chromogen, and hematoxylin was used for counterstaining. Three whole cross-sections were examined at high magnification for each tumor. The area fractions of the tumor tissue staining positive for collagen or pimonidazole were determined by image analysis and used as parameters for connective tissue fraction (CTF_Col) and hypoxic fraction (HF_Pim), respectively [45].

Interstitial fluid pressure

IFP was measured in the center of the tumors with a Millar SPC 320 catheter equipped with a 2F Mikro-Tip transducer (Millar Instruments, Houston, TX, USA) [46]. Marks on the catheter assured correct positioning of the sensor, and a single measurement was carried out in each tumor. The catheter was connected to a computer via a Millar TC-510 control unit and a preamplifier. Data acquisition was carried out by using the LabVIEW software (National Instruments, Austin, TX, USA). By measuring the IFP in the same tumors twice, we have shown that this method produces highly reproducible IFP values [24].

Statistical analysis

Curves were fitted to data by regression analysis. The Pearson product moment correlation test was used to search for correlations between parameters. Probability values (p) and correlation coefficients (R²) were calculated by using the SigmaStat statistical software (SPSS Science, Chicago, IL, USA). A significance criterion of p < 0.05 was used.

Results

Individual CK-160 tumors differed substantially in HF_Pim and CTF_Col and, generally, tumors with high HF_Pim showed low CTF_Col and vice versa. This is illustrated in Figure 1, which shows immunohistochemical preparations of a tumor with high HF_Pim and low CTF_Col (Figure 1a) and a tumor with low HF_Pim and high CTF_Col (Figure 1b).

Figure 1. Immunohistochemical preparations of (a) a CK-160 tumor with high HF_Pim and low CTF_Col and (b) a CK-160 tumor with low HF_Pim and high CTF_Col. Upper panels: pimonidazole staining of tumor hypoxia. Middle panels: collagen staining of tumor connective tissue. Lower panels: high magnification images referring to the boxes shown in the middle panels.

Associations between CTF_Col and the physiologic microenvironment of CK-160 tumors were searched for by measuring IFP in 17 tumors before the tumors were resected and prepared for immunohistochemical assessment of CTF_Col and HF_Pim. A significant positive correlation was found between CTF_Col and IFP (p = 0.00047; R^{2 =} 0.57; Figure 2a), and a significant inverse correlation was found between CTF_Col and HF_Pim (p = 0.0016; R^{2 =} 0.50; Figure 2b).

Figure 2. Plots of (a) CTF_Col versus IFP and (b) CTF_Col versus HF_Pim for CK-160 tumors. Points represent single tumors. Curves were fitted to the data by linear regression analysis.

To investigate whether the fraction of connective tissue in CK-160 tumors can be determined by DCE-MRI, 11 tumors were subjected to DCE-MRI before the CTF_Col of the tumors was assessed by analysis of immunohistochemical preparations. K^trans and v_e differed substantially among individual tumors. This is illustrated in Figure 3a, which shows the K^trans images and frequency distributions of a tumor with low K^trans values and a tumor with high K^trans values, and in Figure 3b, which shows the v_e images and frequency distributions of a tumor with low v_e values and a tumor with high v_e values. There was no correlation between K^trans and CTF_Col or between v_e and CTF_Col. This is illustrated in Figure 3c, which shows a plot of median K^trans versus CTF_Col (p > 0.05), and in Figure 3d, which shows a plot of median v_e versus CTF_Col (p > 0.05). Furthermore, neither IAUGC nor k_ep showed a significant correlation to CTF_Col (data not shown).

Figure 3. Examples of (a) K^trans images and frequency distributions of CK-160 tumors with low (top) and high (bottom) K^trans values and (b) v_e images and frequency distributions of CK-160 tumors with low (top) and high (bottom) v_e values, and plots of (c) median K^trans versus CTF_Col and (d) median v_e versus CTF_Col for CK-160 tumors. Points represent single tumors.

To investigate whether DW-MRI can provide information on the fraction of connective tissue in CK-160 tumors, 15 tumors were subjected to DW-MRI and subsequent immunohistochemical assessment of CTF_Col. Individual tumors differed substantially in ADC, as illustrated in Figure 4a, which shows the ADC maps and frequency distributions of two tumors, one with low ADC values and the other with high ADC values. There was a significant inverse correlation between ADC and CTF_Col, as illustrated in Figure 4b, which shows a plot of median ADC versus CTF_Col (p = 0.0013; R^{2 =} 0.56). However, spatial correlation between ADC and CTF_Col could not be detected, most likely because of significant heterogeneity in CTF_Col within voxels.

Figure 4. Examples of ADC images and frequency distributions of CK-160 tumors with low (top) and high (bottom) ADC values (a), and a plot of median ADC versus CTF_Col for CK-160 tumors (b). Points represent single tumors. The curve was fitted to the data by linear regression analysis.

The CK-160 tumors included in the experiments described above differed in volume from ˜100 mm³ to ˜800 mm³. There was no significant correlation between tumor volume and CTF_Col, HF_Pim, IFP, K^trans, v_e, IAUGC, k_ep, or ADC (data not shown), implying that the intertumor heterogeneity in these parameters cannot be explained by differences in tumor size.

Discussion

A tumor line showing significant intertumor heterogeneity in fraction of connective tissue, fraction of hypoxic tissue, and interstitial hypertension is required to answer the questions addressed in the present communication. These requirements were fulfilled by the CK-160 cervical carcinoma xenograft line. Furthermore, it has been shown that CK-160 xenografts retain several biologic properties of the donor patient's tumor, including radioresponsiveness and metastatic pattern and propensity [36]. Consequently, CK-160 tumors should be excellent preclinical models for studying relationships between the stromal and physiologic microenvironments of tumors as well as for investigating the potential of DCE-MRI and DW-MRI as non-invasive methods for quantifying the fraction of stromal tissue.

However, it should be noticed that only the parenchyma of xenografted tumors is of human origin whereas the stroma is recruited from the host animal and thus is of murine origin. Moreover, immune-deficient animals like BALB/c nu/nu mice do not show a normal inflammatory response, and the stromal conditions in tumors in such hosts may not necessarily be representative for spontaneous animal tumors or tumors in man. Thus, studies of pancreatic carcinoma have shown that the stroma of xenografted tumors may differ from that of human tumors and tumors in genetically engineered mouse models of this disease [47,48].

Interstitial hypertension in tumors is a result of high resistance to blood flow, low resistance to transcapillary fluid flow, decreased interstitial hydraulic conductivity, increased extracellular matrix contraction, and impaired lymphatic drainage [49,50]. The positive correlation between CTF_Col and IFP observed here could be due to inhibition of fluid flow in the interstitium caused by a dense extracellular matrix and/or contraction of the extracellular matrix caused by connective tissue cells. There is significant evidence that tumor-associated fibroblasts may apply tensile forces on extracellular matrix fibers, an ability that is mediated by matrix-bound integrins in response to specific cytokines [50]. Correlations between collagen density and IFP have also been observed in previous studies of experimental tumors [35,51].

Tumor hypoxia is a result of an imbalance between the rate of oxygen supply and the rate of oxygen consumption [52]. The inverse correlation between CTF_Col and HF_Pim observed here was probably a consequence of an association between tumor microvessels and stroma. We have shown previously that the majority of the microvessels in CK-160 tumors are embedded in bands of connective tissue and that there is a positive correlation between the fraction of connective tissue and microvascular density [53]. Most likely, HF_Pim increased with decreasing CTF_Col because of an association beween CTF_Col and microvascular density, implying that hypoxia in CK-160 tumors was primarily a result of inadequate microvascular density and, hence, inadequate blood and oxygen supply.

Few studies have searched for associations between DCE-MRI-derived parameters and the extent of connective tissue in tumors. Farace et al. [54] used two contrast agents, Gd-DTPA and MS-325 (an albumin-binding contrast agent), and studied contrast-enhancement in two experimental tumor lines differing significantly in stromal content. They showed that high-molecular-weight contrast agents may be superior to low-molecular-weight contrast agents like Gd-DTPA in discriminating between tumors with sparse and extensive stroma.

Associations between CTF_Col and K^trans or v_e were not found in the present study of CK-160 tumors. In a previous study of CK-160 tumors, we showed that the K^trans of individual voxels consisting almost exclusively of connective tissue was significantly higher than the K^trans of voxels consisting primarily of parenchymal tissue or necrotic tissue [41]. In that study, it was concluded that the potential of parametric images of K^trans or v_e in providing information on the histomorphology of cervical carcinomas may be limited, mainly because the tumor tissue showed significant morphological heterogeneity at the subvoxel level. The study reported here is consistent with this conclusion and suggests that DCE-MRI may not be an adequate imaging method for quantitative assessment of the fraction of connective tissue in tumors unless imaging at a high signal-to-noise ratio can be carried out at a significantly increased spatial resolution.

However, a significant inverse correlation was found between ADC and CTF_Col in CK-160 tumors, an observation suggesting that the movement of water molecules in the extracellular space was inhibited significantly by the connective tissue. This suggestion is consistent with studies of pancreatic carcinoma, which have revealed that tumors with dense fibrosis have lower ADC values than tumors with loose fibrosis [55,56]. Furthermore, similar to the present observation, Aoyagi et al. [57] found an inverse correlation between ADC and the density of stromal collagen in esophageal squamous cell carcinoma and concluded that ADC may be an important prognostic factor in esophageal cancer. It should be noticed that only two b values of 0 and 500 s/mm² or 0 and 1000 s/mm² were used in these clinical studies [55–57]. Our study is consistent with the study of Aoyagi et al. [57] and opens for the possibility that ADC values assessed by DW-MRI may have prognostic value also in cervical cancer. Interestingly, DW-MRI may not be the only MR method by which the fraction of connective tissue in tumors can be quantified. Significant correlations have been detected between parameters determined by magnetization transfer MRI and the fractional area of fibrotic tissue in pancreatic carcinoma xenografts [34]. Furthermore, evidence has been provided that the quantity of connective tissue in breast carcinoma xenografts may be assessed by the use of fibrin-specific contrast agents [58]. However, DW-MRI is already an established clinical method for characterizing malignant diseases and may thus be preferable to other non-invasive imaging methods.

There is significant evidence from preclinical and clinical studies involving tumors in several sites that ADC decreases with increasing cell density [25,32]. Moreover, it has been shown that treatment-induced tumor cell death results in increased ADC values, independent of treatment modality [33,59]. The observed associations between ADC and pre- and post-treatment tumor cellularity have led to the suggestion that ADC may be a useful biomarker for the outcome of cancer treatment [59]. The observations reported here for CK-160 tumors are by no means inconsistent with this suggestion. Individual CK-160 tumors do not differ significantly in cell density, most likely because they are established from the same cell line [41]. The fact that CK-160 tumors differ significantly in fraction of connective tissue but not in cell density makes them particularly suitable for studying associations between ADC and stromal content without severe confounding effects from other histomorphologic parameters. However, it should be noticed that necroses may influence the intertumor heterogeneity in median ADC in CK-160 tumors. It has been shown that individual CK-160 tumors may differ significantly in the density of small necrotic foci, as reflected by the intertumor heterogeneity in median v_e [41].

Cancer patients having developed tumors with a stromal microenvironment characterized by a high density of extracellular matrix fibers and a physiologic microenvironment characterized by extensive hypoxia and highly elevated IFP have poor prognosis and may benefit from particularly aggressive treatment [5–7,18–21]. Previous preclinical studies in our laboratory have revealed that DCE-MRI with Gd-DTPA as contrast agent may provide clinically useful information on the physiologic microenvironment of tumors. Those studies demonstrated that tumors with low values of K^trans and v_e are likely to be severely hypoxic [22,43] and to exhibit significant interstitial hypertension [24]. The study reported here showed an inverse correlation between ADC and CTF_col, suggesting that DW-MRI may have the potential to provide valuable information on the stromal microenvironment of tumors. Furthermore, CTF_Col was found to be positively correlated to IFP and inversely correlated to HF_Pim. It is thus possible that DCE-MRI and DW-MRI are imaging methods that may complement each other in characterizations of the microenvironment of tumors and jointly provide novel biomarkers of tumor aggressiveness.

This suggestion is consistent with initial clinical studies of locally advanced cervical cancer. Significant correlations have been found between DCE-MRI-derived parameters and hypoxia-associated parameters of the tumor tissue, including microvascular density and oxygen tension [28,60] and between tumor signal enhancement and outcome parameters, including rate of local control after radiation therapy and probability of disease-specific survival [31,40,61,62]. However, the correlations are generally weak, perhaps because the DCE-MRI was carried out under suboptimal conditions or because of inadequate analysis of the DCE-MRI series. Furthermore, there is significant evidence that DW-MRI may be useful in differentiating between malignant and benign cervical tissues, in detecting pelvic lymph node metastases, and in early detection of response to treatment [33,44,59]. In patients treated with chemoradiotherapy, significant correlations have been found between treatment-induced changes in ADC and response to treatment as well as between ADC and rate of disease-free and overall survival [63–65].

In summary, the present preclinical study provides significant evidence that DWI-MRI is superior to DCE-MRI as a non-invasive method for characterizing the stromal microenvironment of tumors. However, it should be noticed that our study has significant limitations. The main limitation is that the experiments were carried out on three different cohorts of tumor-bearing mice. Sequential DCE-MRI, DW-MRI, IFP measurement, and histologic examinations of the same tumors would have allowed a more comprehensive data analysis. However, our experimental strategy was chosen to avoid excessive experimental strain on the host animals.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

Financial support was received from the Norwegian Cancer Society and the South-Eastern Norway Regional Health Authority.