Hyaluronan, fluid pressure, and stromal resistance in pancreas cancer

1Pancreatic ductal adenocarcinoma (PDA) is an aggressive disease of the exocrine pancreas, marked by a 5-year survival rate of <5% and a median survival of ∼6 months, making it the fourth leading cause of cancer-related deaths in the United States. The anatomy and biology of PDA conspire to elude detection and resist eradication. An absence of localising symptoms, together with a lack of biomarkers and screening methods for early detection, irreparably delay diagnosis. Indeed, 80–90% of patients are diagnosed too late to permit surgical intervention, which itself can prolong life but is seldom curative, due to almost inevitable disease recurrence and/or metastasis. For these patients, treatment is confined to chemical and/or radiotherapies, which historically have provided limited benefit at best.

2The deoxycytidine analogue, gemcitabine, has represented the standard-of-care for disseminated PDA, although this may be slowly changing. Gemcitabine monotherapy can improve quality of life in a subset of patients and extend survival modestly. The poor response of PDA to chemotherapy was long thought to result principally from cellular mechanisms of intrinsic resistance in the tumour epithelium; it is now understood, however, to be due at least in part to an inability of even small molecule therapeutics to enter and perfuse the tumour bed. Intriguingly, sustained exposure of PDA to chemotherapies with long circulating half-lives can increase intratumoral drug levels and improve response. For example, nab-paclitaxel, an albumin-coated nanoparticle formulation of taxol with a >10 h half-life, in combination with gemcitabine improved median survival (mOS) substantially compared with historical controls (mOS=12.2 months; ). In a large phase III trial of patients with metastatic disease, the multi-drug regimen FOLFIRINOX – which combines bolus doses of oxaliplatin, irinotecan, leucovorin, and 5-fluorouracil, followed by a prolonged 46h continuous infusion of fluorouracil – improved survival compared with gemcitabine alone (mOS=11.1 versus 6.8 months, respectively; ). An admittedly substantial side-effect profile may limit widespread use of FOLFIRINOX and, despite representing potentially significant advances over the prior collective experience, neither regimen is curative.

3Our expanding knowledge of the genetic repertoire of pancreas cancers promises to guide development of new, more selective strategies. The basic molecular profile of PDA is known. Activating mutations in the KRAS proto-oncogene occur early in disease progression and are found in >90% of invasive PDAs. Subsequent mutations in the CDKN2A, TP53, and/or DPC4 (SMAD4) tumour suppressor genes are also common (>90%, ∼75%, and ∼50% respectively; reviewed in ). Genetic engineering in mice has demonstrated that these mutations, and the order in which they arise, drive, and shape disease progression, although efforts to target them directly to treat pancreas cancer have yet to achieve fruition. A number of signalling pathways are also aberrantly regulated in PDA including EGFR, ERBB2, COX-2, SHH, MMPs, and NOTCH, and which suggest additional potential targets. To date, only inhibition of the epidermal growth factor receptor (EGFR) has been clinically approved, and even then provides only incremental improvement in survival.

4In this review, we discuss the concept of ‘stromal resistance’ in PDA, namely the barriers to effective treatment imposed by the complex and dynamic tumour microenvironment (TME). In particular, we highlight physical mechanisms of drug resistance in PDA and recent efforts to target stromal elements in order to improve delivery of small molecule therapeutics. We discuss further how studies in genetically engineered mouse (GEM) models of PDA have yielded unexpected insights compared with more traditional xenograft and allograft models and may help to explain the previous divide between preclinical promise and clinical reality.

5A dynamic fibroinflammatory, or desmoplastic, response is a hallmark of PDA and essentially pathognomonic for the disease. This desmoplasia evolves during disease progression and includes stromal fibroblasts, immune cells, and excessive deposition of a complex extracellular matrix (ECM) (Figure 1; ). Activated stromal fibroblasts, or myofibroblasts, produce matrix constituents that alter the physical structure of the developing cancer ‘organ’ and also modulate tumour cell behaviour through direct binding interactions with surface receptors. The immune reaction in PDA consists largely of immunosuppressive and pro-tumourigenic elements that infiltrate at the earliest stages of preinvasive disease, with scant evidence of effector immunity. The ECM of this desmoplastic response contains significant levels of fibrous collagen together with a complex mixture of proteoglycans and glycosaminoglycans. In addition to roles in maintaining tissue integrity and providing signalling scaffolds, dense collagen and microfibrillary matrices and fluid-trapping mucopolysaccharides have been shown to impede diffusion (exchange driven by concentration gradients) and convection (pressure-driven fluid flow that drags molecules) through porous materials such as the stromal compartment in carcinomas. An elevated IFP that limits convection and a fibrotic ECM that compromises diffusion combine to impede solute flux across a perfused vessel and through a tumour bed. Despite the presence of such substantial barriers to treatment, the altered fluid mechanics of the desmoplastic stroma in PDA represents an underappreciated and relatively unexplored mechanism of disease resistance.

6The delivery and distribution of small molecules across a tissue are governed by well-defined principles of solute transport involving passive diffusive and convective fluxes (reviewed in ). In general, the distribution of small molecules is primarily diffusion limited, while larger molecules rely also, or even preferentially, on convective flow. Diffusion is driven by concentration gradients between the intravascular and interstitial compartments, and convection by pressure gradients composed of hydrostatic and oncotic components. The balance between hydrostatic and oncotic pressures in tissues and vessels determines the net flow of fluid.

7In contradistinction to most solid tumours, PDAs are hypoperfused; indeed, in standard clinical imaging modalities, they are identified as regions that take up less injected contrast material than the normal surrounding tissues. Neuroendocrine tumours of the pancreas, on the other hand, appear bright on contrast-enhanced CT or MRI, reflecting their hypervascular nature. These radiographic signatures of blood flow provide tell-tale clues distinguishing these dramatically different tumour types. Perhaps not surprisingly, therefore, and despite a profound hypoxia and elevated levels of pro-angiogenic factors, PDAs have a paucity of blood vessels. Even worse, the vast majority of existing vessels in PDA are not functional. As a result, relatively few vessels are effectively perfused with chemotherapy, and those that are then face further barriers to transvascular distribution of drug into the stroma imposed by the ECM. Inhibition of paracrine sonic hedgehog signalling (Shh) in a GEM model of PDA depleted stromal fibroblasts, stimulated angiogenesis, and produced a salutary increase in cytotoxic drug delivery. Although these changes in tumour perfusion were short-lived, they provide proof-of-principle of a primary mechanism of drug resistance in PDA. However, the lack of significant functional perfusion in PDA results not only from a sparse vasculature but also from a profound degree of vascular collapse. In fact, ∼75% of the vessels in PDA appear to be collapsed. The observations raise immediate questions regarding the metabolic regulation in PDA cells and also the mechanisms underlying this vascular collapse.

8A collapsed vasculature suggests the presence of interstitial pressures exceeding the combined intravascular hydrostatic and oncotic pressures and elastic forces of the vascular wall. As noted, a large body of work in a number of experimental model systems has suggested that altered fluid mechanics can limit the efficacy of systemic therapies (reviewed in ). Fewer studies have been performed in autochthonous tumours and none before and after systemic interventions. To perform interstitial fluid pressure measurements in normal tissues and autochthonous PDA, we used a Millar Mikro-Tip pressure catheter transducer connected to a control unit and data acquisition system. In normal tissues, IFP ranged from ∼8 mm Hg for the head of the pancreas, to 0.1 mm Hg for the liver, to −2.0 mm Hg in muscle, consistent with values measured in numerous other studies using a variety of methodologies (for examples, see and; reviewed in ). These tissue IFPs are significantly below the typical intravascular pressures of the arterioles (40–80 mm Hg) and capillaries (15–40 mm Hg) that feed them, as originally hypothesised by. In contrast, IFPs in autochthonous PDAs were dramatically elevated, ranging from 75 to as high as 130 mm Hg, rivalling mean arterial pressures.

9HA is a large linear, negatively charged, and soluble macromolecule comprising N-acetyl glucosamine and glucuronic acid in alternating β-1,3 and β-1,4 linkages that figures prominently in the architecture, integrity, and malleability of tissues, particularly in dynamic processes such as embryogenesis and oncogenesis. HA also binds to the surface receptors, CD44 and RHAMM, activating signalling pathways that can promote cell survival, proliferation, adhesion, migration, and invasion. Its ability to imbibe and retain water contributes to the gel-like interstitia of many tissues and these same viscoelastic properties underlie the use of HA in a variety of cosmetic and clinical applications. These properties together with the extraordinary abundance of HA beginning from the earliest stages of disease suggested a potentially critical role in establishing and defining the unique TME of PDA.

10Homeostatic HA levels in mammalian tissues are maintained by the selective expression and activity of three different synthases, HAS1–3, and six different hyaluronidases. Originally identified as the ‘spreading factor’ from bovine testes, purified hyaluronidase was later shown to represent a fragment of PH20, a GPI-anchored enzyme found on the acrosomal membrane of mammalian sperm that assists in fertilisation. The purified testicular extract has been used historically to promote resorption of fluid and solutes administered subcutaneously. The unexpected palliation of disease in a multiple myeloma patient who received the extract for such an indication spurred investigations into the ability of the enzyme to enhance the efficacy of chemotherapies. A number of cell autonomous and non-cell autonomous mechanisms have been proposed and continue to be studied (reviewed in and ).

11The purified and recombinant forms of native hyaluronidase possess extremely short circulatory half-lives (t1/2<3 min) which, together with the development of hypersensitivity reactions, have precluded their use as systemic agents. Chemical conjugation of recombinant human PH20 with polyethylene glycol preserves enzymatic activity and extends the half-life to >10 h. Systemic administration of this agent can deplete HA from implanted tumour xenografts in immunocompromised mice and decrease IFP. To directly address whether systemic administration of this enzyme could deplete HA from an autochthonous tumour and whether this would ameliorate the elevated IFP, we treated genetically engineered KrasLSL-G12D/+;Trp53LSL-R172H/+;Cre (KPC) mice with established invasive PDA with intravenous injections of PEGPH20. Significant intratumoral depletion of HA and reduction of IFP occurred within 2 h and peaked at 24 h. The reduction in IFP correspondingly resulted in a dramatic increase in patent tumour vessels without affecting overall vessel number, and ready penetration of chemotherapies across the tumour bed. Interestingly, although stromal HA is similarly depleted from a number of normal tissues in the animals, including heart, lungs, intestine, and liver, no overt untoward effects on organ function and overall health were observed.

12When combined with the conventional cytotoxic, gemcitabine, in a randomised, placebo-controlled preclinical trial in KPC mice, PEGPH20 significantly increased objective response rate, decreased metastatic tumour burden, and prolonged median survival. Increased apoptosis was observed in both the pancreatic stellate cell and tumour epithelial cell compartments. After only a few weeks of combination therapy, the effects on HA levels and vascular perfusion persisted even if treatment ceased. Thus, the PDA stroma can be permanently remodelled in this manner, in principle permitting subsequent treatment with a wide variety of sequential and combination regimens. That metastatic tumour burden was decreased and not increased in the setting of increasing vascular access of cancer cells is also reassuring. A similar survival benefit with this combination therapy was independently observed.

13The fluid pressures we observed in autochthonous PDAs are significantly larger than those reported in a variety of in vitro, xenograft and allograft tumour model systems (reviewed in ). Of course, an invasive carcinoma developing in situ in the native organ from spontaneously evolving precursor lesions is very different from the aforementioned model systems in several critical ways (see below). The apparatus used to measure fluid pressure has also varied. A number of systems have been used to measure fluid and tissue pressures over the last 50 years, including the needle, modified wick-in-needle, micropipette, and implanted perforated capsule procedures. Each of these systems has virtues and limitations, not the least of which is the need for sufficient skill to reproducibly fashion the custom glass needles used for the measurements (reviewed in ). Indeed, the micropuncture glass capillary method is known to yield more reliable estimates of fluid pressures than the wick-in-needle because it produces less trauma and requires smaller volumes to record IFP. The polyurethane ultraminiature pressure transducer used in our experiments was developed, in part, to overcome some of the limitations encountered in previous methodologies, and also to both simplify and increase the accuracy of the method. We also took a number of measures to ensure the fidelity of the measurements and that solid stresses (SSs) were not being measured in addition to fluid pressure. First, the instrument was calibrated to 0, 25 and 100 mm Hg before every experiment and confirmed at its conclusion. The catheter is also designed so that the sensor is situated in a side port a few mm from the end of the probe, and not at its very tip, to minimise any potential trauma or artifact during placement. In addition, we did not use the pressure catheter transducer directly to enter the tumour, but rather placed it in the track left behind after insertion and withdrawal of a 25-gauge needle. Third, for every measurement made, we ensured that the catheter could be freely and repeatedly inserted into and withdrawn from the preformed needle track several times without resistance and without any shearing of tissue (i.e., there was no adherence of tissue to the side-port sensor). Finally, we confirmed that IFP in the normal head of the pancreas dropped from a mean of 8 mm Hg to zero or even slightly negative pressure (−2.7 to 0.8 mm Hg) within minutes of the cessation of cardiac function and tissue blood flow, as expected. In normal pancreas tissue, the interstitial fluid space is in equilibrium with and directly chronicles mean vascular pressures (MVP). For all of these reasons, we concluded that the piezoelectric catheter transducer was not measuring SS. In fact, the theoretical possibility of solid pressure artifacts with this apparatus was specifically examined by who found no evidence for such. went on later to describe the Millar SPC320 catheter as a ‘gold standard reference’.

14Our observations and conclusions differ from that derived from an extensive body of work employing novel and highly creative in vitro and tumour explant systems including ‘tissue-isolated tumours’ comprising an externalised ovarian fat pad injected with a rat mammary adenocarcinoma ‘tumour slurry’; ‘tumour spheroids’ comprising cell lines seeded into agarose gels; and a variety of subcutaneous tumour cell line transplant systems (reviewed in ). Shared properties among these experimental systems have allowed a sophisticated working model of tumour physiology to be proposed (Figure 2), characterised by: (1) hypervascularity; (2) markedly increased vascular permeability and hydraulic conductivity (i.e., ‘leaky’ vessels); and (3) ‘open’ systems in which the interstitial fluid space is in equilibrium with the intravascular space and fluid can ‘ooze out of the periphery’. As Jain and colleagues have elegantly demonstrated, elevated IFP in these contexts is due primarily to leaky vessels and dysfunctional lymphatics and can rise as high as microvascular pressure (MVP). Moreover, because the intra- and extravascular fluid spaces are in equilibrium, changes in MVP are rapidly reflected in changes in IFP, which both plunge towards zero when the heart is stopped. The investigators concluded that IFP cannot collapse vessels and that, instead, vascular collapse is caused by SS from proliferating cancer cells. From these observations, it seems reasonable to suggest that ‘collapsed lymphatic and blood vessels are known to contribute to elevated IFP and not the other way around’.

15Unexpected challenges to this framework arise when confronting autochthonous PDAs. Properties of these tumours subvert the working assumptions derived from the aforementioned studies in many important ways (Figure 2). First, both human and murine PDAs (; Provenzano et al, 2013) are profoundly hypovascular, in contradistinction to the systems described above. Second, the vasculature is characterised by widespread collapse. Third, the tumour vessels in PDA appear to be functionally and ultrastructurally intact, have preserved interendothelial junctions, and are notably lacking in the fenestrae that characterise ‘leaky’ systems. Further, despite the extremely elevated interstitial pressures, fluid is not ‘squeezed’ out from the tumour as would be expected if the system was in open communication with its environment, and small molecules are unable to effectively penetrate the tumour when administered either intravenously or intraperitoneally (i.e., directly bathing the tumour). Applying perhaps the most stringent criterion, intratumoral IFP does not drop after cessation of cardiac function in an animal with autochthonous PDA, even up to 40 min later, confirming that the intravascular and interstitial fluid spaces are largely disconnected. How PDA cells survive such a profoundly hostile microenvironment is worthy of study in its own right but likely includes a reliance on alternate metabolic pathways, transient microregional variations in pressures and blood flow, and autophagy, among other possibilities. Nevertheless, to a first approximation, PDA would appear to behave as a ‘closed’ rather than open system, a reality that has profound implications for the generation and maintenance of elevated fluid pressures, as well as the relationship between IFP and SS. These considerations become especially germane when considering therapeutic strategies. Xenografts from both human and mouse respond readily to gemcitabine, in stark contrast to their respective autochthonous counterparts. Agents that inhibit or normalise angiogenesis are not effective in treating PDA in either humans or mice, despite their proven utility in typically hypervascular pancreatic neuroendocrine tumours and other malignancies. It is important to note that SS is, by definition, exerted only at points of direct contact, whereas a fluid pressure transmits force in all vectorial directions. Thus, it is unlikely that a sufficiently large vascular surface area is in direct contact with solid tissue elements (primarily proliferating cells) and exerting a high enough pressure to collapse the majority of vessels in the tumour bed. Indeed, PDAs are paucicellular, particularly with regard to their epithelial cell content, and the mean distances between tumour cell and vessel are relatively high. Thus, SS may contribute to vascular collapse in PDA, but IFP appears to predominate. In the end, it is likely that more than one working model of tumour physiology will be required to address and inform interventions across the spectrum of human solid malignancies.

16Finally, how then to explain the profoundly elevated fluid pressures and vascular collapse in PDA? Do fluid pressures collapse vessels or do collapsed vessels lead to increased fluid pressures? The specific features of the stromal compartment and the intact ultrastructure of the vessels in autochthonous PDA provide the requisite properties to understand the phenomenon. The large extended HA polymers can imbibe significant amounts of water and organise it into a hydrogel, creating an immobilised fluid phase in equilibrium with freely mobile water. The gel, nevertheless, also possesses an elastic modulus, a property of solids. Thus, hydrated HA is viscoelastic and several types of pressure are associated with the HA hydrogel including: fluid pressure; a surface pressure; and a colloid osmotic pressure, contributed to by the Donnan equilibrium effect, and which itself can rise significantly and non-linearly in proportion to increasing HA concentration to become quite substantial. Tethered collagen fibrillary and non-collagenous mirofibrillary structures with varying degrees of cross-linkage compound this complexity, given their ability to create microcompartments within the overall tumour architecture, while providing a scaffold for active cell contractility to constrain an expanding HA hydrogel (Figure 2). In open systems, SS and IFP can be considered additive to estimate a total tissue pressure (TTP). However, in the relatively closed system that more closely approximates an autochthonous PDA, the TTP is the IFP, which is transmitted through the solid matrix structure as well. Indeed, this is not unlike how hydraulic systems operate and generate force. Acute angle-closure glaucoma provides another apt example of the physiologic consequences of elevating fluid pressures in a confined space with free fluid and a hydrogel. Accumulation of aqueous humour (free fluid) in the posterior chamber of the eye acts in concert with the vitreous body – a hydrogel comprising HA, collagen, proteoglycans and both free and bound water – to impinge upon ocular blood flow and the optic nerve, threatening the eye.

17The precise mechanics and interrelationships among the free fluid, solid, and viscoelastic phases in PDA aside, it is important not to lose sight of the most relevant finding from these studies: enzymatic degradation of HA through systemic administration of a clinical-grade reagent renders PDA more uniformly permeable to chemotherapeutics, revealing an unexpected and profound sensitivity to agents that have previously been shown to be ineffective. This finding permits the rational investigation of an entire array of small molecule, antibody, and cell-based therapies for PDA, in association with HA degradation. It has the potential, currently under active clinical investigation (NCT01453153), to influence the fundamental strategy in treating patients with this dreaded disease.

18The complex cellular, molecular, and mechanical features of the desmoplastic response in PDA presents challenges to effective treatment, as well potential vulnerabilities to exploit. In creating a drug-free sanctuary, the tumour also presumably remains sensitive, at least initially, to the chemotherapies it has been shielded from rather than selected to resist. Although intrinsic mechanisms of disease resistance will no doubt emerge, breaching the sanctuary with a systemically deliverable enzyme offers the prospect of unprecedented, sustained vascular access to the PDA primary tumour bed and metastases. This will enable not only rigorous evaluation of new treatment strategies, but also exploration of distinct sequencing and schedules of various treatment modalities in the attempt to remain one step ahead of the tumour. Indeed, re-examining the long list of clinically approved agents previously thought to be ineffective in PDA would be an excellent place to start. Additional mechanisms to reduce IFP should also be explored including targeting other ECM components (such as collagen, versican, or decorin) or inhibiting contractile forces generated by fibroblasts. Restoring physiologic blood flow, oxygen, and nutrients to a pancreas cancer will undoubtedly alter its metabolism, and these changes may present additional therapeutic targets and also new mechanisms of resistance. However, the idea that the complex and impervious pancreas cancer ‘organ’ can be perturbed dramatically is itself encouraging. It cannot be long before this literal and figurative entry into pancreas cancer pathophysiology results in a more effective therapeutic armamentarium.