Background
Solid tumours show interstitial fluid pressures (IFP) that are elevated above that of normal tissues. Tumour growth and development is supported by both the pre-existing host vasculature and by neovasculature generated through the process of angiogenesis. Tumour angiogenesis generates abnormal vessels [
1‐
3] that demonstrate several anomalies including an incomplete or absent endothelial cell layer and basement membrane which makes them hyper-permeable [
4]. These vessels exhibit a high resistance to capillary blood flow and a low resistance to transcapillary flow, resulting in a net efflux of fluid into the surrounding interstitial space where a lack of functional lymphatics allows it to accumulate, distending the elastic extracellular matrix and increasing the interstitial pressure [
5,
6]. An equilibrium is established where the capillary and interstitial pressures are equivalent resulting in reduced fluid movement through the interstitium [
6]. In addition, the tumour interstitium itself is thought to be abnormal, comprising a dense network of collagen fibres, as well as increased fibroblasts, macrophages and other cells involved in inflammation, which further contribute to elevated IFP values [
3]. It is clear from these previous studies that high IFP in tumours arises because of the complex interplay between the abnormal vasculature and the abnormal interstitium. However, the pathophysiologic mechanisms underlying widely varying IFP values in human and experimental tumours of the same and differing types, and the influence of growth site and the host, is less well understood.
Elevated tumour IFP plays a role in the pathophysiological microenvironment that characterises solid tumours contributing to disease progression and therapeutic resistance [
3,
6‐
9]. The mechanisms involved remain to be fully elucidated but several experimental animal studies have shown an improved uptake of therapeutic agents in response to a reduction in tumour IFP suggesting that high tumour IFP acts as a barrier to drug delivery [
10‐
17]. Furthermore, there are clinical data showing that tumour IFP correlates with response to treatment [
18,
19], with strong evidence for high IFP as an adverse prognostic indicator in cervix cancer patients treated with radiotherapy [
7,
9]. Patients in the latter study were significantly more likely to develop distant recurrence if they presented with a tumour IFP value above the group median (19 mmHg), which suggests a role for IFP in metastatic spread. A relationship has also been observed between tumour IFP values and metastasis in experimental melanoma xenografts [
20].
These data, coupled with the breadth of data demonstrating elevated IFP in a wide range of human tumours [
7,
18,
19,
21‐
25], designate high tumour IFP as an important therapeutic problem. Further preclinical investigation is needed to understand the mechanisms underlying the adverse prognostic effect of high IFP and the implications for treatment. However, little is known about the most appropriate experimental model for these studies in relation to clinical tumour behaviour. To date most studies have focused on tumour models grown sub-cutaneously [
13,
26‐
28] although evidence suggests that orthotopic models may be more clinically relevant [
29,
30]. Although IFP has been measured in a variety of different tumour models, to our knowledge no previous study has focussed on the influence of tumour growth site on the development of the pathophysiological tumour microenvironment, or more specifically, tumour IFP. As such, the purpose of this study was to assess IFP in a number of different murine (KHT-C) and xenograft (ME180, SiHa) tumour models growing both ectopically and orthotopically and to examine features of the tumours (lymphatic and blood vascular density, hypoxia, perfusion) that might relate to the IFP levels and to disease progression. The murine and xenograft models were selected on the basis of previous studies within our lab [
31,
32] and the polyoma middle-T (MMTV-PyMT) transgenic spontaneous mammary tumour model was included to allow comparison with the transplanted models [
33]; IFP has only previously been assessed in one other spontaneously arising tumour model [
34]. The orthotopic human cervix cancer xenograft models were included because of the direct relevance of this model to our clinical program [
7,
9]. IFP was examined for each of the different tumour models and growth sites and related to tumour size, metastatic dissemination, tumour hypoxia, proliferation and vascular and lymphatic density.
Methods
Mice and tumour cell lines
Experiments were performed using MMTV-polyoma middle-T transgenic mice (MMTV-PyMT;[
33]) bred in-house, the previously described KHT-C murine fibrosarcoma cell line [
35] and the ME180 and SiHa human cervical carcinoma cell lines stably transfected to constitutively express the fluorescent marker DsRed [
36]. All cell lines were maintained on an alternative
in vitro/
in vivo growth cycle.
In vitro cells were maintained as monolayers in plastic tissue culture flasks using α-MEM medium (Life Technologies, Inc., Burlington, Canada) supplemented with 10% fetal bovine serum (Wisent, Quebec, Canada). The cervical carcinoma cell lines were maintained under G-418 selection (400 μg/ml). Cells between their 2
nd and 5
th in vitro passage were removed from the flasks during exponential growth using 0.05% trypsin for transplantation into mice. KHT-C cells were transplanted into syngeneic 8–12 week old C3H/HeJ male mice (Jackson Laboratory, Bar Harbour, ME). ME180 and SiHa cells were transplanted into 8–12 week old female CB-17/SCID mice obtained from an in-house breeding program. PyMT cells were transplanted into female FVB (wild-type, w.t.) or SCID mice. Tumours were initiated either intramuscularly (i/m) in the left gastrocnemius muscle, or sub-cutaneously (s/c) on the flank by injection of 2.5 × 10
5 or 5 × 10
5 cells respectively in a 50 μl volume of α-MEM media. Tumours growing i/m were monitored by measuring the external leg diameter of the mouse. Tumours growing s/c were measured directly. Orthotopic cervical and mammary gland tumours were initiated from donor tumours using protocols described below. Animals were housed at the Ontario Cancer Institute animal facility and had access to food and water
ad libitum. All experiments were performed under protocols approved according to the regulations of the Canadian Council on Animal Care.
Orthotopic implantation in the cervix or mammary gland
The method for orthotopic implantation of tumour fragments into the cervix has been described in detail previously [
36]. In brief, donor tumours grown i/m were excised under sterile conditions and small fragments (1.5–2 mm in diameter) were sutured into the site of a small incision in the uterus at the level of the cervix. Once tumours were palpable, IFP measurements were taken and tumours, lumbar lymph nodes and lungs were imaged/removed for further analysis. Spontaneously arising donor tumours in MMTV-PyMT transgenic mice were excised under sterile conditions and divided into small fragments of approximately 2 mm in diameter. A fine incision was made in the skin to expose the 4
th mammary gland. A small incision was made in the fat pad of the 4
th mammary glands and a tumour fragment was sutured in place using a single 8-0 silk suture. The skin was closed using stainless steel wound clips. The left and right 4
th mammary glands were implanted with a donor tumour fragment from either the same or a different donor tumour (Additional File
2). This allowed the effect of donor tumour variability on the subsequent development of the recipient tumour microenvironment to be assessed without the confounding factor of inter-animal variability. All surgical procedures were carried out under anaesthesia (2% isofluorane). Buprenorphine (0.1 mg/kg) was administered s/c following surgery to alleviate pain.
Tumour growth was monitored using callipers to measure the width and length of the tumour; once a size of 50–80 mm2 was attained IFP measurements were initiated. Mice were sacrificed once a tumour reached a size of 200–250 mm2. Half of each tumour was fixed in 10% neutral buffered formalin and half was snap frozen in OCT for histological analyses. The lungs were excised for examination of metastases.
Pressure treatment in vitro
KHT-C tumour cells were exposed to elevated pressure (20 mmHg) for various times
in vitro in 10 cm tissue culture dishes seeded with a sub-confluent cell monolayer in a modular incubator chamber (Billups-Rothenberg, Del Mar, Canada). Pressure levels used were based on the average IFP values apparent when grown i/m. Pressure was controlled through adjustments of a dual scale low pressure gauge (0–15 KPa; Cole Parmer, Quebec, Canada) on the outlet port. Control cells were gassed using the same system, but without the addition of a pressure gauge on the outlet port, allowing the gas to flush through freely at atmospheric pressure. The metastatic potential of pressure treated tumour cells was assessed using an experimental lung metastases assay [
37].
IFP measurements
Interstitial fluid pressure was measured using a wick-in-needle technique [
38]. Measurements were made using a 23-gauge needle with side port connected to a pressure transducer (Model P23XL, Viggo-Spectramed, Oxnard, CA) and an electronic data acquisition and recording system (Model MP100, World Precision Instruments, Sarasota, FL) through 470 mm of PE20 polyethylene tubing (Becton Dickinson, Franklin Lakes, NJ, USA). A "wick" was placed in the distal portion of the needle, and the entire system was flushed with a heparin sulphate/saline solution (1:10) [
39]. IFP measurements were taken at three to four different locations in the tumour, and the mean value of these readings was taken to represent the tumour IFP.
Assessment of macroscopic and microscopic metastases
Following sacrifice of tumour bearing animals, the lungs were removed and fixed overnight in Bouin's solution (BDH Inc., Toronto, Canada). A dissecting microscope was used to count the number of visible metastases in each of the five lobes and the total number of lesions counted per lung reported (KHT-C and MMTV-PyMT). In the case of too many metastatic lesions to count, the wet weight of the lungs was taken as representative of metastatic burden. For the orthotopic ME180 tumours that had been transfected to express DsRed, lymph node metastases were visualised by fluorescence and counted as previously described [
36].
Microscopic lung metastases were detected immunohistochemically (ME180 and SiHa) following fixation of lungs in Bouin's solution (BDH Inc., Toronto, Canada). All five lobes were paraffin embedded and four 4 μm sections 150 μm apart were cut from each lobe. The number of visible micro-metastases in each of the five lobes was then counted using a light microscope. Two or more clumped tumour cells were scored as a lesion. Micro-metastases are reported as the total number of lesions counted per lung.
Histological Analyses
Analyses were carried out using immunohistochemistry for tumour hypoxia (EF5), vascular density (CD31), lymphatic vessel density (LYVE-1), and proliferation (Ki67). Tumour bearing animals were injected with the hypoxia marker EF5 [2-(2-nitro-1H-imidazole-1-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide]; obtained from Dr. Cameron Koch, University of Pennsylvania, at 10 mg/kg 2.5 h prior to tumour excision. Once excised, half of the tumour was fixed in 10% neutral buffered formalin, and the remaining half placed in optimal cutting temperature (OCT) embedding medium (Tissue-Tek, Sakura, USA), and snap frozen in liquid nitrogen. Paraffin-embedded tissue was used for all markers with the exception of CD31.
For each marker, two sections were cut 100 μm apart due to intra-tumoural heterogeneity (4 μm sections for paraffin-embedded tissue, 5 μm sections for frozen tissue). The slides were then processed according to standard immunohistochemical protocols. The primary antibodies used were: for EF5, the biotinylated antibody ELK3-51 (1/500; a gift from Dr. Cameron Koch, University of Pennsylvania); for CD31, the rat anti-mouse CD31, clone MEC 13.3 (1/500; Pharmingen, Canada); for Ki67, mouse anti-human Ki67 clone MIB-1 (1/100; DAKO, Canada); for LYVE-1, rabbit anti-mouse LYVE-1 (1/200 Abcam, Canada). For all markers, apart from EF5, primary incubation was followed by a 30 minute incubation with a biotinylated secondary (Vector Labs, Canada) and horseradish peroxidase conjugated ultrastreptavidin labelling reagent (ID labs, Canada). Nova Red (Vector) with Mayer's hematoxylin counterstain was used for chromogenic detection.
The stained sections were analysed using the Aperio imaging system (Aperio Technologies, California). Entire sections were scanned using the ScanScope CS and the total area of positive staining quantified using a positive pixel algorithm designed for brown/blue immunohistochemical stains. The same settings were used for each stain across all images, and the area of positive staining was calculated by dividing the total number of positive pixels (weak, medium and strong staining) by the total number of pixels in the image (positive + negative pixels) to yield the overall percentage of positive pixels. To assess intra-tumour heterogeneity for each stain, 10 (CD31) or 20 (EF5, Ki67 and LYVE1) high-magnification (20×) fields were randomly selected from each tumour section and analysed independently using the same methodology.
Blood flow detection using high-frequency micro-ultrasound functional imaging
Real-time ultrasound biomicroscopic imaging of anaesthetised (2% isofluorane) mice was carried out using speckle variance analysis of high frequency ultrasound (Vevo660, VisualSonics, Inc., Toronto, ON, Canada) images as previously described [
40]. The ultrasound transducer transmits at a central frequency of 40 MHz with a focal length of 6 mm. The lateral and axial resolutions were 68 and 38 μm, respectively. Ultrasound gel (Aquasonic 100, Parker Laboratories, Fairfield, NJ) was used as a coupling agent on the skin. A minimum of five brightness mode (B mode) two-dimensional image planes were acquired per tumour, each with cineloops of 300 frames at a frame rate of 17/s. Speckle pattern and intensity during real-time B mode imaging of stationary tissue remain constant, and the temporal variance of speckle intensity increases with tissue motion. A speckle-variance flow-processing algorithm devised by Yang [
41] was used to calculate changes in speckle intensity between sequential frames as an indication of functional blood flow [
40]. This technique gives a relative indication of the number of perfused vessels in each tissue plane, and has been validated previously against perfusion assessed using injection of Hoechst 33342 [
40]. Each of the 5 B mode image planes were analysed for 3 i/m and 6 cervix ME180 tumours.
Statistical Analysis
Experiments with three or more groups were analysed for statistical significance using the Kruskal-Wallis statistic, and individual comparisons within these groups were carried out using Dunn's test. Experiments with two groups were analysed for statistical significance using the Mann-Whitney statistical test. Correlation was assessed using the correlation coefficient derived from linear-regression analysis.
Discussion
There is increasing interest in the adverse effects of elevated tumour IFP on drug delivery and treatment response. The aim of this study was to characterise this relatively unexplored parameter of the pathophysiological tumour microenvironment in a range of tumour models growing in different sites as a foundation for future experiments and to explore various phenotypic properties such as blood and lymphatic vascular density and tumour hypoxia that may impact on the development of tumour IFP. Systematic differences in tumour IFP as a function of both tumour type and implantation site were observed, consistent with an effect of predetermined genetic factors and interactions between the tumour and host microenvironments in regulating IFP. However, this explained only a portion of the total observed variability in IFP, possibly reflecting stochastic development and remodelling of the tumour vasculature.
IFP levels in individual tumours are influenced primarily by three fundamental pathophysiologic parameters: the trans-capillary and interstitial hydraulic conductivities and the capillary pressure [
5,
53,
54]. In general, tumours are characterized by abnormal, highly permeable vessels and a relatively impermeable interstitium [
6]. Fluid that leaks from the vessels accumulates in the interstitium and causes the pressure to rise. However, there is probably wide variability in each of these parameters among individual tumours, which contributes to heterogeneity in IFP values. In tumours where the trans-capillary conductivity is substantially less than interstitial conductivity, the IFP is low and much less than the capillary pressure. At the other extreme, where trans-capillary conductivity is high and interstitial conductivity is low, IFP becomes almost equal to the capillary pressure. This is thought to occur in many pre-clinical tumour models [
55] and possibly also in human malignancies. Tumours with high flow resistance due to unregulated angiogenesis, high cell density or tumour growth in a confined space with vascular compression would be expected to have both high capillary and high interstitial pressure values. The important role of the vasculature in determining IFP is supported by studies of anti-vascular drugs that have shown reductions in tumour IFP with vascular regression or normalisation [
11,
39,
56].
In our study, tumours grown orthotopically consistently demonstrated IFP values lower than those grown in the i/m ectopic site (Figure
1). This is in contrast to results described by Brekken et al. (2000) [
30] who demonstrated higher IFP values in a human osteosarcoma line grown orthotopically as compared to s/c. This difference highlights the interaction between the tumour and the surrounding normal tissue and its influence on IFP. For example, tumours growing in bone would be expected to have high blood flow resistance and high capillary pressure because of progressive vascular compression as the cell mass increases in a confined, noncompliant space. In addition, there may be fewer normal lymphatic vessels in close proximity to the tumour, effectively reducing interstitial conductivity and driving up IFP until it equals the capillary pressure. We also demonstrated that tumours grown i/m have higher IFP values than those grown s/c, in the cervix, or in the mammary fat pads, probably reflecting similar pathophysiologic mechanisms. Overall, our results suggest that site-specific differences in vascular development and remodelling, and consequently flow resistance and pressure, as well as site-driven differences in the extracellular matrix, lead to systematic differences in IFP values. We speculate that heterogeneity in IFP values among individual tumours of the same type growing at the same site reflect the stochastic nature of unregulated angiogenesis in tumours and the random nature of the resultant vasculature.
In further support of these concepts, the MMTV-PyMT transgenic mammary gland tumours growing orthotopically in the mammary fat pad showed IFP values that were comparable to the spontaneously arising tumours. Also, there were no differences when tumours were transplanted into FVB or SCID hosts. However, when grown i/m, the IFP in the MMTV-PYMT tumours was significantly increased. Since the data showed no relationship between donor IFP and recipient IFP this suggests a site-specific effect. Similar findings in the cervical tumour models suggest individual tumour development irrespective of any tumour microenvironment-induced molecular interactions that may have existed in the original donor tumour.
It is interesting that although the tumour vasculature plays a vital role in the development of both tumour IFP and tumour hypoxia, there was no correlation between these two parameters, an effect that has also been observed both experimentally [
26,
57] and in the clinic [
7]. In addition, there was no relationship between vascular area and IFP, with similar values being observed in all models. Furthermore, in our high frequency ultrasound speckle variance analysis of blood flow in a small number of tumours, vascular perfusion was also comparable between ME180 tumours growing in the cervix versus i/m.
There is evidence in the literature that high primary tumour IFP is associated with a higher incidence of metastatic disease [
7,
20,
23], although Rofstad et al. (2002) suggest that it is neither necessary nor sufficient. In the present study we found no indication of a relationship between high tumour IFP and metastatic disease in any of the models in any of the sites. A controlled in vitro experiment designed to test the influence of pressure on the metastatic ability of cancer cells without the potentially confounding influence of other microenvironmental factors also showed no effect. These data are consistent with clinical data where, although a high pre-treatment IFP in cervical carcinoma patients was associated with a high risk of distant metastases following treatment with radiotherapy, there was no correlation between IFP and metastatic disease at the time of diagnosis [
7]. It is possible that tumour IFP may be more a marker of treatment response than have a direct causative effect on metastatic spread.
Conclusion
The data presented provide, for the first time, a specific analysis of tumour IFP across a selection of different tumour models encompassing human, murine transplanted and spontaneous tumours, growing in different sites. We demonstrate inherent inter-tumour heterogeneity, and a lack of correlation both between tumour donors and recipients and between different tumours growing within the same animal. Our findings indicate systematic differences in IFP as a function of tumour type consistent with predetermined inherent genetic differences that influence vascular development and the composition and organisation of the interstitium. There are also systematic differences as a function of growth site, presumably reflecting the interaction between the tumour and the surrounding host normal tissue. Nevertheless, the heterogeneity of IFP in individual tumours growing under similar conditions suggests that IFP is probably influenced to a large extent by the stochastic nature of vascular development and remodelling during tumour growth. In this context it is surprising that an analysis of various phenotypic parameters failed to show any correlation between IFP and tumour hypoxia, tumour proliferation, blood or lymphatic vascular density. Whether the lack of correlation in these parameters is indicative of no involvement in the development of tumour IFP remains uncertain, however, since the parameters are interdependent and may act together to elevate IFP. It is clear that factors influencing tumour IFP are complex and that further study is required to reveal the mechanisms that elevate IFP and its importance in therapeutic response either as a stand-alone factor or in combination with other factors.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SJL participated in the design of the study, performed the animal experiments, analysed the IFP data and histology slides, and wrote the manuscript. TMKK was responsible for the breeding of the MMTV-PyMT transgenic mice, aided in the development of the mammary tumour surgical transplant technique, and contributed to the manuscript. AB, aided by VXY, was responsible for the analysis of the Ultrasound speckle variance data, and contributed to the manuscript. MM and RPH participated in the design of the study and were involved in writing the manuscript. All authors read and approved the final manuscript.